A BEHAVIORAL STAGE-MATCHED INTERVENTION FOR IMPROVING HYPERTENSION CONTROL.S. Natarajan1; E. Santa Ana2; P. Nietert2; K.M. Magruder2. 1VA New York Harbor Healthcare System and New York University, New York, NY; 2Medical University of South Carolina, Charleston, SC. (Tracking ID #117389)

BACKGROUND:  The Transtheoretical model for health behavior change is a useful framework for facilitating behavior change by tailoring interventions to a person's readiness for change. Among individuals with hypertension, this study evaluated the efficacy of a behavioral stage-matched telephone-delivered counseling intervention for: 1) increasing adherence to exercise, diet, and medications, and 2) improving blood pressure (BP) control.

METHODS:  Baseline stage of change (precontemplation, contemplation, preparation, action, maintenance and termination) for exercise, diet, and medications as well as systolic and diastolic BP were assessed in 120 veterans with hypertension attending a routine health care appointment. Participants were randomized 1:1:1 to receive a stage-matched intervention (SMI), a health education intervention (HEI), or usual care (UC). Interventions were delivered by monthly telephone calls for six months to participants in the SMI and HEI groups. Stage of change (for exercise, diet, and medication) and BP was evaluated at 6-month follow-up using multivariate analysis of covariance (MANCOVA) and logistic regression.

RESULTS:  There were no significant baseline differences between groups. At 6-month follow-up, a significantly greater proportion of participants in SMI were in later (action or maintenance) stages of change for exercise (63.9%) and diet (72.7%) adherence compared to participants in HEI (exercise: 25%, diet: 15.2%) or UC (exercise: 11.1%, diet: 12.1%). At 6 months SMI was a significant independent predictor for being in the later stages of change (action/maintenance) for exercise (P < .0001) and diet adherence (P < .0001). All but 3 participants (96.6%) reported being fully medication adherent. At 6-month follow-up, MANCOVA adjusting for baseline BP indicated that systolic BP was significantly lower in SMI (129 mm Hg, standard deviation [SD] = 13.9) than HEI (139.2 mm Hg, SD = 18.2, P = .032) and UC (140.2 mm Hg, SD = 19.5, P = .015). Diastolic BP was significantly lower in SMI (74.7 mm Hg, SD = 8.9) than UC (81.3 mm Hg, SD = 14, P = .017), but not HEI (79 mm Hg, SD = 9.1, P = .098).

CONCLUSION:  In this randomized study, a Transtheoretical stage of change-matched intervention led to improved adherence for exercise, diet, and taking medications, and lowered BP among veterans with hypertension. This study provides preliminary evidence that stage-matched interventions hold promise for improving hypertension control among veterans.

A COMPARISON OF DIABETES CARE QUALITY IN VA AND COMMERCIAL MANAGED CARE: THE TRIAD STUDY.E.A. Kerr1; R. Gerzoff2; S.L. Krein1; J.V. Selby3; J.D. Piette1; J. Curb4; W.H. Herman5; D.G. Marrero6; V. Narayan2; M. Safford7; C.M. Mangione8. 1VA Ann Arbor Healthcare System and University of Michigan, Ann Arbor, MI; 2Centers for Disease Control and Prevention (CDC), Atlanta, GA; 3Kaiser Permanente Division of Research, Oakland, CA; 4Pacific Health Research Institute, Honolulu, HI; 5University of Michigan, Ann Arbor, MI; 6Indiana University Purdue University Indianapolis, Indianapolis, IN; 7University of Medicine and Dentistry of New Jersey, Newark, NJ; 8University of California, Los Angeles, Los Angeles, CA. (Tracking ID #116059)

BACKGROUND:  Studies have shown improved quality of care in the Department of Veterans Affairs (VA) relative to Medicare. No studies, however, have compared care in VA to that delivered in commercial managed care (CMC) organizations, nor have they focused in depth on chronic, outpatient conditions. We sought to compare the quality of diabetes care between patients in VA and those enrolled in CMC organizations using equivalent and pre-specified sampling and measurement methods.

METHODS:  We enrolled patients with diabetes from 5 VA medical centers (N = 1285) and 8 CMC organizations (N = 6920) in 5 matched geographic regions. We compared scores on 10 identically specified quality measures (e.g., annual hemoglobin A1c [A1c], annual low density lipoprotein cholesterol [LDL]) and 4 satisfaction measures (e.g., satisfaction with quality of diabetes care), adjusted for patient demographic and health characteristics.

RESULTS:  VA patients had better scores than CMC patients on all process measures, ranging from a 10 percentage point difference on performance of an annual A1c (93% versus 83%; P = .006) to a 25 percentage point difference on aspirin use counseling (74% vs. 49%; P < .001). There was no difference in blood pressure control between VA and CMC patients, but VA patients had better control of LDL and A1c (86% vs. 72% for LDL < 130 mg/dl, P = .002; 92% vs. 80% for A1c < 9.5%, P =.006). Satisfaction was similar between the two cohorts.

CONCLUSION:  We found that diabetes processes of care and 2 of 3 intermediate outcomes were substantially better for VA study patients than for CMC patients. Commercial plans may benefit from a better understanding of VA quality improvement programs, especially regarding enhancing LDL and A1c control. However, there was room for improvement in blood pressure control in both VA and CMC.

A COMPARISON OF THE QUALITY OF MEDICAL CARE MEASURED BY INTERVIEW AND MEDICAL RECORD.J.T. Chang1; C.H. MacLean2; C.P. Roth2; N.S. Wenger1. 1University of California, Los Angeles, Los Angeles, CA; 2RAND, Santa Monica, CA. (Tracking ID #117145)

BACKGROUND:  Chart-based measurement has been considered the gold standard for many measures used to estimate quality of care. However, patients can report on many aspects of their care, including some that may be poorly documented in the medical record, and interview can be a cost-effective data collection method. Using a set of process measures developed for vulnerable older adults, we compared performance scores obtained using data from patient interviews and medical records.

METHODS:  The Assessing Care of Vulnerable Elders (ACOVE) quality indicators (QIs), a set of 236 explicit process measures covering 22 conditions, were used to assess care in a random sample of vulnerable older adults from two senior managed care plans. Data were available from both interview and medical records for 245 patients. 60 QIs were measured with data available from both sources. We performed a priori classification of the “gold standard” data source for the 60 QIs as follows: patient interview would be the better source for information for QIs measuring communication and nonprescription medications (16 QIs); the medical record would be a better source for QIs measuring routine medical procedures (14 QIs); the remaining 30 QIs had no preferred data source. Performance scores were computed as the percentage of indicated care processes that patients received.

RESULTS:  Performance assessed by the 30 quality indicators without a preferred data source scored the same by interview 66% (95% CI, 64% to 68%) and medical record 65% (95% CI, 64% to 67%). Quality of care measurement by the 16 QIs for which interview was the a priori preference had a much higher mean quality score by interview 66% (95% CI, 63% to 68%) than medical record 30% (95% CI, 27% to 32%). Examples of such QIs include education for a newly prescribed medication (99% interview v. 15% chart, P < .001), prescribing a bowel regimen with chronic opioid treatment (83% interview v. 0% chart, P = .004), and aspirin therapy in patients with diabetes (73% interview v. 49% chart, P = .019).

CONCLUSION:  Quality measurement using patient interviews mirrors medical record findings for many care processes. Excluding interview data from a chart-based quality assessment system will significantly miss information on communication processes and nonprescription medication use.

A COMPARISON OF TWO HOSPITALIST MODELS WITH TRADITIONAL CARE IN A COMMUNITY TEACHING HOSPITAL.M. Halasyamani1; P. Valenstein1; M. Friedlander1; M. Cowen1. 1Saint Joseph Mercy Hospital, Ann Arbor, MI. (Tracking ID #101659)

BACKGROUND:  Many studies have documented significant length of stay and cost savings when hospitalist care is compared to usual care. However, there have not been any studies that have compared the outcomes of more than one hospitalist model in a single site. We report the resource utilization and clinical outcomes of two hospitalist models compared with a traditional care model within a single site.

METHODS:  This is a retrospective cohort study conducted between July 2001 and June 2002 in a tertiary care community-based teaching hospital. Patients: 9,984 inpatients cared for by community physicians, academic hospitalists, or private hospitalists. Main Outcome Measures: Length of stay, variable costs, 30-day readmission rates, in-hospital and 30-day mortality, all adjusted for potentially confounding factors including age, diagnosis, insurance, severity of illness, and admission from the emergency department.

RESULTS:  Patients of academic hospitalists had significantly lower average length of stay (P = .0005) and odds of 30-day mortality (OR = 0.77; 95% CI 0.77, 0.92), compared with traditional community physicians. Private hospitalists did not differ from community physicians in average length of stay or 30-day mortality. Academic hospitalists did not differ from community physicians in variable direct costs (P = .08); however, community physicians had significantly lower variable direct costs compared with private hospitalists (P = .04). In-hospital mortality or 30-day readmission did not differ significantly among the three physician groups

CONCLUSION:  Different hospitalist models may have distinct effects on patient outcomes. Future studies should examine the specific organizational characteristics of hospitalists that contribute to improved patient care and resource utilization.

A COMPUTER-BASED DECISION AID IS AN EFFECTIVE TOOL FOR EDUCATING WOMEN SEEKING INFORMATION ABOUT BRCA1/2 GENETIC TESTING.M.J. Green1; S.K. Peterson2; M. Wagner Baker3; G.R. Harper4; L.C. Friedman5; W.S. Rubinstein6; D.T. Mauger1. 1Penn State College of Medicine, Hershey, PA; 2University of Texas M.D. Anderson Cancer Center, Houston, TX; 3Penn State Cancer Institute, Hershey, PA; 4Penn State Cancer Institute, Allentown, PA; 5Baylor College of Medicine, Houston, TX; 6Evanston Northwestern Healthcare, Evanston, IL. (Tracking ID #116010)

BACKGROUND:  The demand for cancer genetic testing is increasing, and primary care practitioners have limited time and resources for educating patients about hereditary cancer risk. We conducted a clinical trial evaluating the effectiveness of an interactive, computer-based decision aid (“Breast Cancer Risk and Genetic Testing”) to educate women about BRCA1/2 genetic testing.

METHODS:  This was a randomized, controlled trial of 211 women with family or personal histories of breast cancer enrolled at six study sites from 2000–2002. Participants received standard one-on-one genetic counseling (Counselor Group, n = 105) or education by a computer-based decision aid (Computer Group, n = 106) followed by genetic counseling. The computer program was accessed via personal computer and participants spent 45–60 minutes using it. Outcomes included: knowledge, risk perception, genetic testing intentions, state anxiety and length of the counseling sessions. Outcomes were compared pre- and post-intervention in each group, and were stratified by low (<10%) vs. high (≥10%) risk of carrying a mutation.

RESULTS:  Both groups had comparable demographic characteristics, experience with computers, medical literacy, and baseline knowledge of breast cancer and genetic testing. Mean age was 44 years, 56% completed college, and 93% were white. Mean knowledge scores (range 0–100% correct) increased in the Counselor Group from 53% pre-counseling to 82% post-counseling (P < .0001), and in the Computer Group from 55% to 91% (P < .0001). Mean knowledge scores post-computer use were higher compared with post-counseling alone (P = .01) among low-risk women. Neither frequency of computer use in daily life nor confidence in computer skills were related to changes in knowledge scores. Among low risk women in both the Counseling and Computer Groups, there was a significant decline in participants’ post-intervention estimates of their absolute risk of developing breast cancer, as well as their intention to undergo genetic testing. Further, neither counseling nor computer use resulted in increased anxiety. Finally, for low risk women, counseling sessions were significantly shorter (P = .03) when preceded by computer use (77 minutes) compared to counseling without prior computer use (89 minutes).

CONCLUSION:  The computer-based decision aid was more effective than standard genetic counseling for increasing knowledge of breast cancer and genetic testing among women at low risk for carrying a BRCA1/2 mutation. The program's use among low risk women was associated with shorter subsequent counseling sessions and a diminished perception of breast cancer risk, along with a decreased likelihood of pursuing genetic testing. Primary care practitioners seeking to educate women who are concerned about their inherited risk for breast cancer may find this program to be a useful educational tool.

A LOW-CARBOHYDRATE, KETOGENIC DIET FOR TYPE 2 DIABETES MELLITUS.W.S. Yancy1; M.E. Foy1; E.C. Westman1. 1Duke University, Durham, NC. (Tracking ID #117291)

BACKGROUND:  Low-carbohydrate diets lead to weight loss, and in one study led to improved glycemic control in diabetics. The objective of this pilot study was to examine the safety and effectiveness for glycemic control of the low-carbohydrate, ketogenic diet (LCKD) in patients with type 2 diabetes.

METHODS:  In a Veterans’ Affairs Medical Center outpatient clinic, we recruited overweight (body mass index [BMI] >25 kg/m2) subjects taking oral hypoglycemic agents and/or insulin, or having a hemoglobin A1c >6.0% without medication. Subjects received LCKD counseling, with an initial goal of <20 g carbohydrate/day, and were encouraged to take a multivitamin and drink 6–8 glasses of fluids daily. Diabetes medication dosages were reduced by approximately 50% at diet initiation; diuretic medications were reduced by 50% or discontinued, and subsequently reinstituted if needed. Subjects returned every other week for 16 weeks for body and vital sign measurements, counseling, and further medication adjustment if needed. Fasting blood and 24-hour urine tests were obtained at weeks 0, 8, and 16. Serum electrolytes and kidney function tests were monitored additionally at weeks 2 and 12.

RESULTS:  Nineteen of the 25 subjects who were enrolled completed the study. Eighteen subjects were men; 13 were White, 5 were Black. The mean [+ SD] age was 56 ± 8 years; mean BMI was 39 ± 6 kg/m2 (range 28-51 kg/m2). From baseline to week 16, hemoglobin A1c decreased by 15% from 7.4 ± 1.5% to 6.3 ± 1.1% (P < .001) while diabetes medications were discontinued in 6 subjects, reduced in 7 subjects, and unchanged in 5 subjects. Mean body weight decreased by 7% from 131 ± 19 kg to 122 ± 20 kg (P < .001). Fasting serum glucose decreased from 163 ± 70 mg/dL to 136 ± 48 mg/dL (P = .05) and triglycerides from 242 ± 268 mg/dL to 144 ± 124 mg/dL (P = .005). Changes in other serum lipid measurements were not statistically significant. One hypoglycemic attack requiring assistance occurred during the study. In linear regression analyses, weight change at 16 weeks did not predict change in hemoglobin A1c.

CONCLUSION:  The LCKD reduced glycemia, body weight and serum triglycerides in type 2 diabetic patients but close medical supervision was required to adjust diabetic and blood pressure medications. Controlled trials are needed to determine the safety and effectiveness of this diet compared with conventional weight loss diets.

A MULTIFACETED APPROACH TO TEACHING QUALITY IMPROVEMENT TO RESIDENTS.E.S. Holmboe1; L. Prince2; M.L. Green3. 1Yale University, New Haven, CT; 2University of Connecticut, Farmington, CT; 3Yale University, Waterbury, CT. (Tracking ID #116930)

BACKGROUND:  Understanding and applying quality improvement (QI) is now recognized as a core competency for all physicians. Educators need new curricula to train and evaluate residents and to improve residents’ current and future clinical practice.

METHODS:  A controlled longitudinal cohort study of a multifaceted self-directed curriculum in quality improvement. We studied residents in a university based internal medicine program. A four week QI course targeting PGY-2 residents during an ambulatory medicine rotation consisted of a syllabus with core readings from the recent Institute of Medicine reports, training in clinical care audits, performance of a resident self audit of their continuity clinic diabetic patients, weekly self-reflection facilitated by a faculty member, and completion of a commitment to change (CTC) survey. PGY-3 residents, who did not complete the QI course, served as concurrent controls. The primary outcome was changes in key diabetes process and outcome of care metrics. An independent abstractor performed baseline (1 year prior to start of QI course) and follow-up (1 year after day 1 of QI course) medical record audits. To be eligible for analysis, the resident must have seen the patient at least once in both the baseline and follow-up years.

RESULTS:  In the first year of the program, thirteen PGY-2 residents completed the QI course. However, only 95 (out of a possible 155) diabetic patients were eligible for analysis. Despite the small number of patients, several process and outcome of care metrics improved in the PGY-2 group. The proportion of patients with blood pressure <130/80 rose from 32% at baseline to 59% at follow-up, the proportion of patients with LDL <100 rose from 66% to 73%, and patients receiving yearly monofilament exam from 15% to 27%. Compared to 13 matched control PGY-3 residents, patients in the PGY-2 group were significantly more likely receive foot exams, pneumovax, and a baseline ECG (P < .05), and although not statistically significant, more likely to achieve goal blood pressure and have a hemoglobin A1c measured. The 13 PGY-2 residents proposed (“committed to”) 39 individual changes at the end of the rotation. At the 6 month follow-up, the 13 PGY-2 residents believed they had fully or partially implemented 35 (90%) of their individual changes.

CONCLUSION:  This study demonstrates that a multifaceted self-directed curriculum in quality improvement is highly valued by residents, leads to changes in self-reported practice behaviors, and can also lead to modest but meaningful effects on actual patient care. Future work should involve a larger number of residents and investigate the impact of systems on resident learning in quality improvement.

A NATIONAL COMPARISON OF THE PERFORMANCE OF TRADITIONAL MEDICARE AND MEDICARE MANAGED CARE.B.E. Landon1; A.M. Zaslavsky1; S. Bernard2; M.J. Cioffi1; P.D. Cleary1. 1Harvard University, Boston, MA; 2Research Triangle Institute, Research Triangle Park, NC. (Tracking ID #115742)

BACKGROUND:  Since 2000, the Centers for Medicare and Medicaid Services (CMS) has been collecting information on beneficiaries’ experiences with care for both Medicare managed care (MMC) and fee-for-service (FFS) Medicare using similar versions of the Consumer Assessment of Health Plans Study Survey (CAHPS®).

METHODS:  CMS administered MMC and FFS versions of the CAHPS survey to samples of beneficiaries from Medicare + Choice organizations and from geographic strata within traditional FFS Medicare. The main outcomes measures were four overall ratings (of the plan, personal doctor, care received overall, and care received from specialists), 5 measures summarizing experiences with care (getting care you need, getting care quickly, communication with providers, courtesy and respect of doctor's office staff, and paperwork, information and customer service) and reports of receipt of 3 preventive services (flu shots, pneumonia vaccinations, and smoking cessation advice). Differences between MMC and FFS within states were assessed with weighted t tests after adjusting for casemix and non-response. For estimates at the regional and national level, state estimates were combined after weighting by the MMC enrollment in the state.

RESULTS:  We analyzed responses collected in 2000 and 2001 from 299,058 beneficiaries enrolled in MMC (RR 82%) and 198,811 enrolled in FFS Medicare (RR 68%). MMC and FFS respondents were similar to each other and to the Medicare population as a whole. Nationally, FFS Medicare beneficiaries rated their experiences with care higher than did MMC beneficiaries, for instance in ratings of care received (8.91 on a scale of 1–10 versus 8.86, P < .001 in 2000 and 8.88 vs. 8.78, P < .001 in 2001). Differences between FFS and MMC varied across states, however. Managed care enrollees reported significantly fewer problems with paperwork, information, and customer service (2.62 vs. 2.55, P < .001 in 2000 and 2.59 vs. 2.51, P < .001 in 2001) and managed care enrollees were also more likely to report having received immunizations for influenza and pneumonia (from any source) and smokers were more likely to report having received counseling to quit smoking.

CONCLUSION:  In this report, we summarize beneficiary experiences with traditional FFS Medicare and Medicare managed care over a two-year period. Neither MMC nor FFS Medicare is superior in all aspects. Our data suggest that managed care excelled in delivering preventive services while traditional Medicare excelled in other aspects of care related to access and beneficiary experiences. These relative strengths should be considered when making policy decisions that affect the availability of choice or incent beneficiaries to choose one model over another.

A NATIONAL SURVEY OF PRIMARY CARE PRACTICE-BASED RESEARCH NETWORKS.W.M. Tierney1; B.L. Hudson1; C.C. Oppenheimer2; A. Finn1; J. Benz2; A. Zafar1; D. Lanier3; D.S. Gaylin2. 1Indiana University School of Medicine, Indianapolis, IN; 2National Opinion Research Center, University of Chicago, Washington, DC; 3Agency for Healthcare Research and Quality, Rockville, MD. (Tracking ID #115672)

BACKGROUND:  Most research is performed in academic medical centers where patients, providers, and care are atypical. Practice-based research networks (PBRNs) are a potential resource for performing relevant research in real-world settings and translate research into everyday practice. U.S. primary care PBRNs have not been described.

METHODS:  We identified 111 PBRNs of primary care practices that perform patient-based clinical research that had applied for funding through AHRQ's PBRN initiative or the Robert Wood Johnson Foundation's Prescription for Health program or belonged to the Federation of PBRNs. Using a Web-based survey, we assessed PBRNs’ history, size, location, organization, resources, and operations. We invited each PBRN director to participate and recontacted all non-respondents by e-mail or telephone.

RESULTS:  84 (76%) of the 111 PBRNs completed the survey between May and November of 2003. They contain 2,724 practices in 44 states and Puerto Rico caring for more than 16 million patients, an average of 198,112 patients per PBRN (range 1,200 to 2.7 million). Of the 32 PBRNs with practices in a single primary care specialty, 17 (53%) are family medicine, 10 (31%) pediatrics, and 1 (3%) general internal medicine [GIM]. Of 52 (62%) with multiple specialties, 51 (98%) had family medicine practices, 39 (75%) pediatrics, and 37 (71%) GIM. Mean patient age is 19 (SD = 64) years; 38% are under 18 and 18% over 60 years old; 60% of patients are white, 22% black, and 18% other races; and 18% are Hispanic. Insurance coverage is 45% private, 42% state/federal, and 13% none. Most PBRNs (63%) believe they are young, under development; 37% feel they are mature. The majority (76%) are affiliated with academic medical centers while 7% are affiliated with professional societies. A quarter have completed no studies, while half have completed 4 or more. Research is both descriptive (of populations 57%, diseases or treatments 54%, and individual patients 27%), and interventional (28%). The mean number of patients studied was 1,775 (SD 3,666, median 888, range 16–20,000). Common research foci have included prevention (49%), diabetes (37%), cardiovascular risk (31%), mental health (27%), and cancer (19%). The PBRNs site funding, community involvement, and informatics as needing improvement.

CONCLUSION:  Primary care PBRNs are large, numerous yet young. Minority and under-insured patients are over-represented. Despite being “practice based,” most have ties to academia and have performed relatively few (but large) studies of common primary care issues. PBRNs are a potentially rich national resource for community-based research.

A PATIENT REMINDER LETTER IMPROVES COMPLIANCE WITH FECAL OCCULT BLOOD TESTING.G.P. Barnas1; D. Ehley2. 1Medical College of Wisconsin, Milwaukee, WI; 2Milwaukee VA Medical Center, Milwaukee, WI. (Tracking ID #116878)

BACKGROUND:  Colorectal cancer screening is recommended by the USPSTF and the VA for all patients age 50 and older with either annual fecal occult blood testing (FOBT), a sigmoidoscopy every 5 years, or a colonoscopy every 10 years. Achieving this goal in a majority of patients has been proven difficult in clinical practice. While prior studies have focused on methods to improve providers offering screening studies to their patients, we chose to study if patient compliance with returning FOBT cards once issued could be improved by sending a reminder letter to patients who were given a set of FOBT cards but did not return them within one month.

METHODS:  All patients seen in Primary Care outpatient clinics at the Milwaukee VA Medical Center issued FOBT cards Sept–Dec 2001 were analyzed weekly via electronic medical record review of orders and lab results to determine whether or not they returned the cards within a one month time period. Those who did not were randomized to receive either a mailed reminder letter encouraging them to return their cards or else usual care and follow-up. The percentage of cards returned in both groups over the succeeding months was then analyzed to determine the impact of the reminder letter on the overall return rate.

RESULTS:  A total of 2,230 patients were issued FOBT cards within the four month study period. A total of 1,240 cards were returned within one month of issue indicating an initial patient compliance rate of 55.6%. The 990 patients who did not return their cards were then randomized to receive a reminder letter (496 patients) or usual care (494 patients) as a control group. Of those patients receiving letters, 131 (26.4%) subsequently returned the FOBT cards compared to only 32 (6.5%) of controls (Chi-square, P < .001). This difference represents an absolute increase in the subsequent return rate of 19.9% (for every 5 reminder letters sent—one additional FOBT test was returned). This would translate into an improved overall compliance rate of 67.3% vs. 58.5% if this intervention was implemented clinic-wide. In analyzing the time impact of the reminder letter on the return rate –87.5% of cards that were eventually returned were returned within 21 days of sending the reminder letter.

CONCLUSION:  A standardized one month reminder letter is a useful and low cost method to further enhance patient compliance with colorectal cancer screening. This is especially true in clinics or facilities that have electronic records in place that can be used to monitor the ordering and return status of FOBT cards.

A PILOT RANDOMIZED CONTROLLED TRIAL OF EXPOSURE TO A WOMEN's SAFE SHELTER USING QUALITATIVE METHODS TO EXPLORE RESIDENTS’ PERSPECTIVES ABOUT INTIMATE PARTNER VIOLENCE (IPV) EDUCATION.R.S. Brienza1; L.M. Whitman1; M.L. Green1; L. Ladouceur1; I. Alexander2. 1Yale University School of Medicine, New Haven and Waterbury, CT; 2Yale University School of Nursing, New Haven, CT. (Tracking ID #116669)

BACKGROUND:  Although IPV remains a major problem, patients are not screened routinely. Multiple curricula have been described for IPV screening and intervention, however, deficiencies exist in efficacy evaluation and study design. We conducted a randomized trial of exposure of IM residents to a women's safe shelter.

METHODS:  We studied residents in a university based IM training program during the ambulatory block rotation. We randomized residents to usual methods (lecture, video, role play—controls) or usual methods plus a visit to a women's safe shelter (cases). The shelter visit consisted of a meeting of women survivors facilitated by a counselor focusing on experiences with the health care system. The primary outcome was residents’ perspectives on IPV education, which we assessed via independent focus groups of convenience sub-samples of cases and controls. A non-study investigator facilitated the focus groups using a discussion guide. Transcripts from the focus groups were independently analyzed by different investigators, using common coding techniques and the constant comparative method of thematic analysis. Emerging themes from case and control groups were then contrasted.

RESULTS:  Subjects included 36 Primary Care IM residents randomized to case (n = 18) or control (n = 18) groups during academic year 2001–02. Most participants were female (total: n = 20, 58%; focus group subsample: n = 10, 59%). Five audiotaped focus groups were conducted (2 case, n = 4 & 3; 3 control, n = 4,3, & 3). Case group themes included 1) impact of educational interventions (hearing real women's stories, women's experiences with clinicians, and interactions between different educational methods), and 2) knowledge and practice (physician's role, frustration with process, definition of success, dispelling stereotypes, and IPV facts). Control group themes focused on 1) screening barriers (handling “yes” responses, offending patients, and stereotypes) and 2) information learned to overcome or manage barriers (how to ask, resources, and universality of risk). Thematic differences included the described impact, by the case group, of hearing real women's stories at the safe shelter and planned change to screen all women, not just “red flag” situations. Conversely, controls focused more on desire for clearer guidelines for how and when to screen and intervene.

CONCLUSION:  These results suggest that exposure to women survivors at a safe shelter may have a different impact on IM trainees than standard educational methods alone. Future research using these results to inform quantitative survey design and aimed at evaluating practice behaviors following educational interventions is needed.

A PILOT STUDY OF A LOW CARBOHYDRATE KETOGENIC DIET FOR OBESITY-RELATED POLYCYSTIC OVARY SYNDROME.E.C. Westman1; W. Yancy1; J. Hepburn1; J. Mavropoulos1. 1Duke University, Durham, NC. (Tracking ID #117295)

BACKGROUND:  Polycystic ovary syndrome (PCOS) is the most common endocrine disorder among women of reproductive age, and is frequently associated with central obesity, insulin resistance, and dyslipidemia. Because recent evidence demonstrates that a low carbohydrate ketogenic diet (LCKD) leads to weight loss and improvements in insulin sensitivity, we conducted this uncontrolled trial of the diet for PCOS.

METHODS:  Subjects were recruited from the community. Inclusion criteria were signs or symptoms suggestive of PCOS (chronic anovulation, hyperandrogenemia, hirsuitism), age 18–45 years, body mass index >27 kg/m2, and motivation to lose weight. Subjects received LCKD counseling, with an initial goal of <20 grams of carbohydrate per day, gradually increased as tolerated. Subjects were instructed to take a multivitamin and to drink 6–8 glasses of water daily. Fasting blood samples were obtained at weeks 0, 10 and 24.

RESULTS:  Eleven women were enrolled; 5 (45%) completed the 24-week study. In the 5 adherent subjects, there were significant reductions from baseline to 24 weeks in body weight (101.5 to 89.2 kg, P = .01) and percent free testosterone (2.2 to 1.7%, P = .04). There were non-significant changes in insulin (23.7 to 8.2 mg/dl), glucose (97.4 to 79.2 mg/dl), testosterone (51.8 to 48.0 mg/dl), hgbA1c (6.0 to 5.4%), perceived body hair (3.8 to 2.4 on a 7-point scale), LDL (120.0 to 131.8 mg/dl), and triglycerides (101.8 to 73.2 mg/dl). Two women who had previous difficulty becoming pregnant, became pregnant during the study.

CONCLUSION:  In women with obesity and a clinical diagnosis of PCOS, a LCKD led to weight loss and a reduction in percent free testosterone over a 24-week period in those able to adhere to the diet. Further controlled studies are needed to determine whether this approach is superior to other weight loss methods for the treatment of PCOS.

A RANDOMIZED CONTROLLED TRIAL OF A PATIENT-ACCESSIBLE ELECTRONIC MEDICAL RECORD.S.E. Ross1; C. Lin2; L. Wittevrongel2; L. Moore2. 1University of Colorado Health Sciences Center, Aurora, CO; 2University of Colorado Health Sciences Center, Denver, CO. (Tracking ID #117056)

BACKGROUND:  Several medical institutions have given patients access to online medical records. Giving patients access to online medical records may improve doctor-patient communication and patient self-care. At the same time, direct access to medical records may make patients confused or worried, and may interfere with clinical practice. Patient-accessible medical records have not yet been studied in a clinical trial.

METHODS:  We enrolled 107 patients with congestive heart failure. 54 patients were randomly assigned to the intervention group, which received 12 months of access to a patient-accessible online medical record including clinical notes, laboratory results, and other test results. This system also allowed patients to communicate with the practice electronically. 53 patients were assigned to the control group, which received standard care over the same period. Patients completed surveys at baseline, 6 months, and 12 months. Surveys assessed health status, adherence, and patient satisfaction. The number of visits to the practice, to the emergency department, and to the hospital were tracked. Message volume (electronic and telephone messages) was also tracked.

RESULTS:  After 12 months, general adherence, as measured by the Medical Outcomes Study adherence measure, was significantly improved in the intervention group. (85 vs. 79 on 100 point scale, P = .01). The intervention group demonstrated a trend towards improvement in several questions assessing patient satisfaction with doctor-patient communication. The intervention group did not differ from the control group in health status, number of clinic visits, or number of hosptializations. There were more emergency department visits in the intervention group (20 vs. 8, P = .03), but only 4 of the visits in the intervention group were associated with use of the online medical record in the week preceeding. Overall message volume for the first 6 months was increased in the intervention group (146 messages vs. 88) but not in the second 6 months (110 vs. 103). Nurses responding to messages did not perceive a change in message volume, and reported that no inappropriate electronic messages were sent by patients.

CONCLUSION:  In a population of patients with congestive heart failure, access to a patient-accessible online medical record resulted in improvement in general adherence and a trend towards improvements in patient satisfaction with doctor-patient communication. This access did not change health status or overall utilization of services. It was implemented with modest increases in messsage volume, and patients used it appropriately.

A RANDOMIZED TRIAL COMPARING TELEMEDICINE CASE MANAGEMENT WITH USUAL CARE IN OLDER PATIENTS WITH DIABETES MELLITUS.S.J. Shea1; R.S. Weinstock2; J. Starren1; J. Teresi3. 1Columbia University, New York, NY; 2Joslin Diabetes Center, SUNY Upstate Medical University, Syracuse, NY, Syracuse, NY; 3Hebrew Home for the Aged at Riverdale, Bronx, NY. (Tracking ID #116378)

BACKGROUND:  Telemedicine is a promising method for delivering health care services but large randomized evaluation studies are lacking.

METHODS:  We randomized 1,665 Medicare beneficiaries aged 55–84 years, with diabetes, and living in federally designated medically underserved areas of New York State (890 in Upstate NY, 775 in New York City) to receive telemedicine diabetes case management vs. usual care. At baseline, 70.1% reported household income <$20,000/year and 79.0% responded “No” to the question “Do you know how to use a computer?”. Intervention subjects received a home telemedicine unit that supported 2-way video conferencing with a nurse case manager at a diabetes center, upload of home glucose and blood pressure measurements to case management software, and access to an educational website. Evaluation data were collected in person at baseline and 1 and 2 years follow-up, with masking of evaluation staff to group assignment. Findings reported here are for an interim analysis of 1 year follow-up (N = 1,334). Group mean value at baseline was subtracted from follow-up value for each group; net differences between intervention and usual care are shown. P values are based on standard errors adjusted for clustered randomization by physician practice.

RESULTS:  See Table

Table shows mean values at 1 year follow-up

 Usual CareInterventionNet DP Value
Glyc Hgb (%)  7.17  6.96–0.17<.01
Systolic BP (mmHg)140.7137.5–3.6<.0001
Diastolic BP (mmHg) 70.1 68.5–1.9<.0001
Total Cholesterol (mg/dl)182.4170.4–10.7<.0001
Triglyceride (mg/dl)171.8154.7–16.1<.0001
LDL Cholesterol (mmHg)105.9 95.8–9.3<.0001

CONCLUSION:  Telemedicine is an effective method for providing diabetes case management to medically underserved older patients.

A SEXUAL HISTORY/HIV COUNSELING WORKSHOP USING STANDARDIZED PATIENTS INCREASES KNOWLEDGE AND IMPROVES SKILLS.S.A. Haist1; A. Hoellein1; G. Talente2; M.L. Jessup1; J.F. Wilson1; C. Griffith1. 1University of Kentucky, Lexington, KY; 2East Carolina University, Greenville, NC. (Tracking ID #116327)

BACKGROUND:  To determine the effect of a sexual history/HIV counseling (SHHIV) workshop on knowledge and clinical skills of workshop participating 3rd-year medical students versus students not participating.

METHODS:  A 4-hour SHHIV workshop was developed as part of a new curriculum for a required 3rd-year medical school 4-week primary care internal medicine clerkship. The workshop was interactive with students participating in 4 standardized patient (SP) cases representing different challenges (27 year- old man, “I want an HIV test”). A faculty preceptor then discussed the cases. The students were provided an 11-page SHHIV reference. The workshop was delivered at the beginning of 12 of the 24 rotations during 2001–02 and 2002–03. All students on all rotations were assigned SHHIV textbook readings and participated in a preventive care (PC) SP wWorkshop. At the end of the 4 weeks, all students took a 139-question written exam (9 SHHIV questions) and a 9-station SP exam (1 SHHIV station and 2 PC stations). Total SHHIV (35 items), sexual history (11 items), infectivity (12 items), condom use (5 items) and STD testing (7 items) scores were determined. After each SP station, students completed a station-related open-ended written exercise (list HIV risk factors). Simple means, standard deviations, and multiple regression approaches were used to compare the results of workshop-participating students vs. non-participating students. The 2 PC SP checklist scores were a control variable for the checklist analyses and USMLE Step 1 score was a control variable for both the exam questions and the post-SP open-ended written exercise.

RESULTS:  81 students participated in the SHHIV workshop and 85 did not. Workshop participants scored 7.1 on the SHHIV written questions vs. 6.7 for the non-participants (F = 5.05, P = .03). On the SHHIV SP checklist, total SHHIV scores of participants were higher than non participants (24.9 vs. 20.3; F = 18.2, P < .001) as were subscale scores for sexual history (8.4 vs. 6.7; F = 20.8, P < .001), infectivity (7.7 vs. 5.4; F = 18.3, P < .001), and STD testing (6.1 vs. 5.6; F = 3.8, P = 02). Scores on the opened-ended SP post-encounter exercise were no different for participants vs. non-participants (4.91 vs. 4.76, F = 1.3, P = .27).

CONCLUSION:  Students participating in a 4-hour SHHIV workshop using SPs had better SHHIV clinical skills as assessed by an SP exam and had greater SHHIV knowledge as assessed by 9 written questions compared to students not participating in the SHHIV workshop four weeks later.

A SNAPSHOT OF HEALTHCARE FOR AMERICAN INDIANS: THE NATIONAL HEALTHCARE DISPARITIES REPORT.P. Johannson1; E. Moy1; S. Siegel1; H. Burstin1. 1Agency for Healthcare Research and Quality, Rockville, MD. (Tracking ID #115982)

BACKGROUND:  The National Healthcare Disparities Report, released on December 22, 2003, represents the first in an annual report to Congress that tracks prevailing disparities in healthcare delivery by race, ethnicity, and socioeconomic status in priority populations. However, data on certain populations, including American Indians and Alaska Natives (AI/ANs), are often limited by insufficient numbers to allow reliable estimates for many measures.

METHODS:  The data for the report come from a variety of nationally representative federal data sets. These data are used to assess the healthcare received by AI/ANs across a broad range of access to care and quality of care measures. Differences between AI/ANs and whites are quantified and presented in this abstract as relative rates. In addition, gaps in the ability of extant data to provide reliable estimates for AI/ANs are identified.

RESULTS:  Overall, half of quality of healthcare measures and a third of access to care measures allowed for reliable estimates for AI/ANs populations. Among the measures utilized by the National Healthcare Disparities Report with adequate sample sizes, significant healthcare disparities were found in healthcare quality and access among AI/ANs. We found that AI/AN populations received worse quality of care than the comparison population for numerous measures, including: lower rates of breast cancer screening for women 40 and over (Relative Rate [RR], 0.68), renal transplantation (RR, 0.78), cholesterol screening (RR, 0.87), and prenatal care in the first trimester (RR, 0.82). For measures of diabetes management, sample sizes were insufficient to provide reliable estimates for the AI/AN population. On access to healthcare, AI/AN populations had worse access to care than the comparison population for numerous measures including: no health insurance coverage (RR, 2.51), % of patients with difficulty getting appointments on short notice (RR, 1.71), % of patients filling a prescription for medication in the past year (0.86), % of patients with dental visits (0.55), and outpatient visits per population (0.24).

CONCLUSION:  Significant healthcare disparities were demonstrated in the quality of care and access to care for American Indians, when compared with the white population. However, large differences on many measures between AI/ANs and whites did not obtain statistical significance due to limited AI/ANs sample size in federal data sets. Enhanced data collection efforts targeted at AI/AN populations are needed. These results suggest that targeted improvement efforts are needed to reduce disparities among American Indians and reduce the significant preventable burden of disease.

A SYSTEMATIC REVIEW OF RESIDENCY ETHICS AND PROFESSIONALISM CURRICULA.J.R. Rosenbaum1; M.L. Green1. 1Yale University, Waterbury, CT. (Tracking ID #117557)

BACKGROUND:  In the past two decades, ethical and professional lapses among physicians have received increasing attention. Several bodies, including the ACGME, now mandate or recommend training in these areas. A summary and analysis of the experience to date would help educators identify or develop curricula for their own programs. Thus, we performed a systematic review of reported residency ethics and professionalism curricula.

METHODS:  We identified curricula by searching Medline, PsychInfo, Educational Resources Information Center, two bioethics databases, Association of Program Directors in Internal Medicine Educational Clearinghouse, several medical education journals, and bibliographies of collected articles. We included articles that focused on ethics or professionalism education of residents exclusively or as part of a larger educational program. We abstracted information on demographics, curriculum development, learning objectives, instructional strategies, and evaluation methods and results.

RESULTS:  Our search identified 36 curricula, including 29 that addressed ethics and 7 that focused on issues of professionalism. Inclusion criteria showed good inter-rater reliability (Kappa = 0.77, P < .01). Only 8 curricula included descriptions of how the curricula were developed, and 6 included needs assessments. Thirty-two articulated goals or objectives. Most curricular objectives aimed to increase awareness of ethical issues, knowledge of values and ethical principles that should be used to guide behavior, or the residents’ ability to apply this knowledge effectively in clinical practice. More recent curricula focused more on case based discussion and less on ethical theory. The most commonly used instructional strategies included small group discussions (n = 26), lectures (13), role playing (8), videotape as either a trigger or feedback tool (8), and self-reflection exercises (7). Nine curricula sought to have an impact on institutional culture. Only 8 reports included an evaluation of at least one of the following effects of the curricula on the learners: knowledge (n = 6), attitude (3), behavior/performance (3), or skills (1). The six evaluations that met our criteria for adequate study design suggest that ethics and professionalism curricula can have a measurable, though modest, impact on knowledge, skills, and performance. The successful curricula involved either case based discussions that incorporated residents’ own case experiences or standardized patients as instructional strategies.

CONCLUSION:  A broad array of types of educational interventions for ethics and a smaller number for professionalism have been reported. Given the limitations of the descriptions available, these reports offer little guidance to educators seeking to develop or adapt curricula for their programs. More work needs to be done to assess which instructional strategies are most effective in influencing ethical and professionalism knowledge, skills, and behavior, though the data suggest that meaningful impact is possible.

A SYSTEMATIC REVIEW OF THE MULTIMEDIA COMPUTER FOR OFFICE-BASED PATIENT EDUCATION.J.L. Wofford1; E.D. Smith2; D.P. Miller1. 1Wake Forest University, Winston-Salem, NC; 2University of New Mexico, Albuquerque, NM. (Tracking ID #116183)

BACKGROUND:  Compared with the ubiquity of the multimedia computer in formal education, the use of computer-assisted patient education in the clinical setting is rare. the objective of this study was to explore the potential use of the multimedia computer for patient education in the ambulatory health care setting.

METHODS:  Data Sources: All relevant English-language studies from MEDLINE and Cochrane Library databases, and reference lists of key articles. Study Selection: Randomized controlled trials using computer–assisted patient education. The MEDLINE search was performed in September 2002. Two independent assessors reviewed the retrieved citations. Data Extraction: Seventeen randomized trials of computer-assisted patient education were identified. Data were extracted on trial characteristics and computer strategies. Two reviewers independently rated trial quality using established criteria.

RESULTS:  Data Synthesis: Outcomes examined in these trials included clinical measures (7/17, 41.1%), knowledge retention (8/17, 47.1%), health attitudes (12/17, 70.6%), level of shared decision making (4/17, 23.5%) health services utilization (3/17,17.6%), and costs (4/17, 23.5%). Three trials targeted patients with breast cancer, but the educational domains were otherwise diverse. The quality of trial methodology was adequate with regard to testing the results of randomization (14/17, 82.3%) and reporting the analysis of main effect variables (12/17, 70.6%) but not adequate with regard to reliability of the randomization process (6/17, 35.3%), blinding of outcomes assessment (3/17, 17.6%), or sample size definition (5/17, 29.4%). Two trials made use of the internet, and four trials used touchscreen technology for patient interactions. Most computer strategies used audio but only three used video as a communication strategy. The computer was located in the clinical setting in one third of studies. Separate office visits were required in two trials where the computer was sited at the clinical setting.

CONCLUSION:  The potential for improving office efficiency through computer-assisted patient education has been demonstrated, but better proof of the impact on clinical outcomes will be necessary before computer-assisted patient education becomes acceptable in the office setting.

A SYSTEMATIC REVIEW OF THE OBSERVATIONAL STUDIES OF THE EFFECTS OF SLEEP DEPRIVATION ON RESIDENT PERFORMANCE.K.E. Fletcher1; W. Underwood2; S.Q. Davis3; R.S. Mangrulkar2; L.F. McMahon2; S.K. Saint2. 1Medical College of Wisconsin/ Milwaukee VAMC, Milwaukee, WI; 2University of Michigan, Ann Arbor, MI; 3University of Chicago, Chicago, IL. (Tracking ID #115606)

BACKGROUND:  New work hour regulations have begun for all resident physicians in the United States. While the primary goal is to improve patient safety by decreasing resident fatigue, solutions often create discontinuity of care. We have previously reported that interventional studies found equivocal results with respect to patient safety. Herein, we assess the effect sleep deprivation has on resident performance based on observational studies, thereby further informing the process of how best to improve patient safety in the context of resident work hour restrictions.

METHODS:  We used electronic databases and additional methods to search the English-language literature for studies published since 1966 that evaluated any aspect of resident work hours. All potentially pertinent studies were independently reviewed by 2 investigators.

RESULTS:  The total number of studies included was 118. Of those, 34 were about interventions which had been implemented to reduce the effects of fatigue or work hours (results previously reported). The remaining 84 studies were observational studies; the 44 studies with patient safety relevance are included in this report. Evidence for the effects of sleep deprivation on patient outcomes may come from studies that evaluate 4 types of outcomes: 1) patient-related adverse events; 2) medical errors; 3) performance on tests or simulated situations; and 4) tests of cognitive function. Few studies directly assessed the effects of sleep deprivation on actual patient outcomes. Of these, one assessed actual errors and 3 used self-reported errors. These studies do not provide strong evidence that sleep deprivation worsens patient outcomes. However, a larger group of studies focusing on test performance or simulation does suggest that test scores appear to be worse in sleep-deprived groups. Skills such as laparoscopic simulations may also decline, and tasks requiring vigilance also tend to worsen. Finally, the studies of cognitive function using psychological tests found, in general, that reaction time, concentration and dexterity seem to be adversely affected by sleep deprivation, while memory and problem solving are less affected.

CONCLUSION:  Unfortunately, few studies have related sleep deprivation to actual patient outcomes. Evidence suggests, however, that sleep deprivation may adversely impact certain aspects of cognitive function and performance, especially those activities requiring vigilance and dexterity. Defining the resident tasks that are especially susceptible to sleep deprivation and those susceptible to discontinuity should inform the allocation of responsibilities within a resident team to maximize patient safety.


BACKGROUND:  The identification of medical errors has historically relied upon reviewing sentinel events and medical record audits. Little is known about the voluntary disclosure and perception of medical errors by healthcare workers in the clinical setting. The objective of this study is: To categorize the perception, types and nature of errors as identified by nurses, faculty, and residents by a voluntary reporting system for patients on a teaching service in a community hospital.

METHODS:  This is a descriptive prospective cohort study of resident admissions at a 250-bed community hospital from 9/1/02 to 9/1/03. The chief resident on the post-call day encouraged residents to voluntarily report potential medical errors using a standardized card. Anonymous incident report cards with drop boxes were placed on the medicine floors to allow any healthcare provider to report perceived medical errors on house staff patients. The definition of medical error was adopted from the IOM's report as “failure of a planned action to be completed as intended or use of a wrong plan to achieve an aim” and were screened by the primary investigators. Any disagreements regarding the classification of error were resolved by consensus. The events were then categorized as an error of drug administration, therapeutics, diagnosis, or procedure.

RESULTS:  There were 104 potential medical errors that were voluntarily reported. Upon review, 90 events met the definition of a medical error. Of these errors, 55 (61%) were reported by residents while 35 (39%) were by anonymous reporting. Nurses were responsible for all but one anonymous error reports. Individuals more often reported errors involving other healthcare providers in 74 (82%) than self-disclosure of errors in 16 (18%). The types of medical errors disclosed varied greatly between residents and nurses. Residents disclosed errors of diagnosis in 18 (33%), drug administration in 16 (29%), and therapeutics in 16 (29%). Resident therapeutic errors were defined by medication dosing in 5 (31%) and therapeutic decisions in 11 (69%). Nursing reported mostly drug administration errors in 23 (66%).

CONCLUSION:  In this study of voluntary disclosure, residents were more likely to disclose errors of diagnosis and therapeutics while nurses reported drug administration errors.

ABUSE OF PRESCRIPTION PAIN RELIEVERS AND CO-USE OF OTHER DRUGS: A POTENTIAL BARRIER FOR OFFICE-BASED TREATMENT WITH BUPRENORPHINE?L.D. Woodard1; J.R. Kramer1; P. Richardson1; L. Capistrano1; C.M. Ashton1; R.O. Morgan1. 1Houston Center for Quality of Care and Utilization Studies & Baylor College of Medicine, Houston, TX. (Tracking ID #116676)

BACKGROUND:  Prescription pain reliever abuse has significantly increased the number of patients requiring treatment for opiate addiction. Recent U.S. legislation allows physicians to prescribe buprenorphine to treat opiate-addicted patients in their practices. However, physicians have been reluctant to embrace this treatment option, possibly due to concerns regarding use of other illicit substances among patients who abuse pain relievers. In this study, we: (1) estimated the prevalence of co-use of heavy alcohol, marijuana, and other illicit substances among persons who abuse prescription pain relievers, and (2) determined the demographic factors associated with this co-use.

METHODS:  We used data from the 2000 National Household Survey on Drug Abuse to identify users of prescription pain medication (93% of which were opioid analgesics) for non-medical purposes within the prior year. Logistic regression was used to determine predictors of use of heavy alcohol (≥5 drinks, 5 + times within prior month), marijuana (prior year), and other illicit substances (prior year) among users of pain relievers. Model predictors were age, sex, race, education, employment, household income, veteran status, and metropolitan statistical area.

RESULTS:  Approximately 2.9% of Americans reported non-medical use of pain relievers during the prior year. Of these, large percentages reported heavy alcohol, marijuana, and other illicit substance co-use (23.1%, 46.1%, and 46.0%, respectively). Controlling for other predictors in the model, age, race/ethnicity, and marital status were all broadly associated with heavy alcohol, marijuana, or other illicit substance co-use. Younger persons were more likely to be co-users than those aged 50 years and older (P ≤ .001 for all three types of co-use); Whites were more likely than other race/ethnicities to co-use (P ≤ .001 for all three types of co-use); and single individuals were more likely than those who were married to co-use (P ≤ .0001 for all three types of co-use). Men were more likely than women to co-use heavy alcohol and marijuana (P ≤ .0001 and P ≤ .001, respectively), but not other illicit substances, while individuals with a college education were more likely to co-use other illicit substances than those with less education (P ≤ .001). Education level was unrelated to differences in heavy alcohol or marijuana co-use.

CONCLUSION:  Co-use of heavy alcohol, marijuana, and other illicit substances is high among users of prescription pain relievers. Individuals who are younger, White, male, and unmarried are most likely to use multiple substances. These findings suggest that substance co-use is prevalent among users of prescription pain relievers and may be a barrier limiting physician acceptance of office-based treatment with buprenorphine.

ACCES TO EMERGENCY CONTRACEPTION IN TWO ECONOMICALLY DISPARATE NEIGHBORHOODS IN NEW YORK CITY.R. Agarwal1; J. Olson1; H. Kunins2. 1Albert Einstein College of Medicine, Bronx, NY; 2Albert Einstein College of Medicine, New York, NY. (Tracking ID #117013)

BACKGROUND:  Emergency contraception is a safe and effective treatment for unwanted pregnancy. Despite the FDA advisory committee recommendation in 2003 that pharmacies dispense emergency contraception (EC) without a prescription, EC is currently available only by prescription in most states. Little is known about the actual access to EC, and whether access varies among economically disparate neighborhoods. In this study, we investigated access to EC in two economically and ethnically distinct neighborhoods in New York City.

METHODS:  Pharmacies were identified by zip code through the New York State Education Department. Using 2 internet directories, phone numbers were identified for all licensed pharmacies in the 6 zip codes from the disadvantaged South Bronx (SB) and in the 3 zip codes from the affluent Upper East Side (UES) in Manhattan. We called all pharmacies with listed phone numbers and administered phone interviews with the available pharmacist. Specialty pharmacies were excluded. The interview assessed business hours, time necessary to fill a prescription and if pharmacies kept EC in stock.

RESULTS:  104 general pharmacies with listed telephone numbers were identified. Interviews were conducted with 96 pharmacies in the SB (N = 51) and in the UES (N = 45), for a combined response rate of 92%. Pharmacies in the SB and the UES were equally likely to report that they stocked at least 1 of the 2 brands of hormonal contraception marketed as EC (78% vs. 84%, P = .6). However, SB pharmacies were less likely than UES pharmacies to carry both brands of EC (47% vs. 73%, P = .02). Pharmacies in the SB and the UES were equally likely to be able to fill an EC prescription within one hour (84% vs. 90%, P = .4). SB pharmacies were less likely to be open past 6 pm on weekdays (55% vs. 89%, P = .001) and less likely to be open on both Saturday and Sunday (31% vs. 71%, P < .001).

CONCLUSION:  A majority of pharmacies in both the disadvantaged South Bronx and the affluent Upper East Side of Manhattan stock at least 1 form of EC. In addition, most pharmacies report being able to fill a prescription for EC rapidly. More limited hours of pharmacy operation in the South Bronx, however, may adversely affect poor women's access to EC. Physicians who provide care for women from economically disadvantaged neighborhoods should counsel patients of potential difficulty in accessing EC due to constrained pharmacy hours.

ACCURACY OF COMPUTERIZED MEDICATION LISTS BEFORE AND AFTER AN INTERVENTION TO IMPROVE PRESCRIBING.B. McClimon1; A. Hoth1; R.I. Shorr2; M.J. Barnett3; G.E. Rosenthal1; P. Kaboli1. 1University of Iowa, Iowa City, IA; 2University of Tennessee, Memphis, TN; 3Iowa City VAMC, Iowa City, IA. (Tracking ID #117237)

BACKGROUND:  Due to the lack of reliability of the medical record as an accurate source of medication history, computerized medication records have been recommended to improve medication safety. However, to our knowledge, no studies have evaluated the validity of medication profiles found in computerized medical records. The objective of our study was to determine the accuracy of a computerized medication list before and after a collaborative phamacist-physician (PharmD-MD) intervention in older veterans to improve prescribing.

METHODS:  The sample included 530 primary care patients 65 years and older receiving prescriptions for >5 medications in a VA primary care clinic. A structured medication history and medical records review was performed to confirm the list of medications the patient was taking. Patients were randomized to usual care or an intervention in which therapeutic recommendations were presented to patients’ primary care providers. Accuracy of the computerized medication list was assessed at baseline and 3 months and included omissions (medications not on computer list) and commissions (medications on list that were no longer being taken). Changes over time between groups were assessed using analysis of covariance (ANCOVA).

RESULTS:  Patients (mean age, 74 years; 98% male) were taking a mean of 13.9 ± 5.1 medications at baseline and 13.1 ± 4.8 at 3 months; 66% were prescription, 22% over-the-counter (OTC), and 11% vitamins/herbals. At baseline, only 5% had complete agreement between the computer medication list and what the patient was taking and increased to 8% at 3 months, with no difference between the groups (P = .37). Rates of omissions decreased from 24% in the intervention group to 20% at 3 months, while omission rates decrease from 22% in the control group to 20% at 3 months (P = .27). The most frequent omissions were aspirin, vitamins/herbals, acetaminophen, and non-steroidals. Of omissions, 34% were prescription, 35% OTC, and 31% vitamins/herbals. The commission rate for the intervention group was 12% at baseline and 9% at 3 months, and for the control group was 13% at baseline and 11% at 3 months (P = .15). The most frequent commissions were aspirin, non-steroidals, and anti-hypertensive agents. Of commissions, 66% were prescription, 29% OTC, and 5% vitamins/herbals.

CONCLUSION:  We found very few computerized medication lists to be accurate and only 4 of 5 medications being taken were on the computer list. A PharmD-MD intervention had only minimal impact on improving the accuracy of the list. Barriers to accurate medication lists should be investigated and medication databases used in research should be evaluated for accuracy.

ACCURACY OF TESTING FOR CORONARY ARTERY DISEASE IN WOMEN: A SYSTEMATIC REVIEW.M.S. Beattie1; L. Metz1; K.E. Fleischmann1; D.G. Grady1; R.F. Redberg1. 1University of California, San Francisco, San Francisco, CA. (Tracking ID #116780)

BACKGROUND:  Many studies have suggested the accuracy of testing for coronary artery disease (CAD) in women may be different compared to men. This systematic review evaluates the accuracy of exercise echocardiography (echo) and exercise myocardial perfusion imaging (MPI) in women and men. It also evaluates the test characteristics of MPI with Thallium (Tl) compared to technetium sestamibi (MIBI) imaging.

METHODS:  PubMed, Cochrane, and DARE databases were searched for articles published from 1/90 to 7/03. Inclusion criteria were: 1) Cardiac cath was a gold standard, 2) Data was presented that allowed calculations of true positive, true negative, false positive, and false negative, 3) Clear definitions of positive test and positive cath were provided. Exclusion criteria were: 1) Less then 10 women were studied, 2) Testing was done exclusively after infarction, angioplasty, bypass surgery, or during acute coronary syndrome, 3) Pharmacologic agents were used as stressors. 34 authors were contacted to obtain gender-stratified data. Of these, 5 sent gender-stratified data, however 2 of these studies were excluded because they contained less then 10 women.

RESULTS:  3,389 titles were identified, 340 articles were reviewed, and 17 were included. Of these, 13 used MPI (6Tl, 5 MIBI, 2 both) and 4 used echo. The MPI studies included 888 women and 2,451 men. The echo studies included 344 women and 345 men. The table shows mean weighted positive and negative likelihood ratios (LR) with 95% confidence intervals.

CONCLUSION:  Positive LR for both echo and MPI, in both men and women, varied from 1.5 to 3, differences that are not clinically significant. Negative LR, similarly, were not clinically different, and varied from 0.1 to 0.4. Verification (or referral) bias plays a large role in estimating sensitivity and specificity, decreasing sensitivity by up to 30% and increasing specificity by up to 60%. Only 2 studies controlled for this.

inline image

ADHERENCE TO ANTIHYPERTENSIVE MEDICATIONS RELATED TO THERAPEUTIC CLASS AMONG ASIAN AMERICANS AND PACIFIC ISLANDERS COMPARED TO WHITES.R.P. Gelber1; D.A. Taira2; J.W. Davis2; T.B. Seto3. 1University of Hawaii John A. Burns School of Medicine, Honolulu, HI; 2Hawaii Medical Service Association, Honolulu, HI; 3The Queen's Medical Center, Honolulu, HI. (Tracking ID #115873)

BACKGROUND:  Medication adherence is a complex behavioral process, involving knowledge, motivation, environment, and resources. Research on adherence has emphasized the need to consider ethnic and cultural backgrounds when developing adherence plans. However, few studies have attempted to identify factors that are most predictive of adherence for specific racial and ethnic groups, particularly Asian Americans and Pacific Islanders (AAPI). Objectives: To assess variation in adherence related to antihypertensive therapeutic class among Japanese (n = 13,836), Filipino (n = 3,818), Chinese (n = 2,376), Korean (n = 469), and part-Hawaiian (n = 3,924) patients, as compared to whites (n = 4,144), and to examine factors associated with adherence in these patients.

METHODS:  The study population was comprised of 28,567 adult patients enrolled in a large health plan in Hawaii (July 1999–June 2003), who had a diagnosis of hypertension and had filled at least one prescription for an antihypertensive medication. Members with antihypertensive medication in their possession at least 80% of the time were considered adherent. Multivariable logistic regression models were used to identify factors associated with adherence and to compare adherence across ethnic groups.

RESULTS:  Overall adherence rates were less than 65% among all ethnic groups. After adjustment for patient age, gender, morbidity level, health plan type, isle of residence, comorbidities, and year of treatment, Japanese patients were more likely than whites to adhere to antihypertensive therapy [OR = 1.2 (1.1–1.3)], whereas Filipino [OR = 0.65 (0.60–0.70)], Korean [OR = 0.81 (0.69–0.95)], and Hawaiian [OR = 0.86 (0.80–0.96)] patients were less likely to adhere. These results were consistent across therapeutic class, suggesting that observed differences were related to factors other than therapeutic class. For all racial and ethnic groups, beta blockers and calcium channel blockers had the highest adherence rates, followed by ACE inhibitors and angiotensin receptor blockers. Members taking thiazide diuretics had the lowest adherence rates. Other factors associated with lower adherence included younger age, high morbidity level, living in remote areas, and history of congestive heart failure.

CONCLUSION:  Our findings of substantial disparities among AAPI sub-groups highlight the need to examine sub-groups separately. Further study is needed to determine appropriate interventions, particularly for Filipino, Korean, and Hawaiian patients and to examine the impact of non-adherence on health and economic outcomes.

AFRICAN-AMERICAN PATIENT's REASONS FOR UNDERGOING OR REFUSING ELECTIVE KNEE/HIP JOINT REPLACEMENT SURGERY.J.P. Lopez1; K. Kwoh1; S. Ibrahim2. 1University of Pittsburgh, Pittsburgh, PA; 2Veterans Administration, Pittsburgh, PA. (Tracking ID #117177)

BACKGROUND:  Knee or hip osteoarthritis (OA) is a leading cause of disability in the elderly and total joint replacement is an effective treatment option for end-stage disease. There is marked racial disparity in the utilization of this effective treatment option. Racial variations in expectations and perceptions regarding joint replacement may contribute to this disparity. We conducted 8 focus group interviews to examine African-American (AA) patient's perceptions and expectations regarding reasons for undergoing or refusing total joint replacement.

METHODS:  Elderly AA men and women with chronic hip/knee pain attending a community-based primary care clinic in inner city Cleveland were asked about their perceptions and expectations regarding care for their hip/knee pain. Each 90-minute focus group session was attended by 8–12 invited participants, and was led by a trained moderator and AA facilitator. The sessions were audiotaped, transcribed and 2 independent analysts coded the transcripts for thematic structure using NUD*IST® software. The main themes were tabulated, and actual text was analyzed to record specific nodes that emerged from each theme.

RESULTS:  There were 61 participants; mean age was 63.5 years (range 49–77). The mean years of education was 13 (range 10–22); 75% were Protestant, 12 % were from other Christian denominations and 13% were non-Christian; 49% were married; 48% reported annual household income of less than or equal to $20,000; 72% of the sample were women. Major reasons patients cited for having joint replacement were: 1) all other treatment options were exhausted; 2) to relieve pain; 3) to increase mobility; 4) knew of someone else who had successful joint replacement surgery; 5) need to replace worn out previous knee/hip prosthesis. Major reasons cited for refusal to undergo the procedure if offered included: 1) fear of surgery in general, and of being cut; 2) personal experience with others who had unsuccessful surgeries; 3) desire to try other treatment options first; 4) unsatisfactory experience during rehabilitation from previous joint replacement surgery; 5) no desire for repeat a procedure after the previous prosthesis had worn out.

CONCLUSION:  Elderly AA men and women with chronic knee/hip arthritis pain expressed specific reasons for and against joint replacement that should be addressed in clinical decision-making regarding this treatment option.

AFRICAN-AMERICANS’ PARTICIPATION IN HIV RESEARCH: IF YOU ASK, THEY WILL SAY YES.M. Garber1; G. Switzer2; B. Hanusa2; R.M. Arnold2. 1University of Pittsburgh Medical Center, Pittsburgh, PA; 2University of Pittsburgh, Pittsburgh, PA. (Tracking ID #117079)

BACKGROUND:  Despite the increasing incidence of HIV infection in the African-American population, national participation rates of African-Americans in HIV treatment trials remain low. This study assessed factors associated with research participation among HIV-infected African-Americans at the Pittsburgh AIDS Center for Treatment (PACT), an ambulatory care clinic within the University of Pittsburgh Medical Center.

METHODS:  In December 2002, 286 African-American patients were registered at PACT. Within the period July to November 2003, 228 African-American patients were seen at the clinic and 200 (87%) of them were recruited to complete a one-time only, anonymous questionnaire. The survey, which asked about actual participation and willingness to participate in future HIV research, was developed after reviewing the literature and identifying variables influencing research participation; it was reviewed by content experts and a focus group of patient volunteers. Univariate 2 and multivariate polychotomous logistic regression were done using Stata 7.0. (Stata Corporation, TX).

RESULTS:  Of 203 patients approached, 2 had dementia and 1 declined participation. 56% were male and 78.5% were between ages 25–50. 63% had achieved trade school or some college education and 44% had Medicaid. 155 (78%) had known about their disease for over 2 years and 21% were identified as having AIDS based on CD4 count; 122 (61%) had been registered at our clinic for over 2 years. Of the 200 respondents, 114 (57%) had been asked to participate in medical research. 98 (86%) of those asked, agreed to participate and 80 (70%) enrolled in a research study. 55 (69%) of enrollees felt good about contributing to research and 65 (81%) would participate in future studies. Of patients who agreed but did not participate, 36% cited time constraints and 29% reported drug side effects as deterrents. Univariate analyses showed that patients who had known about their HIV infection for 2 years, had been at PACT for 2 years, and with higher levels of education, were significantly more likely to enroll in research. Multivariate analyses revealed that only length of time at PACT was significantly associated with enrollment in research. Of 86 patients who had never been asked to participate in research, 59% said they would be willing to participate in future research. Multivariate analyses indicated that prior research participation (P = .023) and a commitment to staying in the Pittsburgh area (P = .046) were significantly related to willingness to participate in future research.

CONCLUSION:  The major barrier to participating in research among HIV-infected African-Americans at a university-based clinic is never having been asked.

ALCOHOL AND FOLATE INTAKE AND BREAST CANCER RISK.C. Duffy1; A.R. Assaf1; M.G. Cyr1; G. Burkholder1; E. Coccio2. 1Brown University, Providence, RI; 2Memorial Hospital, Pawtucket, RI. (Tracking ID #116920)

BACKGROUND:  Research suggests breast cancer risk is higher in postmenopausal women who consume alcohol. Folate may modulate the increased risk of breast cancer in alcohol drinkers via its role in DNA replication and cell division. However, results of studies have differed in the alcohol intake at which folate modulates risk. The relationship among alcohol, folate and breast cancer risk needs to clarified. We sought to examine this relationship in a more ethnically and socio-economically diverse group of women.

METHODS:  The observational study (OS) of the Women's Health Initative (WHI) enrolled 93,724 postmenopausal women aged 50–79 at 40 clinical centers throughout the US. Women were eligible to participate if they were postmenopausal, unlikely to move within three years, and not enrolled in any other clinical trial. Total folate was calculated from a modified Block food frequency questionnaire (FFQ), and reported supplement use, both obtained at baseline. Alcohol intake was determined from the same FFQ and converted into grams per day. Breast cancer cases were reported on a yearly medical update form, and were centrally adjudicated. We examined the relationship between alcohol and breast cancer by multivariate logistic regression, controlling for known or suspected risk factors (age, family history, reproductive history, use of hormone therapy, breast biopsies, breast feeding, income, education, BMI, tobacco, and physical exercise). We then tested for an interaction between alcohol and folate. SAS version 7.0 was used for all analysis.

RESULTS:  The mean age of women in the OS was 63.5 years, 83% were non-Hispanic white and 17% were of other ethnic/racial groups. Sixty-six percent were currently married, while 41% had at least a college degree. Seventeen percent of women had incomes less than $20,000 per year. The median alcohol intake was 5.6 g/day and average baseline folate intake was 446 mcg/day. There were 1783 cases of breast cancer in just over 5 years of follow-up. In the adjusted model with non-drinkers as the reference, women consuming <4.33 g alcohol/d had an OR = 1.07 (0.94, 1.21), between 4.33 and 15 g alcohol/d an OR = 1.18 (1.02, 1.37) and for >15 g alcohol/d an OR = 1.26 (1.07, 1.48).We did not find a relationship between folate and breast cancer risk in our adjusted model, and there was no interaction between alcohol and folate.

CONCLUSION:  In this large, ethnically and socio-economically diverse cohort of postmenopausal women, moderately low level consumption (3–7 drinks/week) was associated with increased breast cancer risk. We did not however, find evidence for folate modulating this relationship. Increasing folate intake in women who consume moderately low levels of alcohol may not reduce risk of breast cancer.

ALCOHOL BRIEF INTERVENTION IN PRIMARY CARE: A SYSTEMATIC REVIEW. N. Bertholet1; J. Daeppen1; M. Fleming2; V. Wietlisbach3; B. Burnand4. 1Alcohol Treatment Center, Lausanne, Switzerland; 2University of Wisconsin-Madison, Madison, WI; 3Institute of Social and Preventive Medicine, University of Lausanne, Lausanne, Switzerland; 4Institute of Social and Preventive Medicine, Lausanne, Switzerland. (Tracking ID #115583)

BACKGROUND:  The purpose of the present review was to evaluate the evidence of the effectiveness of brief interventions aimed at reducing chronic alcohol use and harm related to alcohol consumption, conducted among individuals actively attending primary care but who were not seeking help for alcohol problems.

METHODS:  Randomised trials reporting at-least one outcome related to alcohol consumption and conducted in outpatients who were actively attending primary care centre or provider were selected using Cochrane Central Register of Controlled Trials, MEDLINE, PsycINFO, ISI Web of Science, ETOH database, and bibliographies of the retrieved references and previous reviews. Selection and data abstraction were performed independently and in duplicate. We assessed validity of the studies and performed a meta-analysis for studies reporting alcohol consumption at 6 or 12 months follow up.

RESULTS:  We included 24 reports, reporting results of 19 trials and including 5,639 individuals. Seventeen trials reported a measure of alcohol consumption, eight reporting a significant effect of intervention. The meta-analysis showed a mean pooled difference of –41 (95% CI: −54; −28) g of pure ethanol per week in favour of brief intervention group. Evidences for other outcomes (laboratory values, health related quality of life, morbidity and mortality, health care utilisation) were inconclusive.

CONCLUSION:  Our systematic review indicated that brief intervention might be effective for both men and women in reducing alcohol consumption compared to a controlled intervention, in a primary health care population. The meta-analysis confirmed the reduction in alcohol consumption at 6 and 12 month. Further research should precise the components of effectiveness of brief intervention and the evidence of effects on morbidity, mortality, and quality of life related outcomes.

ALCOHOL CONSUMPTION AND THE PREVALENCE OF THE METABOLIC SYNDROME IN WHITE, BLACK, AND MEXICAN AMERICANS.M.S. Freiberg1; H.J. Cabral2; R.S. Vasan3; R.C. Ellison4. 1Boston University School of Medicine, Boston, MA; 2Boston University School of Public Health, Boston, MA; 3Framingham Heart Study and Boston University School of Medicine, Framingham, MA; 4Boston University School of Medicine and Public Health, Boston, MA. (Tracking ID #116836)

BACKGROUND:  While alcohol has been known to favorably influence select lipid components, the association between alcohol consumption and the metabolic syndrome (MBS) has not been adequately investigated.

METHODS:  We analyzed data on 8,198 subjects from the Third National Health and Nutrition Examination Survey (NHANES III). Eligible subjects fasted ≥8 hours, were evaluated for all components of the MBS, and answered questions pertaining to alcohol consumption as well as potential confounders. Subjects were diagnosed with the MBS if they had ≥3: abdominal waist circumference >102 cm in men, 88 cm in women; serum triglycerides ≥150 mg/dL, serum high density lipoprotein cholesterol <40 mg/dL in men, <50 mg/dL in women; blood pressure ≥130/85 mmHg or taking blood pressure medication; or fasting serum glucose ≥110 mg/dL or taking anti-diabetic medication. Alcohol consumption was defined as: (1) currently drinking (≥1 drink per month) or not currently drinking (<1 drink per month); (2) the number of alcoholic drinks per month of any type of alcohol categorized into not currently drinking, 1–<20, 20–<50, or ≥50 drinks per month; and (3) the number of beverage-specific drinks per month categorized into not currently drinking, 1–<20, or ≥20 drinks per month. Multiple logistic regression analyses were performed to estimate the odds ratio (OR) for the prevalence of the MBS in subjects consuming alcohol as compared to current non-drinkers. SUDAAN was used to obtain proper variance estimates given the NHANES III complex sampling design.

RESULTS:  After adjustment for age, sex, race/ethnicity, education, income, tobacco use, physical activity, and diet, subjects who consumed alcoholic beverages ≥1 –<20, 20–<50, and ≥50 drinks per month had OR for the prevalence of the MBS of 0.66, 0.34, 0.33 respectively as compared with current non-drinkers (P < .05 for all values). The OR for prevalent MBS was particularly reduced for beer and wine drinkers as compared to current non-drinkers. The association of alcohol intake ≥20 drinks per month with the MBS was consistent across race/ethnicities, but was most striking in white men and women (OR = 0.35 and 0.22 respectively P < .05). Alcohol consumption was significantly associated with a lower OR for the prevalence of 3 components of MBS: low serum HDL cholesterol, serum triglycerides, and high waist circumference (P < .05).

CONCLUSION:  Moderate alcohol consumption, especially beer and wine, is associated with a lower prevalence of the metabolic syndrome and with favorable influence on serum lipids and waist circumference. The observed association is strongest in white individuals. Additional investigations are warranted to confirm these findings.

ALCOHOL COUNSELING REFLECTS QUALITY OF PRIMARY MEDICAL CARE.R. Saitz1; N.J. Horton2; J.H. Samet1. 1Boston University, Boston, MA; 2Smith college, Northampton, MA. (Tracking ID #102573)

BACKGROUND:  Some primary care physicians do not screen for or address alcohol problems because they believe patients do not want to discuss their alcohol use. We hypothesized that 1) receipt of alcohol counseling is associated with higher patient-perceived quality of primary medical care, and that 2) quality of primary care is associated with alcohol consumption.

METHODS:  We studied a prospective cohort of adult hazardous drinkers in an urban academic primary care practice. All were subjects in a randomized trial testing the provision of physicians with patients’ alcohol screening results. Trained research staff interviewed hazardous drinkers (current use of risky amounts or a positive response to a CAGE alcohol screening question) after a visit with their primary care physician. They were asked if they had received alcohol counseling (advice on safe drinking limits, to cut down or abstain, or referral), and they were assessed on 3 domains of primary care performance (of 7) from the Primary Care Assessment Survey (PCAS)—Communication, Comprehensiveness (contextual knowledge of patient), and Trust. Six months later subjects completed the 30-day Timeline Follow-back interview to assess alcohol consumption. Multivariable longitudinal regression models were used to test the associations between 1) counseling and the 3 PCAS scales (possible range 0–100), and then 2) between PCAS scales and consumption, adjusting for clustering within physician, sex, race, education, comorbidity, randomized group, level of physician training, having met the physician previously, drinks per drinking day and alcohol problems.

RESULTS:  Of 312 subjects enrolled, 301 reported information on receipt of alcohol counseling during the visit; 288 subjects who saw 40 physicians completed the PCAS measure of quality of primary care. Subjects’ (n = 288) characteristics were: mean age 43; 61% men; 57% black; 77% saw a faculty physician and 71% had seen the physician previously; 132 (46%) reported alcohol counseling. Subjects drank on average 6 drinks per drinking day; 223 had assessment of consumption 6 months later. Adjusted mean PCAS quality scores were higher in patients who had received alcohol counseling: Communication 85 vs. 76, P < .0001, Comprehensiveness 67 vs. 59, P < .01, Trust 79 vs. 77, P = .06. Quality of primary care was not associated with odds of hazardous drinking 6 months later (adjusted odds ratio 1.0 for all 3 scales).

CONCLUSION:  Alcohol counseling by primary care physicians is associated with higher quality care from the patient's perspective. Quality of care was not associated with subsequent drinking. Nonetheless, these results indicate that screening and intervention for alcohol problems can enhance the quality of primary medical care.

AMBULATORY MANAGEMENT OF HYPERCHOLESTEROLEMIA IN PATIENTS WITH ATHEROSCLEROSIS: DIFFERENCES BY SEX AND RACE.S.D. Persell1; S. Maviglia2; D.W. Bates2; J.Z. Ayanian3. 1Northwestern University, Chicago, IL; 2Brigham and Women's Hospital, Boston, MA; 3Harvard Medical School, Boston, MA. (Tracking ID #114034)

BACKGROUND:  Women and black patients with cardiovascular disease receive fewer invasive procedures. Less is known about sex and race differences in risk-factor management. We assessed outpatient cholesterol management and control by sex and race.

METHODS:  We performed a 1-year observational cohort study of the care received by 243 patients with coronary heart disease, cerebrovascular disease, or peripheral vascular disease and a low-density lipoprotein cholesterol (LDL-C) >130 mg/dl receiving primary care at 10 general medicine clinics affiliated with an academic medical center. We analyzed 1082 office visits eligible for cholesterol management. Main outcome measures were intensification of cholesterol lowering medication or LDL-C monitoring at clinic visits, and attainment of LDL-C <130 mg/dl within 1 year.

RESULTS:  Medication intensification or LDL-C monitoring occurred at 31.2% of women's and 38.5% of men's visits, (P = .01), and at 37.3% of black and 31.7% of white patients’ visits (P = .09). Independent predictors of these actions included female sex (adjusted odds ratio [OR], 0.71, 95% CI, 0.50–1.00), preceding LDL-C (OR per 10 mg/dL, 1.17, 95% CI, 1.10–1.24), medication non-adherence (OR, 1.81, 95% CI, 1.25–2.61), having a new clinical problem addressed (OR, 0.47, 95% CI, 0.35–0.63), and seeing a primary care clinician other than the patient's primary care physician (OR, 0.12, 95% CI, 0.06–0.25). Visits of black patients were more likely to result in these actions, OR 1.20 (0.81–1.78), but this difference was not statistically significant. In one year, an LDL-C <130 mg/dl was achieved less often for women than men (41% vs. 61%, P = .003), and for black patients than white patients (39% vs. 58%, P = .01). Adjustment for clinical characteristics and cholesterol management attenuated the relationship between having LDL-C <130 mg/dl and sex (OR women vs. men, 0.58, 95% CI, 0.29–1.13) but not race (OR black vs. white, 0.40, 95% CI, 0.19–0.84).

CONCLUSION:  In this high-risk population with uncontrolled LDL-C, women received less intense outpatient cholesterol management than men. For black and white patients management was similar. Women and black patients were less likely than men and white patients, respectively, to achieve moderate LDL-C control after 1 year. Physicians’ outpatient cholesterol management decisions may account for some of the disparity in cholesterol control between women and men but do not explain differences between black and white patients.

AN ASSESSMENT OF BREAST CANCER RISK CONDUCTED AT THE TIME OF SCREENING MAMMOGRAPHY.R.C. Burack1; N. Patel2; J. George1; R. Everson1. 1Wayne State University, Detroit, MI; 2Karmanos Cancer Institute, Detroit, MI. (Tracking ID #117308)

BACKGROUND:  The Breast Cancer Prevention Study demonstrated that tamoxifen use was associated with a 47% breast cancer risk reduction among women whose Gail model 5-year breast cancer risk was 1.7%. However, identification of these higher risk women is a challenge in the primary care setting. As an alternative, we assessed the feasibility of conducting Gail model risk assessment during mammography visits.

METHODS:  Breast cancer risk factor information was collected at the time of mammography visits. 5-year breast cancer risk was estimated for each woman based on the Gail model risk algorithm. The number of breast cancers expected and “prevented” over 5 years among higher risk women was estimated assuming a 47% risk reduction with chemoprevention.

RESULTS:  Among 18,817 visitors, 14,174 (8,568 African-American [AA] and 5,506 White) women presented for screening mammography. 5.3% of AA and 31.0% of White women had Gail scores 1.7%. The “number needed to screen” (NNS) to identify higher-risk women is <50 for all age groups of AA women and <10 for all White sub-groups. However, the “number needed to screen and treat” (NNT) with tamoxifen is substantially higher for all groups.

CONCLUSION:  The observed age-ethnicity differences in Gail score risk reflect variation in the distribution of risk factors and breast cancer incidence. Risk assessment at the time of mammography offers an effective method to identify higher-risk women. This information can then be shared with the referring primary care provider who can use it in support of risk-related counseling.

inline image

AN ASSESSMENT OF MEDICAL STUDENT KNOWLEDGE AND LEARNING ABOUT THE HEALTH CARE SYSTEM.J.R. Agrawal1; J. Huebner2; J. Hedgecock3; A. Sehgal4; P. Jung5; S. Simon6. 1Brigham and Women's Hospital, Cambridge, MA; 2University of Washington, Seattle, Seattle, WA; 3American Medical Student Association, Reston, VA; 4Case Western Reserve University, Cleveland, OH; 5University of Maryland College Park, College Park, MD; 6Harvard Medical School, Boston, MA. (Tracking ID #116271)

BACKGROUND:  The health care delivery system and its problems remain important issues for future physicians. How well medical students learn about these topics is not well known.

METHODS:  We conducted a mail survey among a national probability sample of first-fourth-year medical students in the United States. We received responses from 295 first-year students (57%) and 475 fourth-year students (56%).

RESULTS:  When asked where the US ranked in the 2000 World Health Organization's ranking of 191 nations based on health systems performance, only 11% of respondents correctly identified the answer (near 40th place), while 89% overestimated US performance in this assessment. 32% of respondents incorrectly asserted that the US had the highest life expectancy of any nation, and over one-third (34%) reported (incorrectly) that the US had the lowest infant mortality rate. Nearly all were aware of the adverse health consequences of uninsurance and the fact that the numbers of uninsured have risen over the past decade, but 40% of both first- and fourth-year students underestimated the numbers of uninsured individuals in the US. 27% were not aware that the US has the highest health cost per-person of any nation, and 61% were not aware that the administrative costs of private health insurance exceed that of Medicare. Of note, fourth-years did not perform any better than first-years on knowledge questions. 96% of all respondents felt that knowledge of health policy is important to their career as a physician, and over half expressed dissatisfaction with their teaching on the subject in medical school, with fourth-years more significantly more likely to express this dissatisfaction than first-years (57% vs. 49%, P < .05). 89% of all respondents wanted increased exposure to health policy with fourth-years significantly more likely to favor additional required medical school coursework (67% vs. 48%, P < .05).

CONCLUSION:  Our study suggests that medical students demonstrate some gaps in basic knowledge concerning the US health care system particularly in the areas of cost, finance, quality, and access, and moreover that these deficiencies are not addressed in the medical school classroom.

AN EMPIRICALLY DERIVED TAXONOMY OF FACTORS AFFECTING PHYSICIANS’ WILLINGNESS TO DISCLOSE MEDICAL ERRORS.L.C. Kaldjian1; E.W. Jones1; T. Tripp-Reimer1; S.L. Hillis1; G.E. Rosenthal1. 1University of Iowa, Iowa City, IA. (Tracking ID #116005)

BACKGROUND:  Voluntary disclosure of medical errors by physicians is necessary if patient safety is to improve, but the factors that motivate or impede disclosure are complex. We employed qualitative methods to develop a taxonomy of factors that affect physicians’ willingness to disclose errors.

METHODS:  A MEDLINE search of literature from 1975 to 2003 was performed to identify factors that facilitate or impede disclosure. Articles were included if they had a first author, focused on the clinical experience of physicians, and discussed error disclosure or reporting. 20% of the articles were analyzed for inter-rater and intra-rater reliability for classifying factors identified in individual articles. We then conducted 5 exploratory focus groups (medical students, interns, residents, academic and community physicians) to identify additional factors and vernacular expressions. Factors identified by the focus groups were combined with those from the literature review and then reduced by aggregating similar factors together in order to facilitate pile sorting that was performed by 20 participants. Pile sorting results were subjected to hierarchical cluster analysis and a scheme of 8 clusters was selected. Pile-sorting participants were then invited to one of 3 focus groups to validate the selected cluster scheme.

RESULTS:  606 articles underwent full review and 316 articles met inclusion criteria, revealing 93 factors (35 beliefs and values and 58 fears and barriers). The 5 exploratory focus groups identified 16 additional factors, which were added to the results of the literature review for a total of 109 items. To facilitate pile sorting, similar factors were aggregated together, reducing the number of factors for sorting from 109 to 61. Cluster analysis of the item piles led to the selection of an 8-cluster scheme. The process of validation of this scheme led to relocation of 4 items to different clusters and dismissal of 2 items (1 redundant and 1 refractory to categorization). The 8 domains of the final taxonomy (with the number of items in each domain in parentheses) are as follows. Factors that facilitate disclosure: Respect for Patients (4), Personal Integrity (13), Improving Professional Relationships and Education (5), Improving Care of Patients and Communities (6). Factors that impede disclosure: Attitudinal Barriers (8), Uncertainties (4), Helplessness (8), Fears and Anxieties (8). For example, the domain “Personal Integrity” contained: accountability, duty, courage, integrity, altruism, humility, compassion, following conscience, doing the right thing, golden rule, alleviating guilt, pursuing forgiveness, and religious faith or spiritual beliefs. The domain “Uncertainties” contained: uncertainty about how to disclose errors, uncertainty about which errors should be disclosed, causal uncertainty regarding the causes of adverse events, and disagreements about errors between supervisors and trainees.

CONCLUSION:  Factors affecting the voluntary disclosure of medical errors are many and complex. Our empirically derived taxonomy describes this complexity and organizes it into 8 domains, providing a conceptual framework for measuring the impact different factors may have on disclosure of errors to health systems, fellow professionals, and patients; it is relevant to patient safety, medical education, and professionalism and can help design voluntary error reporting systems and educational interventions.

AN EVALUATION OF THE ABUSE ASSESSMENT SCREEN, A BRIEF SCREENING TOOL FOR DOMESTIC VIOLENCE, IN AN URBAN GENERAL MEDICINE POPULATION.C. Foldes1; D.R. Korenstein2; J. Wisnivesky2; A. Frieman2; T. McGinn2; A.L. Siu2. 1Baylor College of Medicine, Houston, TX; 2Mount Sinai School of Medicine, New York, NY. (Tracking ID #115020)

BACKGROUND:  The prevalence of domestic violence (DV) has been measured in multiple medical settings, but has not been well studied in a predominately middle-age Hispanic and African American urban general medicine population. Physicians rarely screen for DV, often citing time as a limiting factor. This study has three main objectives (1) to evaluate the performance characteristics of the Abuse Assessment Screen (AAS), a validated 4 item screening tool for lifetime DV, (2) to determine the prevalence of lifetime DV in an older multi-racial urban general medicine population, and (3) to measure predictors of lifetime DV.

METHODS:  In this cross-sectional study, the AAS was administered in English and Spanish to 305 randomly selected women attending a general medicine practice. The prevalence of lifetime DV was measured using the AAS. A positive response to 1 question is considered a positive screen for DV. The performance characteristics of the AAS were also measured. Logistic regression analysis was used to estimate adjusted odds ratios (OR) and 95% confidence intervals for the factors associated with lifetime DV.

RESULTS:  Participant ages ranged from 21–85 years (median 52). 43.9% were Hispanic, 40.7% were African American, and 7.9% were Caucasian. Prevalence of lifetime DV using the AAS was 45.9% (95% CI 40.2–51.7). A DV history was associated with having a partner with a substance abuse history [OR 5.6 (95% CI 3.1–10.2)], a mother with a DV history [OR 3.3 (95% CI 1.9–5.9)], poorer self-perception of health [OR 2.1 (95% CI 1.1–3.9)], and not being married or widowed [OR 1.9 (95% CI 1.04–3.6)]. Speaking primarily Spanish was protective [OR 0.49 (95% CI 0.2–1.0)]. The sensitivity of the first question of the AAS, AAS#1, compared to the entire AAS was 97.9% (95% CI 93.9–99.6%).The likelihood ratio (LR-) for a negative AAS#1 is 0.021.

CONCLUSION:  In this study of predominately middle-age Hispanic and African American women, the prevalence of lifetime abuse is similar to other studies in general medicine settings, 45.9%. We found AAS#1 to be highly sensitive and specific. To our knowledge this is the first evaluation of how one question performs to detect lifetime DV. Using a lifetime prevalence of DV of 45.9% and the LR- of 0.021, the post-test probability of a negative response to AAS#1 is 1.8%, essentially ruling out a history of DV. This question, “Have you ever been emotionally or physically abused by someone important to you?”, could easily be incorporated into a general medicine provider's practice.

AN EVALUATION OF THE VA COLORECTAL-CANCER SCREENING PERFORMANCE MEASURE: DOES IT MEASURE QUALITY OF CARE?L.C. Walter1; N.P. Davidowitz1; P.A. Heineken1; K.E. Covinsky1. 1VA Medical Center, San Francisco, CA. (Tracking ID #103221)

BACKGROUND:  The Department of Veterans Affairs (VA) manages the largest health care system in the U.S., and the Institute of Medicine has recommended many principles of VA quality measurement be applied to the U.S. health care system as a whole. VA measures quality of care by calculating scores on several performance measures, such as colorectal-cancer (CRC) screening. The objective of this study was to determine whether compliance with the national VA CRC performance measure reflects quality medical care.

METHODS:  We reviewed medical records of all 229 patients audited for CRC screening in 2002 at a VA facility by national VA auditors. We abstracted data to calculate the CRC performance measure (defined by the VA Technical Manual as having fecal occult-blood testing within 1 year, sigmoidoscopy within 5 years, or colonoscopy within 10 years among patients aged ≥52 yeas with life expectancies >6 months). We determined the frequency in which there were valid reasons for not screening patients, which included poor prognosis (Charlson comorbidity-age combined risk score ≥4) and patient refusal.

RESULTS:  The mean age of audited patients was 69 years (35%≥75 years). Auditors calculated a CRC screening rate of 58%. This score indicated poor quality of care because it was less than the national goal of 65%. However, 69% of patients had a Charlson comorbidity-age combined risk score ≥4, which is associated with >50% chance of death within 10 years and illustrated by such patients as an 89 year-old woman with severe dementia, a 94 year-old man with metastatic prostate cancer, and a 76 year-old man on dialysis who died prior to the audit date. Of the 97 patients counted as receiving poor quality care for not being screened, 37% declined screening, 10% failed to go to scheduled tests, and 9% had documentation by clinicians that screening was contraindicated. Auditors missed screening documentation in 21%. Of the 132 “screened” patients, 42% had tests for diagnostic purposes.

CONCLUSION:  The VA CRC performance measure is a poor indicator of the quality of medical care. Many audited patients classified as receiving poor quality care for not being screened had valid reasons, such as poor prognosis or patient refusal, and screening these patients would have represented poor quality care. Many patients classified as receiving good quality of care had CRC testing for diagnostic purposes rather than because of screening recommendations. The VA should consider significantly changing the CRC performance measure since it classifies high testing rates as good care regardless of who received it, why it was done, or whether the patient wanted it.


BACKGROUND:  Evidence-Based Medicine (EBM) is a required component of Internal Medicine Residencies’ curriculum. The “best” method of teaching EBM has not yet been identified. The objectives of most residency programs’ EBM Journal Clubs (JC) are to teach critical appraisal skills and keep up with current literature. JCs have been shown to improve biostatistic and critical appraisal skills, but have not been evaluated for improving broader EBM knowledge including asking clinical questions and searching the literature. The objectives of this study are to: 1) evaluate the effect of an EBM JC on residents’ EBM knowledge using a knowledge assessment test; and 2) evaluate the effect of level of training and degree of participation in JC on residents’ change in EBM knowledge.

METHODS:  In August of 2001 the University of Wisconsin Internal Medicine Residency Journal Club was modified to incorporate a comprehensive EBM curriculum. The JC is a twice-monthly conference comprised of two components. During the first 40 minutes a PGY-3 presents a clinical scenario followed by a clinical question, search strategy, critical appraisal of a retrieved article and how to apply the evidence in patient care. During the second portion of the Journal Club a faculty member presents a brief EBM core topic in an interactive format. Residents’ EBM knowledge was assessed by a 29 question multiple choice test developed to evaluate the implementation of this new EBM curriculum: the test includes questions about asking clinical questions, search strategies, critical appraisal, interpreting trial results, and study types. PGY-1, PGY-2 and PGY-3 residents took the EBM knowledge assessment test in June of 2001 prior to the implementation of the new JC and again in June of 2002.

RESULTS:  Nine PGY-3s who were presenters showed weak evidence of improvement in EBM knowledge with a mean baseline score of 14.3 out of 29 (95% CI 9.9 to 18.7) and a mean increase of 5.0 correct answers (CI 0.0 to 10.0, P = .08). 23 PGY-1s and 23 PGY−2s who were participants showed no evidence of improvement with a mean change of −0.3 (CI −1.9 to 1.3), and 1.3 (CI −1.5 to 4.1), respectively.

CONCLUSION:  An EBM JC that incorporates didactic teaching and an EBM curriculum improved the EBM knowledge of presenters at JC. However, this format did not appear to improve participants’ (PGY-1 or PGY-2s’) EBM knowledge. Because previous studies have shown that workshops can change critical appraisal skills, we plan to add a half-day EBM workshop during the PGY-2 year and retest for improvement after this intervention. We will also evaluate other residency teaching venues to reinforce the EBM curriculum.

AN EXPLORATION OF THE INTERACTION OF RATIONING AT THE PROVIDER AND SYSTEM-WIDE LEVELS. S. Hurst1; R. Forde2; D. Pegoraro3; A. Perrier4; S. Reiter-Theil5; A. Slowther6; M. Danis7. 1Centre Medical Universitaire, Geneva, Switzerland; 2The Norwegian Medical Association, Oslo, ; 3Fondazione Lanza, Padova, ; 4Geneva University Hospital, Geneva, ; 5Universitaet Basel, Basel, ; 6Ethox Institute of Health Sciences, Headington, ; 7National Institutes of Health, Bethesda, MD. (Tracking ID #116032)

BACKGROUND:  Limited heath care resources are a ubiquitous reality that have been studied at the organizational level and at the level of the individual clinician. Yet little research has focused on the interactive effects of rationing at these two levels.

METHODS:  We developed a survey instrument to examine the frequency, criteria and strategies used of bedside rationing, and the perceived availability and equity of access to resources in the health care system. Content validity was assessed through expert assessment and scales were tested for internal consistency reliability. The questionnaire was translated into Norwegian, and back translated to English to assess accuracy. We administered the survey instrument to general internists in Norway. Data were analyzed using descriptive statistics and correlations were assessed by Spearman's rank correlation.

RESULTS:  Scales for bedside rationing, rationing criteria, perceived scarcity, perceived equity were acceptably reliable (Cronbach's alpha = .84, .79, .82, .62 respectively). Survey respondents (N = 127) ranged in age from 27–74, were predominantly male (87%), and were largely outpatient-based (79%). Although only 31% of physicians agreed that they should ration, 85% reported that they did ration in their practices, with time being the most frequently rationed resource. Most frequent strategies for rationing expensive resources were explanation to patients (89%), substitution of less expensive options (82%), refusal of requests (81%), and delay of interventions (71%). All respondents perceived some resources as scarce with the most prominent being: access to nursing home, mental health services, and rehabilitation for stroke. Respondents witnessed adverse outcomes from rationing infrequently, but a small minority of respondents had encountered severe adverse events: death (17%) or permanent disability (5%). While 96% of respondents thought everyone in Norway should have equal access to needed medical services and thought they were given enough means to treat their patients according to their best interest (74%) and fairly (76%) , 55% thought access was not equal. Frequency of rationing by clinicians was correlated with perceived scarcity of resources (Spearman's rho, P = .000).

CONCLUSION:  Through the development of a survey instrument to examine rationing at the provider and health system levels, we have observed a relationship between provider rationing and perceived system-wide scarcity. Results from Norway, which provides universal health insurance, may serve as a benchmark for studies in other countries.

AN OBJECTIVE ASSESSMENT OF THE EFFECTIVENESS OF AN EBM FACULTY DEVELOPMENT PROGRAM ON CLINICIANS’ KNOWLEDGE.D.A. Feldstein1; K.M. Hla1; C.L. Gjerde1; B.M. Anderson1; P.K. Kokotailo1. 1University of Wisconsin-Madison, Madison, WI. (Tracking ID #116570)

BACKGROUND:  Community physician preceptors are increasingly responsible for teaching evidence-based medicine (EBM) to students. Primary care community physicians often do not have a good understanding of EBM concepts and do not incorporate EBM into their teaching. The Primary Care Faculty Development Program (PCFDP) teaches EBM to practicing clinician preceptors. The goal of this yearlong program is to make the participants proficient in using and teaching EBM. We previously determined efficacy of the program by participants’ self-assessments of pre- and post-program EBM skills. In the few studies assessing practicing physicians’ improvement in EBM skills after an educational intervention only one major study has looked at objective improvement. This was following a 3 day intensive workshop. There is bias in self-assessment of skills, and the effectiveness and retention of a longer intervention remains unclear. The main objective of this study was to determine participant's change in EBM knowledge following the EBM curriculum using an objective knowledge assessment test.

METHODS:  The EBM curriculum is taught in four sessions throughout the year- comprising 18 hours of teaching. Teaching formats used include large group, small group, in-class problem solving and computer-based hands-on sessions. The EBM knowledge assessment test was developed from a test previously used to measure resident EBM knowledge. It contains 25 multiple choice questions covering the following areas: 1) asking clinical questions; 2) searching; 3) EBM resources; 4) critical appraisal; 5) calculating ARR, RRR, NNT; 6) characteristics of diagnostic tests and post-test probability of disease; and 7) interpreting confidence intervals. Participants for the 2002-2003 year were given the test at the beginning of the first session prior to any EBM teaching (pre) and again after the last session (post) 10 months later.

RESULTS:  16 of 22 (73%) participants took both pre- and post-tests and were included in the final analysis. There was a significant improvement in the participants EBM test scores. The mean (and 95% CI) of correct answers on the pre-test was 13.6 (11.9–15.2), and on the post-test was 19.4 (18.2–20.6). This was a mean increase of 5.8 correct answers (4.2–7.4). The relative increase in correct answers following the intervention was 43%.

CONCLUSION:  Our data demonstrates that community primary care physicians significantly increased their knowledge across a wide range of EBM skills through our yearlong program. The length of the program indicated that there was good retention of EBM knowledge at ten months using this format. We plan to retest participants six months after completion of the program to see if there is continued retention.

ANTIBIOTIC PRESCRIBING HAS DECREASED:DIFFERENCES REMAIN AMONG PHYSICIANS, RESIDENTS, AND MIDLEVEL PROVIDERS.C.L. Roumie1; N.B. Halasa1; K. Edwards1; Y. Zhu1; R.S. Dittus1; M.R. Griffin1. 1Vanderbilt University, Nashville, TN. (Tracking ID #115978)

BACKGROUND:  We measured trends in antibiotic prescribing for adults, and determined if physicians, nurse practitioners/physician assistants (NP/PA), and resident physicians had similar prescribing patterns.

METHODS:  Antibiotic prescribing per 1,000 adults during all outpatient visits and visits for respiratory conditions where antibiotics are rarely indicated (acute bronchitis, cough, upper respiratory tract infection, asthma, allergic rhinitis) were estimated using the National Ambulatory Medical Care Survey (NAMCS) and the National Hospital Ambulatory Medical Care Survey (NHAMCS), 1995 through 2000. Factors associated with inappropriate prescribing for respiratory diagnoses including time period, provider type, and additional provider and patient characteristics were evaluated using multivariate logistic regression.

RESULTS:  From 1995–1996 to 1999–2000, visits associated with antibiotic prescriptions decreased 4%, resulting in 12.6 fewer antibiotic prescriptions per 1000 adults. This decrease was confined to visits for respiratory diagnoses where antibiotics are not indicated. The proportion of visits associated with an antibiotic prescription fell from 42% to 33% (adjusted Odds Ratio [OR] 0.66, 95% Confidence Interval [CI] 0.52 to 0.82). There was no compensatory increase in antibiotic prescribing associated with other diagnoses. Compared to physicians, who prescribed antibiotics for 39% of such visits, NP/PAs and resident physicians prescribed antibiotics for 50% (adjusted OR 1.80, 95% CI, 1.23 to 2.64) and 25% (adjusted OR 0.58 95% CI, 0.42 to 0.81) of visits, respectively.

CONCLUSION:  Inappropriate outpatient antibiotic prescribing for respiratory conditions has declined yielding a decrease in antibiotics use among adults. Variations in adherence to guidelines remain among different practitioners; successful campaigns reducing antibiotic use for specific respiratory diseases should be extended to all practitioners.

Table 1. Factors associated with antibiotic prescribing for Targeted Respiratory Diagnoses
Provider/Practice Characteristics% Prescribed antibioticOdds Ratio (95% CI)Adjusted Odds Ratio* (95% CI)
Provider MD38.51.0 (ref)1.0 (ref)
NP/PA49.81.58 (1.12, 2.24)1.80 (1.23, 2.64)
Resident24.80.53 (0.39, 0.71)0.58 (0.42, 0.81)

ANTICOAGULATION TESTING AND PATIENT SATISFACTION.R. Chaudhry1; R. Stroebel1; D. Dupras1; S. Scheitel1; E. Tangalos1. 1Mayo Clinic, Rochester, MN. (Tracking ID #115930)

BACKGROUND:  Point of Care INR (POCINR) testing by fingerstick (Coagucheck) combined with protocol directed, dosing instruction can be utilized for the monitoring of warfarin therapy for patients on chronic anticoagulation. POCINR testing has several advantages for patients including ease of access, improved turn-around time, less discomfort, and face-to-face review of written dosing instructions by the team RN. Patient satisfaction with the new technology in a face-to-face clinical encounter has not yet been ascertained.

METHODS:  The Division of Community Internal Medicine (CIM) in collaboration with Laboratory Medicine implemented POCINR testing for patients on chronic anticoagulation in 2002. Prior to POCINR testing patients had their ProTime measured by venipuncture and were provided with the results protocol directed instructions on warfarin dosage by a telephone call from a RN. 216 consecutive patients were surveyed one month after the POCINR fingerstick testing and face-to-face counseling were implemented.

RESULTS:  A total of 187 (87%) patients responded to the survey. When comparing the patients’ past experiences with venipuncture and telephone management to POCINR, 77% had less pain, 81% reported blood collection to be easier, 96% reported receiving results faster, 68% reported improved communication on warfarin dosing, 76% reported improved opportunity to ask questions, 74% reported improved ease of making future appointments, 91% of patients reported overall satisfaction with the new system and 94% of patients preferred POCINR testing and face-to-face counseling. All results were statistically significant (P < .0001).

CONCLUSION:  The overwhelming majority of our patients on chronic anticoagulation preferred POCINR testing and face-to-face counseling over venipuncture (ProTime) and telephone management.

ANTIVIRALS AND ANTIBIOTICS FOR INFLUENZA IN THE UNITED STATES, 1995–2001.J.A. Linder1; D.W. Bates1. 1Division of General Medicine, Brigham and Women's Hospital, Boston, MA. (Tracking ID #101660)

BACKGROUND:  Antiviral medications cost-effectively reduce influenza-related morbidity and potentially mortality. Antibiotics may be indicated for some patients with influenza, but concern exists about antibiotic overprescribing. Our goals were to: 1) measure the rates of antiviral and antibiotic prescribing for patients diagnosed with influenza and 2) evaluate independent predictors of antiviral and antibiotic prescribing.

METHODS:  The National Ambulatory Medical Care Survey and the National Hospital Ambulatory Medical Care Survey (NAMCS/NHAMCS) are probability samples that collect data from physician practices and emergency departments throughout the United States. We analyzed visits with a diagnosis of influenza that occurred in the 6 influenza seasons from October 1, 1995 to May 31, 2001 (n = 1,050). We used SUDAAN software, which accounts for the complex sampling design of the NAMCS/NHAMCS, to measure the rates of anti-influenza antiviral medication (amantadine, rimantadine, zanamivir, and oseltamivir) and antibiotic prescribing. We evaluated independent predictors of antiviral and antibiotic prescribing using multivariable logistic regression.

RESULTS:  There were 18.6 million visits (95% confidence interval [CI], 14.5 to 22.7 million visits) with a diagnosis of influenza in the U.S. from the 1995–1996 season to the 2000–2001 season. The sample had a mean age of 34, was 44% male, and 85% white. Visits were to community outpatient practices (88%), emergency departments (8%), and hospital-based outpatient departments (4%). Of community outpatient practice visits, 54% were to family practice physicians, 20% were to pediatricians, and 19% were to internists. Physicians prescribed antivirals in 18% of visits and antibiotics in 30% of visits. There was no significant linear change in antiviral or antibiotic prescribing over time. In multivariable modeling, the lone independent predictor of antiviral prescribing was Medicare insurance (odds ratio [OR], 0.1 versus private insurance; 95% CI, 0.0–0.6). Independent predictors of antibiotic prescribing were Medicaid insurance (OR, 2.0 versus private insurance; 95% CI, 1.1–3.6) and visits to emergency departments (OR, 0.5 versus community outpatient practice visits; 95% CI, 0.3–0.9). Men were marginally less likely to receive antibiotics than women (OR, 0.6; 95% CI, 0.4–1.0).

CONCLUSION:  Although cost-effective, antiviral medications were prescribed to only 18% of patients diagnosed with influenza. Antibiotics were prescribed to only 30% of patients. Antiviral and antibiotic prescribing both appear to be affected by non-clinical factors, possibly related to drug benefits. To reduce influenza-related morbidity and possibly mortality, efforts could be undertaken to increase appropriate antiviral prescribing, especially in seasons when vaccine delivery is delayed, there is insufficient vaccine supply, or there is poor influenza-vaccine matching.

ARE CLINICIAN-EDUCATORS AND CLINICIAN-INVESTIGATORS PROMOTED AT THE SAME RATE? THE PROSPECTIVE STUDY OF PROMOTION IN ACADEMIA.B. Beasley1; S. Simon2; S.M. Wright3. 1University of Missouri-Kansas City, Kansas City, MO; 2UMKC, Kansas City, MO; 3Johns Hopkins Bayview, Baltimore, MD. (Tracking ID #116455)

BACKGROUND:  Academic medical institutions are grappling with how to optimally evaluate and promote clinician-educators. Yet, clinician-educators have continued concerns about their promotability. No prior prospective studies of promotion have been performed across institutions to address this issue.

METHODS:  In the Spring of 2000, we began to prospectively study 183 junior Department of Medicine faculty who volunteered to participate. The faculty were at the rank of Assistant Professor from 80 institutions in 35 states. We obtained baseline demographics, job characteristics, institutional characteristics, the factors that motivate these individuals in their work, and the systems that are available to them to support scholarship. In the spring of 2003, we contacted all 183 faculty to determine the primary outcome measure: whether they had been promoted to Associate Professor and the date of their promotion.

RESULTS:  We classified 107 (58%) participants as clinician-educators (CE’s) and 63 (34%) as clinician-investigators (CI’s) by either their self-selection or by set criteria. Thirteen participants (7%) were unable to be placed into these categories. The 7th-year promotion rate for CE's was 32%, while for CI's it was 53% (Hazard Ratio for CE's compared to CI’s = 0.7, 95% CI 0.6–0.9). Other variables that were significant positive predictors of promotion included meeting with Chairman/Chief more than yearly, (HR = 1.7, 95% CI 1.2–2.4 compared to “never meet”), working more than 60 hours per week (HR = 1.5, 95% CI 1.1–1.9 compared to 50 or less), having less than 10% patient care without learners (HR = 1.4, 95% CI 1.1–1.8 compared to more than 30%), having seen the institution's promotion guidelines (HR = 1.5, 95% CI 1.2–1.9), and having a grant management office (HR = 1.2, 95% CI 1.0–1.5). Significant negative predictors included having 5% or less research time (HR = 0.6, 95% CI 0.4–0.8 compared to greater than 30%), being located in the Northeast (HR = 0.7, 95% CI 0.5–0.9 compared to other regions), and having a low job satisfaction score (HR = 0.7, 95% CI 0.6–0.9 compared to high). Variables that were not significant included gender, race, medical school class rank, being AOA, and family responsibilities. Multivariable analyses are being performed, and these will offer explanations as to why CE's are promoted at a diminished rate.

CONCLUSION:  Differences were noted between clinician-educators’ and clinician-investigators’ promotion rates. These may be explained by particular job characteristics, systems to support scholarship, and institutional variables.

ARE PHYSICIANS LESS LIKELY TO DISCLOSE SERIOUS ERRORS THAT ARE NOT APPARENT TO PATIENTS?T.H. Gallagher1; A.D. Waterman2; J.M. Garbutt2; I. Fischer2; A.G. Ebers2; M. Krauss2; W.C. Dunagan2; W. Levinson3; V.J. Fraser2. 1University of Washington, Seattle, WA; 2Washington University in St. Louis, St. Louis, MO; 3University of Toronto, Toronto, Ontario. (Tracking ID #117144)

BACKGROUND:  Disclosure of errors is desired by patients, recommended by ethicists, and required by some state laws and accreditation standards. However, little is known about what factors influence whether physicians actually disclose errors.

METHODS:  We surveyed all 1505 practicing physicians in internal medicine and family practice at Washington University/BJC Healthcare (St. Louis), University of Washington, and Group Health Cooperative (Seattle) as part of a survey of 2,583 physicians. The response rate to date is 56%. Respondents randomly received one of two scenarios: an overt error (10-fold insulin overdose and hypoglycemic arrest) and an error that would be less apparent to a patient (failure to check lab test ordered after starting medicine resulting in hyperkalemia and cardiac arrest). In both scenarios the patient recovers uneventfully. 5 closed ended questions assessed how much information respondents would disclose about the error. Respondents’ attitudes about error disclosure and previous experience disclosing errors were also assessed. All associations are significant at P < .01.

RESULTS:  Respondents were predominately male (76%), and had been practicing an average of 16 years. 51% were in private practice. 98% agreed that serious errors should be disclosed to patients, and 50% reported having personally disclosed a serious error. Physicians disclosed less information in response to the inapparent error (hyperkalemia) compared with the overt error (insulin overdose). For example, physicians in the hypkeralemia scenario were less likely to chose a disclosure statement that made explicit reference to the error compared to the insulin scenario (39% vs. 70%). Hyperkalemia scenario respondents were also more likely not to mention the error's cause (25% vs. 10%). In both scenarios respondents disclosed significantly less information if they did not strongly agree that serious errors should be disclosed, if they felt disclosure would risk a lawsuit, if they felt less responsible for the hypothetical error, if they agreed that they might not disclose an error if the patient was unaware that the error happened, and if they were in private vs. academic practice. In addition, physicians disclosed more information if they had previously disclosed a serious error, were satisfied with their most recent serious error disclosure, and if they felt relief after disclosing this error to the patient.

CONCLUSION:  Characteristics of an error, such as whether the error is overt vs. inapparent to the patient, as well as physicians’ attitudes about and experience with disclosure, may affect how physicians discuss medical errors with patients. Accounting for such factors in their policies and training may help patient safety programs improve the disclosure process.

ARE PRIMARY CARE PATIENTS WITH AT-RISK DRINKING READY TO CHANGE?E. Williams1; D. Kivlahan1; R. Saitz2; J. Merrill1; K. Bradley3. 1University of Washington, Seattle, WA; 2Boston University, Boston, MA; 3VA Puget Sound, Seattle, WA. (Tracking ID #117376)

BACKGROUND:  Patients’ readiness to change may influence the effectiveness of interventions for at-risk drinking. We evaluated readiness to change among Veterans Affairs (VA) primary care patients who screened positive for at-risk drinking.

METHODS:  Male outpatients at 7 VA GIM clinics were eligible who returned a mailed survey, screened positive for at-risk drinking on the Alcohol Use Disorders Identification Test (AUDIT) consumption questions (≥4 points), and responded to 3 questions about readiness to change. Patients were categorized into stages of change (precontemplation, contemplation, action) based on a validated algorithm. The AUDIT was used to categorize patients based on severity of drinking problems: <8 (least severe), 8–15, 16–19; and 20–40 points (most severe). We used logistic regression to model alcohol-related variables associated with readiness to change (contemplation or action), adjusting for demographic characteristics and self-reported medical and psychiatric comorbidity.

RESULTS:  Of 4,110 patients with at-risk drinking, 83% were classified in contemplation or action categories (34% and 49% respectively). The prevalence of contemplation or action increased as AUDIT scores increased (Figure). Using logistic regression, the severity of drinking problems was a strong predictor of contemplation or action (Adjusted OR 4.8 for AUDIT scores 8–15, CI 3.7–6.0; AOR 12.7 for AUDIT 16–19, CI 5.1–31.4; AOR 27.3 for AUDIT 20–40, CI 8.6–87.9). Other significant predictors were prior alcohol treatment (AOR 1.7; CI 1.3–2.2) and advice to decrease or stop drinking from a primary care provider in the past year (AOR 3.1; CI 2.2–4.3).

CONCLUSION:  Patients with more severe drinking problems, and those reporting prior alcohol-related treatment or advice are more likely to report readiness to change. However, even among patients with the least severe drinking problems, most at-risk drinkers report either contemplating change or actively trying to decrease their drinking. These data suggest reason for optimism in disseminating brief interventions for at-risk drinking in primary care settings.

inline image

ARE SENIOR MEDICAL STUDENTS PREPARED FOR THE ACGME COMPETENCIES? STUDENTS’ SELF-ASSESSMENT OF THEIR TEACHING SKILLS.D. Torre1; J. Zebrack1; D. Simpson1; J. Sebastian1; R.C. Anderson2; J. Jevtic1; J. Beaversdorf1. 1Medical College of Wisconsin, Milwaukee, WI; 2Medical College of Wisconsin, Pewaukee, WI. (Tracking ID #115977)

BACKGROUND:  Skills in self-assessment and teaching are strongly associated with three of the ACGME core competencies: communication, professionalism and practice-based learning/improvement (facilitate the learning of students and other health professionals). While many medical schools offer opportunities for fourth-year medical students (M4) to teach, most M4 students have few formal opportunities to practice self-assessment skills specific to their role as teachers. We therefore designed a teaching exercise to determine the extent to which high achieving M4 students accurately assess their own teaching skills.

METHODS:  Thirteen M4 student volunteers, all members of our local chapter of the Alpha Omega Alpha Honor Medical Society, served as trained evaluators during an objective structured clinical exam (OSCE) that was administered to all second year medical (M2) students at the end of their Introduction to Clinical Examination Course. M4 students used a standardized checklist to evaluate the proficiency of M2 students (n = 190) in performing an abdominal, cardiovascular and chest examination on a standardized patient. After completing the formal checklist, M4 students were instructed to provide M2 students with additional formative feedback and “coaching.” At the end of the OSCE session, all M2 students completed a 12-item form to evaluate the teaching effectiveness of their M4 student preceptor. Using a parallel form, M4 students assessed their own teaching skills immediately after each session

RESULTS:  Each participating M2 and M4 student returned an evaluation form that used a 5-point Likert scale (5 = excellent, 1 = poor) to assess specific components of the teaching encounter. Comparison of ratings showed significant differences with higher M2 ratings compared to M4 self-assessments in each of the relevant ACGME competency areas such as communication (e.g. asked/answered questions clearly)[P < .01], professionalism (provided constructive feedback without belittling me)[P < .01] and practice-based learning (took advantage of teaching opportunities)[P < .01]. With respect to overall teaching effectiveness, the M2 evaluations of their M4 teachers were significantly higher than the M4 self-assessment ratings (4.9 vs. 4.1, p value < .001).

CONCLUSION:  Consistent with the literature in self-assessment, high achieving M4 students underestimated their competence as educators. In preparation for residency, inclusion of self-assessment activities within a variety of instructional venues may help enhance students’ proficiency as teachers and physicians

ARE THE INCENTIVES IN INCENTIVE-BASED PHARMACEUTICAL BENEFIT PLANS OPERATING AS INTENDED?W.H. Shrank1; R. Kravitz2; S. Ettner3; P.A. Glassman4; H.N. Young5; S.M. Asch6. 1UCLA/Greater Los Angeles VA Healthcare System, Los Angeles, CA; 2Univerity of California, Davis, Sacramento, CA; 3University of California, Los Angeles, Los Angeles, CA; 4Greater Los Angeles VA Healthcare System, Los Angeles, CA; 5University of California, Davis, Sacramento, CA; 6RAND, Los Angeles, CA. (Tracking ID #116112)

BACKGROUND:  A majority of Americans with pharmaceutical benefits are now enrolled in three-tiered, incentive-based pharmaceutical benefit plans. These plans use patient cost-sharing requirements as incentives to steer medication choices towards “preferred” formulary options that are less expensive for insurers. Three-tier plans are predicated on the following assumptions: 1) physicians believe it is important to control patients out-of-pocket costs, 2) physicians feel accountable for controlling patients’ costs when prescribing, 3) physicians are aware of patient's out-of-pocket costs when prescribing, and 4) physicians and patients discuss costs and benefits of treatment options at the time of prescribing. Little is known about whether these assumptions are being met. Recent Medicare legislation endorsing an increasing role of private health plans in the provision of prescription drugs to seniors underscores the importance of examining these assumptions directly.

METHODS:  A self-administered, written survey was offered to physicians participating in the California Medical Association Leadership Conference. 131 responses were received of the 205 participants present (64% response rate). Many of the non-respondents were not physicians and the response rate of physicians in the room was likely much higher.

RESULTS:  While 90% of physicians either strongly or somewhat agreed that is important to manage patient's out-of-pocket costs for prescription drugs, only 25% strongly or somewhat agreed that it is their “responsibility” to help. Physicians believe that they do not possess the necessary information to predict patients’ out-of-pocket expenses at the time of prescribing. 59% of physicians report that they never or seldom are aware of the “preferred” formulary option when prescribing, and 70% never or seldom are aware of patients’ out-of-pocket costs at the time of prescribing. Only 15% of physicians report discussing out-of-pocket costs with patients all or most of the time. Instead, 69% of physicians surveyed somewhat or strongly agreed that it is the responsibility of the pharmacist to be familiar with patients’ out-of-pocket costs. Nevertheless, 67% of physicians either strongly or somewhat agreed that phone calls from pharmacists are “bothersome.”

CONCLUSION:  Physicians report that several basic assumptions of the three-tiered, incentive-based pharmaceutical benefits structures are not currently met. Physicians do not have the necessary information at the time of prescribing to help patients manage their out-of-pocket costs, and they do not hold themselves accountable for doing so. Instead, physicians rely on patients to respond to their financial incentives at the pharmacy, communicating their price preferences indirectly and likely less efficiently. Further research should identify whether improved information systems to assist physicians at the time of prescribing or simplified benefit structures would allow the incentives in tiered pharmacy benefit plans to operate as intended.

ARE THE TYPE AND NUMBER OF FOOD STORES IN A NEIGHBORHOOD ASSOCIATED WITH THE HEALTH OF ITS RESIDENTS?A.F. Brown1; A. Ang1; A.R. Pebley1. 1University of California, Los Angeles, Los Angeles, CA. (Tracking ID #116523)

BACKGROUND:  Residence in poor neighborhoods has been associated with poorer access to supermarkets, but greater access to small markets and convenience stores. In prior research, small markets were found to have limited fruits, vegetables, and whole grain products and to charge higher prices. We studied the relationship between the type and number of food stores in a neighborhood and self-rated health.

METHODS:  The Los Angeles Family and Neighborhood Study (LA FANS) uses a representative sample of 65 neighborhoods in Los Angeles County, with an oversample of poor areas, to evaluate area influences on health. Food stores in each census tract were identified using commercial data and were categorized as supermarkets, small markets, convenience stores, and specialty stores using North American Industry Classification System codes. Self-rated health was ranked from 1 (Poor) to 5 (Excellent). Neighborhood SES is an unweighted mean of data from the 2000 Census (poverty, education, employment, occupational status, and median housing value). Information on individual characteristics was obtained from in-person surveys. We constructed multilevel ordered logistic models to evaluate the independent influence of population-adjusted number and type of food sources on self-rated health after adjustment for neighborhood SES and individual age, sex, race/ethnicity, income, education, immigrant status, and chronic conditions.

RESULTS:  Among the 2623 adult respondents (response rate 70%) included in this analysis, mean age 40 years, health status was reported as excellent by 21%, very good 25%, good 31%, fair 19%, and poor 4%. There was a higher density of small markets and convenience stores in poorer neighborhoods. In adjusted analyses, lower self-rated health was associated with residing in neighborhoods with more small markets (OR = 0.91, P = .008). Residing in a neighborhood in the middle or lowest SES tertile was also associated with poorer health, as were lower individual income, less education, male sex, older age, non-immigrant status, and having a chronic condition.

CONCLUSION:  The type and number of food stores in a neighborhood is significantly associated with health, independent of other individual and neighborhood characteristics. A higher concentration of small markets, which tend to have less variety and charge higher prices than supermarkets, is associated with lower self-rated health. These findings may be particularly important for persons with a chronic condition, such as obesity, diabetes, or hypertension, where there is compelling evidence that diets high in fresh vegetables are associated with better outcomes. It is therefore important to evaluate the role of the availability of healthy foods and other neighborhood characteristics on health.

ASIAN AMERICANS’ HEALTH CARE EXPERIENCES AND RATINGS: DIFFERENCES BETWEEN VIETNAMESE AND CHINESE IMMIGRANTS.Q. Ngo-Metzger1; M. Massagli2; B. Clarridge3; D. Sorkin1; R.S. Phillips4. 1University of California Irvine, Irvine, CA; 2Massachusetts General Hospital, Boston, MA; 3UMass-Boston, Boston, MA; 4Harvard University, Boston, MA. (Tracking ID #115485)

BACKGROUND:  Asian Americans are a fast growing minority group. We surveyed a national cohort of limited-English proficient (LEP) Chinese and Vietnamese immigrants to evaluate their health care experiences.

METHODS:  We surveyed 4410 patients with office visits during the last month at 11 health centers in 8 U.S. cities. We measured demographics, health status, and health care experiences at the last visit and in the last 12 months. Patients could answer the mailed survey in English, Vietnamese, or Cantonese or Mandarin Chinese. We used logistic regression to examine how pts demographics and problem experiences (with front desk staff, access to care, communication, and trust) predicted ratings of the last visit and of care received in the last 12 months (excellent/ very good/ good vs. fair/poor), with separate analyses for Vietnamese, Cantonese, and Mandarin Chinese patients, adjusting for age, health status, and whether patients saw their usual doctors.

RESULTS:  We received 3258 surveys (response rate 74%): 1292 from Vietnamese; 1121 from Cantonese; and 678 from Mandarin Chinese patients. Ninety-nine percent were foreign-born, 90% were LEP,and half had <9 years of education. Vietnamese reported less problems with care compared to Chinese patients. One quarter of Vietnamese reported that doctors did not explain things in a way that they could understand, compared to 39% of Chinese patients (P < .001). Eleven percent of Vietnamese reported having problems trusting their doctors, as compared to 34% of Cantonese and 44% of Mandarin patients (P < .001). Vietnamese were less likely to rate the care received at the last visit and in the last 12 months fair or poor, and more likely to recommend the clinic to others (P < .001). In multivariable analyses, problems with front-desk staff (adjusted OR 7.2, 95% CI 4.7–11.1 for Cantonese; aOR 10.4, 95% CI 6.0–18.1 for Mandarin; aOR 26.6, 95% CI 13.1–54.0 for Vietnamese) and difficulty getting care (aOR 3.2, 95% CI 1.9–5.2 for Cantonese; aOR 3.4, 95% CI 1.8–6.3 for Mandarin; aOR 4.6, 95% CI 1.8–11.7 for Vietnamese) were important predictors of poor ratings of care received in the last 12 months. Not having trust in the doctor was a significant predictor of poor ratings of the last visit (aOR 5.3, 95%CI 3.0–9.4 for Cantonese; aOR 3.1, 95% CI 1.3–7.5 for Mandarin; aOR 15.1, 95% CI 5.9–38.4 for Vietnamese).

CONCLUSION:  In a national cohort of Asians, Vietnamese reported less problems and gave higher ratings of care than Chinese patients. For all patients, problems accessing care and with front desk staff were associated with poor ratings of care received in the last 12 months. Trust in the doctor was a significant predictor of last visit ratings.

ASPIRIN, STATINS, OR BOTH THERAPIES FOR PRIMARY PREVENTION OF MYOCARDIAL INFARCTION: A COST-UTILITY ANALYSIS.M. Pignone1; S. Earnshaw2; J.A. Tice3; M. Pletcher3; M. Wilson2. 1University of North Carolina at Chapel Hill, Chapel Hill, NC; 2RTI Health Solutions, Raleigh, NC; 3University of California, San Francisco, San Francisco, CA. (Tracking ID #116539)

BACKGROUND:  Aspirin and statin drugs are both effective in preventing myocardial infarction (MI) for patients with no history of previous cardiovascular events, but no studies have compared these agents in terms of their cost-effectiveness for primary prevention of coronary heart disease (CHD).

METHODS:  We developed a Markov model to compare the cost and utility of 4 strategies: aspirin, statins, both drugs, or neither drug among men and women without a history of cardiovascular disease at 4 different ages (35–45, 45–55, 55–65, 65–75) and 4 different 10-year CHD risk levels (2.5%, 7.5%, 15%, and 25%). The model was run over 20 years. All patients began in the healthy state and could transition annually to: non-fatal MI, angina, aspirin-related gastrointestinal (GI) bleeding, statin-related myopathy, stroke (including hemorrhagic stroke), or death. Patients were assumed to receive both aspirin and statins after MI and stroke. Treatment efficacy data and risk of adverse events were taken from published meta-analyses; utility estimates were drawn from published literature. Direct cost data were drawn from published literature, inflated to 2003 dollars. Specifically, annual costs of aspirin and statins were $16 and $713, respectively, based on Red Book prices. Costs and benefits were discounted at 3%.

RESULTS:  Among 45–55 year-olds, aspirin was both more effective and less costly compared with no treatment when the Framingham 10-year global CHD risk was greater than or equal to 7.5%. At 2.5% 10-year risk, the cost per quality adjusted life year (QALY) gained was $13,800 for men and $27,700 for women. Statins alone, compared with no therapy, had cost-utility ratios of $173,000, $89,700, $50,800, and $32,800/QALY at 10 year risk levels of 2.5%, 7.5%, 15%, and 25%, respectively, for men and $238,900, $96,200, $51,600, and $33,300/QALY for women. The addition of statins to aspirin increased effectiveness at an incremental cost (men/women) of $190,500 / $267,600, $103,100 / $113,100, $62,500 / $65,900, and $48,100 / $49,100 /QALY at risk levels of 2.5%, 7.5%, 15%, and 25%, respectively. Statins had similar cost-utility ratios to aspirin when statin cost was $200–350 per year.

CONCLUSION:  Aspirin is more cost-effective than statins as the initial drug for CHD prevention at all risk levels for men and women, and appears cost-saving for 45–55 year old patients with 10-year CHD risk of 7.5% or higher. At their current price, adding statins to aspirin is cost-effective when pre-treatment 10-year CHD risk is ≥25%.

ASPIRIN USE AMONG ADULTS WITH DIABETES: RECENT TRENDS AND EMERGING GENDER DISPARITIES.S.D. Persell1; D.W. Baker1. 1Northwestern University, Chicago, IL. (Tracking ID #114460)

BACKGROUND:  Despite high cardiovascular risk among adults with diabetes, aspirin use has been low in the past. It is not clear whether aspirin use increased following the 1997 American Diabetes Association position statement recommending aspirin for diabetic adults at increased risk for cardiovascular disease (CVD). We conducted this study to determine recent trends in the use of aspirin among patients with diabetes and to identify patient characteristics associated with underuse.

METHODS:  We conducted a time-series analysis using cross-sectional data from statewide telephone surveys conducted in 7 states in 1997 and 20 states in 1999 and 2001 as part of the Behavioral Risk Factor Surveillance System. This analysis included 875, 3205, and 4272 subjects in 1997, 1999, and 2001, respectively. The study population included adults age 35 or older who reported physician diagnosed diabetes but not only gestational diabetes. We classified individuals as regular aspirin users if they reported taking aspirin every day or every other day. To examine trends over time, we compared aspirin use in states for which data was available in consecutive years. We conducted a stratified analysis based on the presence or absence of diagnosed CVD and used logistic regression to assess the independent association of regular aspirin use with age, sex, race, education, income, and CVD risk factors.

RESULTS:  Aspirin use increased from 37.5% in 1997 to 48.7% in 2001. In 2001, 74.2% (95% CI, 70.9–77.5) of diabetic adults who reported physician diagnosed CVD, but only 37.9% (95% CI, 35.1–40.1) of those without CVD used aspirin regularly, including less than 40% of adults who had diagnosed hypertension, hypercholesterolemia, or who smoked. After adjusting for cardiac risk factors and socioeconomic characteristics, among those without CVD, aspirin use was less common among women aged 35–49 (adjusted odds ratio [OR] 0.22, 95% CI 0.14–0.35) and 50–64 (adjusted OR 0.53, 95% CI 0.37–0.80), and for men aged 35-49 (adjusted OR 0.46, 95% CI 0.28–0.75) compared to men >65 years old. For those with diagnosed CVD, aspirin use was lower among women (adjusted OR 0.42 compared to men, 95% CI 0.29–0.60), adults under age 50 (adjusted OR compared to >65 year-olds 0.52, 95% CI 0.29–0.94), and the uninsured (adjusted OR 0.46, 95% CI 0.24–0.87). The disparity in aspirin use between men and women appeared between 1997 and 2001.

CONCLUSION:  Aspirin use among adults with diabetes increased in recent years. However, use among diabetics without known CVD remains low, even among people with other risk factors for CVD. Women and people under 50 are less likely to be using this effective, inexpensive therapy.

ASSESSING RESIDENTS’ ATTITUDES TOWARDS A NIGHT FLOAT SYSTEM USING A MULTI-DIMENSIONAL SCALED QUESTIONNAIRE.H. Jasti1; B.H. Hanusa2; R. Granieri2; M. Elnicki2. 1Univ of Pittsburgh*, Pittsburgh, PA; 2Univ of Pittsburgh, Pittsburgh, PA. (Tracking ID #116683)

BACKGROUND:  To address residents’ working hours, a Night Float (NF) system has been instituted in residency programs across the country. There are few recent studies that have examined the perceptions and attitudes of residents in multiple broad areas of a NF system.

METHODS:  We developed a 115-item questionnaire to measure attitudes towards 6 potential domains of the NF rotation. We focused on 3 of these domains, namely patient care, education, and satisfaction. For each domain, a principal components analysis was done and factor scores were computed. Analysis of variance compared the means of the factor scores by postgraduate year of training, namely PGY 1 (interns) vs. PGY 2 and 3 (residents).

RESULTS:  The response rate was 90% (n = 149). The first domain of 12 questions dealt with patient care. Factor analysis yielded 3 factors: continuity of care, patient-physician relationship, and patient care by physician. The 3 factors explained 57% of the variance. The second domain, consisting of 10 questions reflecting residents’ thoughts on education during NF, generated 5 factors, explaining 78% of the variance. The third domain of 9 questions, dealing with overall satisfaction, resulted in one factor, explaining 60% of the variance. In the patient care domain, both interns and residents felt that continuity of patient care was preserved (mean 18.6 vs. 18.9, P = .67). However, interns were more likely than residents to feel that the patient-physician relationship was well maintained (mean 5.9 vs. 5.1, P = .01). Residents were more likely to think a “shift-work” mentality developed with NF (mean 6.7 vs. 5.4, P < .001). In the education domain, residents were more likely than interns to state that there was more emphasis on service (mean 3.7 vs. 3.3, P = .03) and less focus on learning about the impact of patient interventions (mean 7.3 vs. 6.5, P = .01). In the overall satisfaction domain, interns were more satisfied with the NF rotation than the residents (mean 35.5 vs. 30.7, P < .001).

CONCLUSION:  Our questionnaire was able to extract factors with good face validity from different domains assessing residents’ attitudes towards the NF system. While interns were more satisfied with the NF rotation, including the preserved patient-physician relationship and education, residents felt there was less of an emphasis on learning. Some of the differences may be institution dependent. In our NF system, interns focus mainly on patient cross-coverage, while the residents only admit new patients. Interns and residents have different educational goals. Although the current NF system seems to meet those of the interns, the development of additional strategies to enhance the educational component for residents may increase their overall satisfaction.

ASSESSING THE FREQUENCY OF FAILURE TO ADHERE TO BLACK-BOX WARNINGS IN OUTPATIENTS.K.E. Lasser1; D.L. Seger2; T. Yu2; J. Fiskio2; A.C. Seger3; A.S. Karson4; D.W. Bates3. 1Cambridge Health Alliance/Harvard Medical School, Cambridge, MA; 2Partners HealthCare System, Wellesley, MA; 3Brigham and Women's Hospital, Boston, MA; 4Massachusetts General Hospital, Boston, MA. (Tracking ID #115334)

BACKGROUND:  Adverse drug events (ADEs) are a leading cause of injury and mortality in the United States, with one estimate being 100,000 deaths per year. However, relatively little is known about the nature and frequency of ADEs or potential ADEs in the ambulatory setting. We sought to determine the frequency with which outpatient drugs are prescribed in violation of black box warnings, which are released by the Food and Drug Administration, and represent one measure of potential ADEs.

METHODS:  We identified all drugs that had a Physicians’ Desk Reference black box warning in 2002 about drug-drug interactions (DDIs), drug-disease interactions, and drug-lab interactions. Using the electronic longitudinal medical record (LMR), we then identified all ambulatory patients seen from January-December 2002 at all Partners HealthCare-affiliated outpatient practices who were prescribed a drug with a black box warning. We used these data to calculate the frequency with which drugs were prescribed in violation of warnings for DDIs.

RESULTS:  We identified 95 drugs with a black box warning about DDIs, drug-disease interactions, and drug-lab interactions. Of these, 55 (57.9%) had a warning that was so vague that it required clarification from medical specialists at 3 Partners HealthCare-affiliated hospitals. For example, the drug valproate contains a black box warning to check liver function tests at “frequent intervals,” but does not specify how often to monitor liver function tests. Sixty-nine of the 95 drugs (72.6%) were prescribed in the 2002 LMR. Of a total of 324,578 patients prescribed a medication in the 2002 LMR, 33,779 (10.4 %) were prescribed a drug with a black box warning about DDIs, drug-disease interactions, and drug-lab interactions. Of the 1,107 patients who were prescribed a drug with a DDI warning, 401 (36.2%) were also prescribed a contraindicated drug. Of 1,139 orders for drugs with a DDI warning, 419 orders (36.8%) were prescribed with a contraindicated drug. Implicated drugs included: Dihydroergotamine (1 DDI in 1 patient), Thioridazine (3 DDIs in 3 patients), Ergotamine (9 DDIs in 9 patients), Propoxyphene (641 DDIs in 377 patients) and Ketorolac (14 DDIs in 13 patients).

CONCLUSION:  Black box warnings about drug-disease and drug-lab interactions are often imprecise and could be improved. Physicians frequently prescribe drugs despite black-box warnings, which may place patients at important risk of harm. Future analyses should assess how frequently such prescribing results in patient harm.

ASSESSMENT OF A RESEARCH-BASED HEALTH ACTIVISM CURRICULUM FOR MEDICAL STUDENTS.S. Cha1; J.S. Ross1; G. Sacajiu1; P. Lurie2. 1Montefiore Medical Center, Bronx, NY; 2Public Citizen, Washington, DC. (Tracking ID #116904)

BACKGROUND:  Few opportunities exist in medical education to teach physicians the research and advocacy skills necessary to inform and advocate for socially equitable health policy. At Montefiore Medical Center, Bronx, NY, we designed a curriculum for fourth-year medical students in Research-Based Health Activism (RBHA). Our objective was to assess the course and its impact on students’ career goals.

METHODS:  The RBHA curriculum is a one month course, teaching the major skills of health activism to enable students to design community and advocacy-based research projects. The course includes sessions on (1) health policy, (2) physician activists as role models, (3) advocacy strategies, and (4) research methods. At the conclusion of the course, students present a research proposal and advocacy plan, and are encouraged to finish these projects at their home institutions. Six students completed the course in 2002 and 15 in 2003. Students from 2003 completed pre- and post-course surveys using Likert scales assessing course expectations, success at meeting educational objectives, and career goals. Main educational objectives included self-assessed ability to generate a research question, design an advocacy plan, and design and present a research proposal. Data was analyzed using descriptive statistics and McNemar's test.

RESULTS:  Students represented all regions of the country, 12 (80%) from outside of the New York City region. Ten (67%) were female, 7 (47%) identified as non-white, and 13 (87%) intended to begin training in a primary care specialty. All students “agreed” or “strongly agreed” that the course met all educational objectives. The number of students intending to pursue careers “significantly” or “exclusively” involving advocacy was 6 (40%) prior to the course and 9 (60%) after the course (P = .51). The number of students intending to pursue careers “significantly” or “exclusively” involving research was 2 (13%) prior to the course and 10 (67%) after the course (P < .01). The sessions identified as most valuable were within the physician activists as role models and advocacy components of the course. All students completed a research project addressing social or economic disparities in health care. Three of the six 2002 students finished projects at their medical schools.

CONCLUSION:  Our curriculum in RBHA met educational objectives for a diverse group of students. The course significantly increased interest in careers involving research. While there was a nonsignificant increase in interest in careers involving advocacy, many students entered the course with a strong interest in advocacy.

ASSESSMENT OF INDICATIONS CITED FOR THROMBOPHILIA SCREENING IN AN URBAN HEALTH REGION.H.L. McArthur1; S.H. Chou1; M. Poon1; D. Southern1; E.A. MacKay1. 1University of Calgary, Calgary, Alberta. (Tracking ID #116031)

BACKGROUND:  Thrombophilia testing is an expensive diagnostic process for which there is considerable uncertainty regarding optimal patient selection for screening. Our specific objectives were to describe the indications cited for testing in an urban health region as well as to review the yield of these studies by indication.

METHODS:  An administrative database of the thrombophilia tests performed in Southern Alberta, Canada, in 2002 was examined. Testing of one or more of: Factor V Leiden, APC Resistance, Factor II Mutation, Antithrombin III, Protein C, Protein S and Lupus anticoagulant was performed. Indications for testing were categorized as first pulmonary embolism (PE); first deep vein thrombosis (DVT); recurrent PE or DVT; unusual site of venous thromboembolism (VTE); arterial thrombosis; family history of thrombosis or thrombophilia in a first degree relative; family history of thrombosis or thrombophilia in a non-first degree relative; and other.

RESULTS:  Of the 906 thrombophilia studies assessed, the documented indications for testing were: a first PE (11.4%); a first DVT (15.8%); recurrent PE or DVT (12.8%); an unusual site of VTE (6.5%); arterial thrombosis (13.7%); a family history of thrombosis in a first degree or non-first degree relative (16.4% and 2.2%, respectively); a family history of thrombophilia in a first degree and non-first degree relative (7.7% and 1.3%, respectively); and other (11.8%). There was no clear indication for testing in four studies. In analysis of the test results by indication, test results were normal for 77.7% of the patients with a first PE; 71.3% of the patients with a first DVT; 70.7% of the patients with recurrent PE or DVT; 71.2% of the patients with an unusual site of VTE; 79.0% of the patients with arterial thrombosis; 68.9% of the patients with a family history of thrombosis in a first degree relative; 75.0% of the patients with a family history of thrombosis in a non-first degree relative; 44.3% of the patients with a family history of thrombophilia in a first degree relative; 75.0% of the patients with a family history of thrombophilia in a non-first degree relative; and 72.9% of patients with other reasons cited for testing.

CONCLUSION:  The most commonly cited indications for thrombophilia testing were a family history of thrombosis in a first degree relative, a first DVT and arterial thrombosis. In contrast, the yield of positive test results was highest for a less common indication—a family history of thrombophilia in a first degree relative. This combination of findings suggests that the current pattern of testing does not provide an optimal balance of effectiveness and efficiency of screening for thrombophilias, and that real-time filtering of referrals may be needed to limit testing for less appropriate indications.

ASSOCIATION OF ANEMIA WITH DIASTOLIC DYSFUNCTION IN PATIENTS WITH CORONARY ARTERY DISEASE: FINDINGS FROM THE HEART AND SOUL STUDY.D. Nair1; M. Shlipak1; B. Angeja1; H.H. Liu1; N.B. Schiller1; M.A. Whooley1. 1University of California, San Francisco, San Francisco, CA. (Tracking ID #116303)

BACKGROUND:  Anemia is associated with left ventricular hypertrophy and with adverse outcomes in patients with systolic heart failure, but the relationship of anemia with diastolic dysfunction has not been studied. We performed a cross-sectional study to evaluate the association of anemia with diastolic dysfunction and left ventricular hypertrophy in outpatients with coronary artery disease.

METHODS:  Patients with known coronary disease from two VA Medical Centers, one University-based medical center, and nine local health clinics were recruited into the Heart and Soul Study. We measured serum hemoglobin (Hgb) and performed transthoracic echocardiography in 822 participants. We defined diastolic dysfunction as diastolic-dominant pulmonary vein flow, and left ventricular hypertrophy as left ventricular mass index >90 g/m2. Using multivariate logistic regression models, we determined the independent association of moderate (11 < hgb < 13) and severe (hgb < 11) anemia with left ventricular hypertrophy and diastolic dysfunction after adjustment for other baseline characteristics.

RESULTS:  Participants with moderate or severe anemia (n = 197; 24%) were older and less likely to be male or white, compared with participants without anemia. The prevalence of DD was 8% in participants without anemia, and 13% and 24% in participants with moderate and severe anemia. After multivariate adjustment, moderate anemia (adjusted OR 2.0, 95% CI 1.1–3.5; P = .02) and severe anemia (adjusted OR 6.6, 95% CI 1.9–23.2; P = 0.003) remained strongly associated with diastolic dysfunction. When entered as a continuous variable, each 1 mg/dl decrement in Hgb was associated with a 40% (95% CI: 10% to 60%) increased odds of diastolic dysfunction. In contrast, anemia had a minimal association with left ventricular hypertrophy. Compared with participants with no anemia, the OR was 1.4 (95% CI 1.0–2.1; P = .08) in participants with moderate anemia and 1.6 (95% CI 0.6–4.6; P = .38) in those with severe anemia.

CONCLUSION:  We found anemia to be strongly associated with diastolic dysfunction, but not with left ventricular hypertrophy, in this community-based study of outpatients with established coronary artery disease. Diastolic dysfunction may be a critical mediator via which anemia promotes the development and progression of heart failure.

inline image

ASSOCIATION OF BODY WEIGHT WITH PERCEIVED HEALTH AND SEVERITY OF SYMPTOMS FROM CHRONIC CONDITIONS.D.E. Arterburn1; M. McDonell2; S. Hedrick2; P.H. Diehr3; S.D. Fihn2. 1University of Cincinnati, Cincinnati, OH; 2University of Washington, Seattle, WA; 3Fred Hutchinson Cancer Research Center, Seattle, WA. (Tracking ID #116058)

BACKGROUND:  Obesity has important effects on health and health-care costs. There is also evidence to suggest that obesity may increase the severity of symptoms among patients with coronary heart disease (CHD), obstructive lung disease (OLD), and depression. We hypothesized that obese adults with CHD, OLD, or depression would report greater impairments in health-related quality of life (HRQOL) due to their angina, dyspnea, or depression symptoms than those with normal body mass index (BMI).

METHODS:  We performed a cross-sectional analysis of data from 16,393 participants in the Ambulatory Care Quality Improvement Project (ACQUIP), a multicenter study of veterans enrolled in General Internal Medicine clinics. BMI was self-reported, and HRQOL was assessed using the SF-36 and condition-specific measures for symptomatic CHD, OLD, and depression. We used multiple linear and logistic regression to assess the associations between BMI and individual HRQOL domains. All analyses were statistically controlled for study site, sociodemographic characteristics (age, gender, race, income, education, marital status), health habits, obesity-associated chronic conditions (hypertension, diabetes mellitus, stroke, arthritis, and cancer), and 13 other chronic conditions.

RESULTS:  Twenty-seven percent of patients had normal body weight (BMI 18.5 to 24.9); 42.7% were overweight (BMI 25 to 29.9); 28.7 % were obese (BMI 30 and over); and 1.6 percent of patients were underweight (BMI < 18.5). We found statistically significant inverse-U-shaped relationships between BMI and HRQOL scores in all 15 domains we examined. Underweight patients and those with class III obesity (BMI over 40) reported the worst HRQOL scores. Above a BMI of 25, we observed linear, inverse relationships between BMI and HRQOL scores; however, compared with normal weight patients, those who were overweight reported significantly better HRQOL scores in 10 out of 15 domains. Among patients with CHD and OLD, underweight patients and those with class III obesity reported the greatest limitations in physical function due to their angina and dyspnea.

CONCLUSION:  Body mass index was strongly associated with perceived health and severity of symptoms from CHD, OLD, and depression. Our results suggest that, when considering HRQOL outcomes among veterans, the optimal BMI may be above the “normal” range. Further research is needed to test the validity of the 1998 NIH BMI categories as predictors of health outcomes among veterans.

ASSOCIATION OF INTENSITY OF HYPERTENSION THERAPY WITH PATIENT RACE AND ETHNICITY AND BLOOD PRESSURE CONTROL.L.S. Hicks1; M.S. Horng2; D. Fairchild3; E. Orav2; D.W. Bates2; J.Z. Ayanian4. 1Brigham and Women's Hospital and Harvard Medical School, Boston, MA; 2Brigham and Women's Hospital, Boston, MA; 3New England Medical Center, Boston, MA; 4Harvard University, Boston, MA. (Tracking ID #116785)

BACKGROUND:  Numerous studies have examined potential biological mechanisms for racial and ethnic differences in blood pressure (BP) control. However, there has been little examination of racial and ethnic differences in intensity of hypertension (HTN) care and whether these differences are associated with differences in BP control.

METHODS:  We examined the electronic medical records of hypertensive patients with more than one visit to 14 general internal medicine clinics during 7/1/01–6/30/02. For each visit, we collected the mean BP and determined whether HTN drug therapy was intensified (an increase in dose of medication or addition of new antihypertensive medication). We developed an algorithm to determine if each patient had a net increase in intensity of drug therapy (compared to no net increase) in response to repeatedly elevated BPs over the one-year study period. We compared the association of receiving a net increase in therapy with patient race/ethnicity by chi square tests. Using repeated measures logistic regression, we determined adjusted odds of obtaining BP control (mean BP <140/90) at a subsequent visit when therapy was intensified the visit before, and tested the interaction of intensification of therapy and patient race/ethnicity in predicting subsequent BP control.

RESULTS:  Of the 850 patients examined, 671 (78.9%) had a net increase in their therapy during the study period. Net increases occurred more often for Black (81.5%) and White (80.9%) patients compared to Latino patients (70.8%) (P = .02). After adjustment for baseline systolic and diastolic BP, intensifying therapy was associated with a higher odds of obtaining subsequent BP control (OR 1.55, P < .001) compared to not intensifying therapy. Black (OR 0.78, P = .03) and Latino (OR 0.55, P < .001) patients were less likely to obtain subsequent BP control compared to White patients. There were no significant interactions between race/ethnicity and intensification in the multivariate model, suggesting that the association between intensification of therapy and improved blood pressure control did not differ by race/ethnicity.

CONCLUSION:  We found that Latino patients with HTN were significantly less likely to have their antihypertensive drug therapy increased compared to other racial and ethnic groups and that intensification of therapy is associated with subsequent BP control similarly for all racial/ethnic groups. Equal treatment in terms of aggressiveness of drug therapy may reduce disparities in HTN and its related outcomes. Potential targets for intervention to reduce disparities in cardiovascular outcomes should focus on the need to intensify drug therapy more aggressively among high-risk populations.

ASSOCIATION OF NUTRIENT INTAKE AND ELEVATED C-REACTIVE PROTEIN IN U.S. ADULTS.J. Lee1; D. Mann2; S. Natarajan2. 1Cornell University, New York, NY; 2New York University, New York, NY. (Tracking ID #116941)

BACKGROUND:  Elevated C-reactive protein (CRP) is considered an important marker for cardiovascular disease (CVD) risk. High body mass index (BMI), sedentary lifestyle, smoking, diabetes, hypertension and lipid abnormalities are related to elevated CRP. No clear associations between nutrient constituents and CRP levels have been established among the general population. We hypothesized that diets high in fruits, vegetables, fish, fiber, antioxidant vitamins, potassium, magnesium and calcium are inversely associated with elevated CRP. Conversely, we hypothesized that diets low in these nutrients and high in saturated fat would be directly associated with elevated CRP.

METHODS:  The Third National Health and Nutrition Examination Survey (1988-1994), a complex, stratified sample of the US population was analyzed. Participants were interviewed, provided blood for analysis and completed a detailed food frequency questionnaire and 24-hour dietary recall. The association of the primary independent variables, macro- and micronutrient intake, with serum CRP levels was evaluated using SUDAAN to incorporate the complex sampling frame and control for known and potential confounders (age, race, education, employment, BMI, exercise, smoking, alcohol use, diabetes, hypertension, CVD, and supplement and vitamin use).

RESULTS:  The analytic sample was 15,393 adults. The multivariate odds ratio (OR) [with 95% confidence intervals] for elevated CRP (dichotomized at <85% of sex specific distribution) was 0.97 [.94–1.00] for fish ingestion (P < .05) as a continuous variable (times/month). Adults ingesting the top quartile of fish (OR 0.71 [0.59–0.86]), fruits and vegetables (OR 0.79 [0.63–0.99]), and carotenoids (OR 0.76 [0.61–0.95]) had significant inverse associations (protective) with elevated CRP when compared to the remaining quartiles. Individuals reporting a diet profile of top quartile fish, fruits and vegetables, and carotenoids intake (5.1% of US adults) had an OR for elevated CRP of 0.54 [0.34–0.87]. We found no associations between elevated CRP and other nutrients of interest, including magnesium, sodium, potassium, folate, B vitamins, vitamin E, vitamin C, protein, carbohydrates, and saturated fat.

CONCLUSION:  Individuals consuming diets high in fish, fruits and vegetables, and carotenoids had lower likelihood of elevated CRP. This supports current recommendations of `heart healthy’ diets rich in fruits, vegetables, legumes and fish to reduce risk of cardiovascular disease and events. Whether or not dietary supplementation of fish, fruits and vegetables, and carotenoids lowers CRP and corresponding cardiovascular risk is a question proposed for further research.

ASSOCIATION OF THE `5TH VITAL SIGN’ (PAIN) AND DEPRESSION IN PRIMARY CARE.M.J. Bair1; L.S. Williams2; K. Kroenke3. 1Roudebush VAMC & Regenstrief Institute, Indianapolis, IN; 2Roudebush VAMC/Indiana University School of Medicine, Indianapolis, IN; 3Regenstrief Institute/Indiana University School of Medicine, Indianapolis, IN. (Tracking ID #116406)

BACKGROUND:  The Veterans Health Affair's emphasis on pain assessment, operationalized as the “5th vital sign”, and clinical reminders to screen for depression highlight the potential impact of these conditions on veterans’ health. Since pain and depressive symptoms often coexist we sought to better understand their relationship in the primary care setting.

METHODS:  We conducted a cross-sectional survey of VA general medicine patients presenting for routine visits. Acute care visits were excluded. Depressive symptoms (primary outcome) were assessed by the Patient Health Questionnaire (PHQ-9). Patients with PHQ-9 scores >10 were identified as having significant depressive symptoms. Patients were routinely asked by clinic nurses, “How much pain do you have today?” on a 0 to 10 pain scale (PS) and then completed a self-administered questionnaire. We categorized pain (primary predictor variable) as none (PS 0), mild (PS 1–3), moderate (PS 4–6), or severe (PS 7–10). We used logistic regression to examine the association of pain with depressive symptoms, controlling for patient demographics.

RESULTS:  Study participants (N = 301) had a mean age of 60 (range 26–86), 91% were male, and 85% were white. Mean PHQ-9 (depression) score was 6.6 (±6.3) and 28% (n = 83) had significant depressive symptoms (PHQ > 10). Pain was present in 74% (n = 221) of the sample: mild 21%, moderate 31% and severe 22%. All levels of pain severity were independently associated with significant depressive symptoms compared to those without pain: odds ratios of 2.2 (95% CI, 1.1–4.4) for mild pain, 5.2 (2.2–12.5) for moderate pain, and 12.0 (4.1–34.4) for severe pain.

CONCLUSION:  Pain severity measured during routine vital sign assessment is strongly associated with significant depressive symptoms in VA general medicine patients. Routine pain assessment may be a useful indicator (“red flag”) and simple tool to risk statify patients who should be assessed for depression.

ATRIAL FLUTTER AND THE RISK OF THROMBOEMBOLISM: A SYSTEMATIC REVIEW AND META-ANALYSIS.W.A. Ghali1; B. Wasil1; R.F. Brant1; D.V. Exner1; J. Cornuz2. 1University of Calgary, Calgary, Alberta; 2University Hospital of Lausanne, Lausanne, . (Tracking ID #116179)

BACKGROUND:  There is considerable uncertainty and controversy surrounding the risk of thromboembolism associated with atrial flutter, with some suggesting that risk is not elevated while others consider risk similar to that for atrial fibrillation. We conducted a systematic review and meta-analysis of observational studies on this issue.

METHODS:  A literature review was conducted using Medline, EMBASE, bibliographies, and clinical experts. Identified studies were classified into those that report on thromboembolism risk associated with attempted cardioversion vs. those that report on longer-term risk. The review process and data extraction were performed by two reviewers. Study event rates were assessed graphically, and a Chi-square test was used to assess heterogeneity across studies. Meta-regression with weighted logistic regression was used to assess the association between study-level clinical factors and reported thromboembolic event rates.

RESULTS:  We found 13 studies reporting on the thromboembolism risk associated with attempted cardioversion of atrial flutter, and 3 studies reporting on longer-term risk from sustained atrial flutter. Reported event rates in the 13 studies on cardioversion risk ranged from a low of 0% (in 7 separate studies) to a high of 7.3%. However, 4 studies reported a thromboembolism risk above 2% around the time of cardioversion, and five studies notably yielded upper 95% confidence intervals exceeding 10%. The test for heterogeneity was highly significant (P < .001), so results were not pooled. Instead, a meta-regression analysis was performed, and this provided some explanation for the heterogeneity in results across studies. Studies were more likely to report high event rates after cardioversion when they included patients with a prior history of thromboembolism, and reported lower event rates when at least some patients were anticoagulated, or if patients underwent echocardiography before cardioversion. The 3 studies reporting on longer-term thromboembolism risk suggest a yearly event rate of approximately 3% with sustained atrial flutter—somewhat lower than for sustained atrial fibrillation, but higher than for sinus rhythm.

CONCLUSION:  These findings suggest that atrial flutter is indeed associated with an increased risk of thromboembolism after cardioversion as well as in the longer-term. Anticoagulation is therefore warranted in patients undergoing cardioversion for atrial flutter. Meanwhile, patients with sustained atrial flutter should undergo a similar risk assessment to that recommended in current practice guidelines for atrial fibrillation, and higher-risk individuals should be considered for long-term anticoagulation.

ATTITUDES AND EXPERIENCE WITH HELP SEEKING AMONG WOMEN EXPERIENCING INTIMATE PARTNER VIOLENCE.J.M. McCauley1; J.C. Campbell2; L. McNutt3. 1Johns Hopkins Community Physicians, Baltimore, MD; 2Johns Hopkins University, Baltimore, MD; 3University at Albany, SUNY, Albany, NY. (Tracking ID #117031)

BACKGROUND:  Background. Intimate Partner Violence (IPV) is associated with many physical symptoms and psychological distress, but only one in three abused women reports discussing IPV with a health professional. Our objectives were to describe: 1) attitudes toward IPV disclosure 2)steps women have taken to get help for IPV.

METHODS:  Methods: We used a cross-sectional design and surveyed women (ages 18–44) in a low-income clinic in Albany NY, using a computerized screening tool. Emotional, physical and sexual abuse was measured with a modified Composite Abuse Scale (CAS), with 0-3 response = no abuse, 4–10 score = moderate abuse and >11 = severe abuse. We developed an attitude scale from the literature and performed factor analysis. Frequencies and chi square analysis are reported.

RESULTS:  Results: Of the 293 respondents, 53% were 18–29 years old, 65% had <12 years education; 63% were African Americans; 18% had experience moderate and 17%, severe abuse. Factor analysis of the questions identified 4 potential constructs related to disclosure: A) Relationship perceptions (Cronbach's alpha (CA) = .83) B) Fear/Shame (CA = .87) C) Staff Trust (CA = .72) D) Reliance (CA = .72). Emotional /Physical/Sexual Abuse Selected Characteristics (% who agree) None Mod High p A) Problems not that serious 72.1 42.5 18.0 0.000 A) Good outweighs the bad 74.3 60.5 47.4 0.004 B) I will get my partner in trouble 11.3 12.5 25.0 0.082 B) Ashamed to talk 10.7 17.1 25.0 0.060 B) I worried talking may affect my children 12.0 29.3 29.4 0.008 C) Don’t think the staff will help 15.0 22.5 31.6 0.056 D) I can take care of myself 71.9 62.5 63.4 0.377 D) I would rather rely on my friends 48.6 36.6 34.2 0.162 D) I would rather rely on God 57.6 68.3 64.1 0.418 Of women with severe IPV, 45% had discussed it currently or in the past with PCP and/or a social worker; 61% had made safety plans; 52% had sought legal help; 45% had taken “a pill for stress.”

CONCLUSION:  Conclusions: Most women with severe abuse recognized that they had a serious problem. However almost half felt that the good still outweighed the bad in their relationship, even when the abuse was severe. Results demonstrated the importance of self reliance and reliance on God more than friends or clinic staff. These findings suggest that clinicians should explore not only the “good” aspects of a relationship that may present a barrier to asking for help for IPV, but also incorporate a women's spiritual and self-reliant feelings in offering help for IPV.

AVOIDING FREE CARE AT ALL COSTS: A SURVEY OF UNINSURED PATIENTS OPTING NOT TO SEEK CARE AT A COUNTY HOSPITAL.S.J. Weiner1; J. VanGeest2; R. Abrams3; A.H. Moswin4; R.B. Warnecke1. 1University of Illinois at Chicago, Chicago, IL; 2American Medical Association, Chicago, IL; 3Rush-Presbyterian-St. Luke's Medical Center, Chicago, IL; 4Michael Reese Hospital, Chicago, IL. (Tracking ID #103851)

BACKGROUND:  Core safety net providers, as defined by the Institute of Medicine (IOM), offer access to services “regardless of ability to pay.” In assessing the ability of these institutions to serve medically indigent patients, the IOM focused on their financial viability and geographic distribution. Another critical dimension, however, is their capacity for serving the growing numbers of uninsured. Last year, uninsured patients made approximately 20,000 visits to three hospital ERs at non-core safety net providers that were sites for this study, all within four miles of a large county hospital. This study sought to determine whether these uninsured patients are opting out of a free service for one that is costly because they are in a relatively stable financial situation and can afford to self-pay, or because overcrowding is making access overly burdensome.

METHODS:  A face-to-face survey was conducted in Spanish or English with 157 uninsured patients who presented in the emergency rooms at three medical centers, all within four miles of a county hospital. Demographic data was gathered along with information on health status, insurance, income, credit history, and experiences seeking care.

RESULTS:  65% of subjects were unemployed, 72% reported a household income of less than $20,000, 48% reported they were in “fair or poor health,” 56% reported that they worry about their credit “alot,” and 33% were were unable to pay a previous bill at the current site. 65% had prior experience seeking care at the county hospital. 64% of these individuals said they were “very satisfied” or “somewhat satisfied” with the quality of the care they had received, but 75% said they were “not at all” or “not too” satisfied with the wait time. Overall, 41% said they were “not at all” or “not too” likely to go back again. There was no relationship between willingness to return and any of the measures of income, health status and credit problems obtained. There was, however, strong correlation with wait time (P = .007, LR).

CONCLUSION:  The high number of indigent patients in this study who have opted not to use the safety net even when its services are nearby and free, represent overflow from an overwhelmed provider not segmentation of the uninsured by those who are relatively well off. Those surveyed were predominantly very low income, in debt, and perceived themselves to be in relatively poor health—all characteristics of individuals who depend on the safety net. Access clearly involves more than geographic proximity and affordability. Excessive wait times can make medical care functionally inaccessible.

BABEL BABBLE: PHYSICIANS’ USE OF JARGON WITH DIABETES PATIENTS.C.M. Castro1; C. Wilson1; D. Schillinger1. 1University of California, San Francisco, San Francisco, CA. (Tracking ID #115435)

BACKGROUND:  Patients with limited health literacy not only experience problems reading medical materials, but are more likely to experience problems with verbal communication, such as understanding medical terminology. Little is known about primary care physicians’ use and patients’ comprehension of jargon, and whether patient comprehension varies with the communication context.

METHODS:  We audiotaped outpatient encounters between 38 primary care physicians and 74 English-speaking patients with diabetes mellitus and limited health literacy and coded for jargon, defined as physicians’ words or phrases with meanings specific to health and/or health care. We characterized each jargon term as technical (“hemoglobin A1c”) vs. lay jargon (“your weight is stable”), and assigned use of each term to 1 of 4 possible functions in the visit (assessing symptoms, delivering test results, making recommendations, providing health education). We selected 19 diabetes-related jargon terms for a phone survey among a subset of 19 patients and measured comprehension of these terms. We first asked patients to self-report their comprehension using a 4-point Likert scale ranging from no understanding to total understanding and to define each term in their own words. We repeated these steps after presenting the sentence context as transcribed from audiotapes. We scored comprehension of open-ended questions through a consensus process. Finally, we examined differences in comprehension in relation to communication variables (e.g. sentence context vs. no context; explanatory vs. non-explanatory context; jargon from patient's own visit vs. other patients’ visit; lay jargon vs. technical jargon).

RESULTS:  60 of 74 encounters (81%) contained at least one jargon term (mean 3; range 0–14). 37% of jargon occurred as part of physicians’ recommendations and 29% as part of health education. Among the subset of 19 patients, only 22% of self-reported and 18% of open-ended responses showed either “some” or “total” understanding. Open-ended comprehension was not affected by (a) providing sentence context (21% vs. 15%), even if explicitly clarifying (27% vs. 17%); (b) whether jargon was from the patients’ own visit (26% vs. 17%); or (c) whether the jargon was lay or technical (16% vs. 18%).

CONCLUSION:  Primary care physicians caring for diabetes patients with limited health literacy commonly employed clinical jargon during routine office visits, particularly when providing health education and recommendations. Patients tended to have poor comprehension of diabetes-related jargon terms, which did not improve as a function of physicians’ clarification or the nature of the jargon. Future research should explore alternative means to convey clinical concepts such that patients can consistently understand.

BABY BEWARE: PATTERNS OF PRESCRIPTION OF TERATOGENIC MEDICATIONS.E.B. Schwarz1; J. Maselli2; R. Gonzales2. 1University of California, San Francisco, San Francisco, CA; 2Division of General Internal Medicine, UCSF, San Francisco, CA. (Tracking ID #115289)

BACKGROUND:  One of every 28 babies born in the United States is born with a birth defect. In the majority of cases, the cause of the defect is unknown. Certain medications are known to increase the risk of birth defects, and are classified by the FDA as class D or X on the basis of the risk the drug poses to a fetus. The goal of this study is to identify the teratogenic medications most frequently prescribed to women of reproductive age and the providers who most frequently prescribe these medications. In addition, we explore awareness of the risks associated with these medications by assessing rates of concurrent contraceptive counseling.

METHODS:  The 1998–2000 National Ambulatory Medical Care Surveys provide information on 12,681 visits made by women of reproductive age to a national sample of non-federally employed office-based physicians who are primarily engaged in direct patient care. By design, the stratified random sampling strategy allows extrapolation of results to the US population of ambulatory physicians and patients. The primary outcome was defined as an office visit associated with a prescription of a potentially teratogenic class D or X medication. A secondary outcome was visits with documentation of contraceptive counseling or prescription of a contraceptive method.

RESULTS:  One of every twenty-five outpatient prescriptions written for women of reproductive age are for potentially teratogenic medications. Those most frequently prescribed to women of reproductive age, in descending order, were anxiolytics, anti-seizure medications, antibiotics, and statins. Internists and family/general practitioners wrote ten times as many prescriptions for class D or X medications to women of reproductive age as gynecologists (4.0 million vs 0.4 million annual prescriptions; or 6.4% of visits to generalists vs 1.6% of visits to gynecologists). Gynecologists were twice as likely as generalists to document contraception at the same visit a potential teratogen was prescribed (13.5% vs. 6.7%, P = .39). Most visits in which a class D or X medication was prescribed had no documentation that contraception was discussed at the same visit. Visits with prescription of a class D or X medication were no more likely to have documentation of concurrent contraceptive counseling than visits with prescription of a class A or B medication (4.1% vs. 5.4%, P = .24).

CONCLUSION:  Increased awareness of the teratogenic risk associated with certain medications and more frequent provision of contraceptive counseling, may decrease the number of children and families who suffer with birth defects. Ongoing collection of information about the provision of contraceptive and family planning services is essential to an accurate assessment of the risks associated with use of class D and X medications.

BARRIERS TO ACE INHIBITOR USE IN PATIENTS WITH TYPE 2 DIABETES.R.N. Sinha1; A. Ettipio2; C.O. Hershey1; N.M. Khan1. 1SUNY at Buffalo, ECMC, Buffalo, NY; 2SUNY at Buffalo, Buffalo, NY. (Tracking ID #115349)

BACKGROUND:  Studies have shown ACE inhibitors (ACEI) provide cardioprotective benefits in patients over 55 years of age with Type 2 diabetes mellitus(DM). Renoprotective benefits have also been demonstrated and ACEI are recommended for treating hypertension in this population. We reviewed our patients with Type 2 DM to find out the barriers to ACEI use.

METHODS:  We retrospectively reviewed clinic charts in our county teaching hospital clinic from Jun ’02–Sept ’03. The findings were discussed with the Attending Physicians(PMD) involved and their responses further clarified the reasons ACEI were not prescribed for their patients.

RESULTS:  275 Type 2 DM patients Male: Female ratio of 123:152;45%:55% were found. 199 patients were on an ACEI, 76(28%) were not. The 76 patients not on an ACEI were analyzed in detail. 55 patients (72%) were African-American & 17 (22%) were Caucasian. Fifty-five (68%) had prescription benefits through Medicaid, 20(26%) were self-paying & 4(5%) had commercial insurance. 63/76 (83%) had hypertension and 13/76 (17%) were normotensive. The median(range) age was 58.5(37–92) years. Mean(SD) duration of diabetes for the 76 patients was 8.39 (6.6)years, HbA1C 8.2% (1.98); Body Mass Index(BMI) 31.4 kg/m2 ( 8.1); Serum Creatinine(SCr) 1.46 mg/dL (1.35); Mean Arterial Pressure(MAP) was 95.2 mmHg (9.85). 39 of the 76 were never on an ACEI while 37 had been prescribed an ACEI at least once. Analysis of the 37 patients previously on an ACEI revealed the following reasons for ACEI omission or discontinuation. 26 patients had an adverse reaction (cough-11; angioedema-7; increase in serum creatinine-3; rash-5). 6 patients were on dialysis, 2 were initially started on an Angiotensin receptor blocker (ARB), and there was unclear documentation for the remaining 3 patients. PMDs gave three main reasons for not using an ACE in the other 39/76 patients: 1) The UKPDS study revealed no difference in nephropathy outcomes if either a beta blocker or ACEI was used to control hypertension. 2) The patient was normotensive or well controlled on other medications. 3) Lack of appreciation of the possible benefit of ACEI usage in patients with diabetes in particular those without microalbuminuria.

CONCLUSION:  The main barriers to ACEI use found in our study were adverse drug reactions, a lack of physician awareness of the data supporting ACEI (& ARB) use, and somewhat conflicting data between clinical trials. Further physician education and clarification with regards to the outcomes, risks and benfits of the ACEI class of drugs would be helpful.

BARRIERS TO THE TREATMENT OF HEPATITIS C INFECTION: PATIENT, PROVIDER, AND SYSTEM FACTORS.J.A. Morrill1; M. Shrestha2; R.W. Grant3. 1Massachusetts General Hospital, Boston, MA; 2M.G.H.-Revere HealthCare Center, Boston, MA; 3Partners Healthcare, Boston, MA. (Tracking ID #116825)

BACKGROUND:  Hepatitis C virus (HCV) infection affects over 4 million Americans (approximately 2% of the U.S. population). Although associated with side effects, antiviral therapy represents the only means of preventing the potentially devastating complications of chronic HCV infection. Despite this benefit of therapy, many (if not most) patients infected with HCV remain untreated.

METHODS:  We identified a cohort of 275 patients with HCV infection followed at an urban community health center. We report here data for the first 155 patients reviewed to date. Demographic, laboratory, and encounter data were obtained from the electronic medical record. We used structured data abstraction forms to determine treatment status (treated or untreated) and to assess medical, psychiatric, and substance abuse history from 12/01 to 12/03. We also conducted structured interviews with each patient's primary care provider (PCP) to ascertain the PCP's rationale for referring (or not referring) the patient for HCV treatment. Chi-square testing (for categorical variables) and Student's t-test (for continuous variables) were used to compare the characteristics of treated vs. untreated patients.

RESULTS:  Patients in our cohort were an average of 48.2 years of age, 49.7% were male, and most were unemployed (60.6%), unmarried (66.4%), and covered by “safety net” health insurance (i.e., Medicaid or Massachusetts Free Care; 62.6%). Most patients (79%) were white, 12% were Asian (most often Cambodian refugees), 3% were Hispanic, and 2% were Black. The mean HCV viral load among these patients was 310,739 IU/ml, and the mean ALT value was 61.3 U/L. Only 24% of the patients had ever received treatment. Compared with untreated patients, treated patients were more often married (53% vs. 27%, P = .004) and had lower rates of ongoing EtOH use (8% vs. 36%, P = .001), past drug use (37% vs. 57%, P = .03), referral to a substance abuse counselor (0% vs. 12%, P = .03), and major psychiatric disorders (38% vs. 61%, P = .02), as well as a lower ratio of missed clinic visits to total clinic visits (0.10 vs. 0.19, P = .01). There were no significant differences in age, gender, employment status, insurance coverage, or ethnicity between treated and untreated patients. According to their PCPs, the major reasons for not offering HCV treatment included: negative viral load (16.2%), psychiatric comorbidity (16.2%), current or past substance abuse (12.8% and 11.1%, respectively), and a history of poor clinic attendance (6%). In 6% of cases, no clear reason for not treating was identified.

CONCLUSION:  We found low treatment rates in our cohort of urban HCV positive patients that were similar to the rates reported in other studies. Treated patients were more likely to be married and less likely to have problems with psychiatric disease, substance abuse, and compliance with clinic attendance. PCP-identified barriers to referral and treatment—particularly psychiatric and substance abuse comorbidity—highlight potential targets for interventions to improve the rate of appropriate treatment and thereby stem the predicted tide of future chronic liver complications in patients currently infected with HCV.

BEDSIDE INTERACTIONS FROM THE OTHER SIDE OF THE BEDRAIL: PATIENTS’ PERSPECTIVES.K.E. Fletcher1; D. Rankey2; D.T. Stern2. 1Medical College of Wisconsin/Milwaukee VAMC, Milwaukee, WI; 2University of Michigan, Ann Arbor, MI. (Tracking ID #115348)

BACKGROUND:  Bedside encounters between patients and physician teams hold potential. Students and residents benefit, but how are patients affected? In a pilot study, we interviewed patients to identify aspects of bedside interactions that were important to them. Based on that study, we designed a patient questionnaire to quantify the importance of these issues.

METHODS:  The subjects were medical inpatients at a VAMC. Exclusion criteria were being non-English speaking, under 18 or unable to give consent. The team was contacted to ensure eligibility. Each day after collecting data, we destroyed the patient lists. Hence, the total number of available patients (461) contains overlap because some patients’ names appeared on more than one day. Twenty-nine patients refused, and others were unavailable during times the survey was administered. We collected anonymous surveys from 97 patients who were eligible, available and willing to complete them. The survey consisted of 44 questions including short answer, multiple choice and Likert-type questions. Data analysis included descriptive statistics and factor analysis.

RESULTS:  The sample was predominantly male and white. Mean age was 62, and 49% reported their health to be fair or worse. Overall satisfaction with the hospital experience and with the team of doctors were both high (95% and 96% were very or mostly satisfied, respectively). Patients reported learning about several issues during interactions with the teams. The 3 most highly rated areas of learning (1–4 scale, 1 = very important) were new problems (1.10), tests to be done (1.14) and treatments to be done (1.16). Most patients (74%) felt that their teams cared about them very much, and 84% said that the team caring about them was very important. Patients were made comfortable when the team showed that they cared, listened and appeared relaxed (reported by 63, 57 and 54%, respectively). Patients were made uncomfortable when the team used language they did not understand (22%) and when several people examined them at once (13%). Many (58%) patients felt personally involved in teaching; only 6% did not like the teaching. The majority of patients liked having medical students and residents involved in their care (69% and 64%, respectively). Factor analysis resulted in 3 factors: an interactiveness factor (alpha = 0.65), a team-oriented factor (alpha = 0.63) and a satisfaction with understanding factor (alpha = 0.77).

CONCLUSION:  Patients have much to teach us about team bedside interactions. Patients’ reactions are generally positive, but patients are different with respect to what makes them comfortable and uncomfortable. Taking their preferences into account could improve the experience of being in a teaching hospital.

BELIEFS ABOUT SMOKING RISKS, INTENTIONS TO QUIT AND ENROLLMENT IN A SMOKING CESSATION PROGRAM.A. Gurmankin1; K. Volpp2. 1Rutgers, The State University of New Jersey, New Brunswick, NJ; 2Philadelphia VA Medical Center, Philadelphia, PA. (Tracking ID #115522)

BACKGROUND:  Despite the well-known risks of smoking and the effectiveness of smoking cessation programs (SCPs), many smokers do not intend to quit and do not enroll in SCPs. According to the health belief model, feeling threatened by a risk strongly influences the adoption of the corresponding protective behavior. Thus, an improved understanding of the role of health concerns of smoking on quitting intentions is critical to developing strategies for clinicians to encourage patients to quit.

METHODS:  431 primary care patients at the Philadelphia Veteran's Affairs medical center (VA) completed a survey assessing their intention to quit smoking, reasons to quit, beliefs and worry about risks, and comparative risk perception. All items used a 1–5 response scale, except worry about risks, which used a 1–7 scale. Eligibility for an SCP was determined by contraindications for the nicotine patch. Eligible subjects (n = 179) were invited to participate in a free 4-session SCP at the VA.

RESULTS:  Compared to those without an intention to quit in the next 6 months (“precontemplators”) (n = 172), those with this intention (“contemplators”) (n = 259) reported greater motivation to quit because of concern about serious illness if don’t quit (4.2 vs. 3.4, P < .0001), knowledge of others who have died from serious illness caused by smoking (3.8 vs. 2.2, P = .0002), concern that smoking will shorten life (4.3 vs. 3.8, P = .0002), noticing physical symptoms that smoking is hurting health (4.0 vs. 3.5, P = .0002), and were more worried about the health risks of smoking (5.3 vs. 4.2, P < .0001). Compared to precontemplators, contemplators more strongly agreed that quitting smoking now can lower the chance of the risks of smoking (4.5 vs. 4.2, P = .0033), more strongly disagreed that doctors exaggerate the badness of the risks of smoking so that smokers will quit (2.3 vs. 2.8, P = .0026), and rated their chance of the risks of smoking compared to a same-sex/age nonsmoker as higher (heart disease: 4.1 vs. 3.7; lung cancer 4.2 vs. 3.8; bladder cancer: 3.8 vs. 3.5; emphysema: 4.3 vs. 4.0; bronchitis: 4.2 vs. 4.0, P < .05 for all). Among eligible subjects, contemplators were nonsignificantly more likely to enroll in the SCP compared to precontemplators (36% vs. 24%, P = .109).

CONCLUSION:  Concerns and beliefs about the risks of smoking are associated with the intention to quit smoking, and those with an intention to quit are more likely to enroll in an SCP. These results suggest that educating smokers about the risks may motivate them to attempt to quit.

BEYOND ELECTROLYTES: EATING DISORDERS AND MEDICAL COMORBIDITY.C. Carney Doebbeling1; L.E. Jones2. 1Indiana University School of Medicine & Regenstrief Institute, Indianapolis, IN; 2University of Iowa College of Public Health, Iowa City, IA. (Tracking ID #117218)

BACKGROUND:  Primary care physicians often provide first-line assessment and ongoing care of women with eating disorders (ED). The objective of this study is to characterize the medical conditions occurring in women with anorexia nervosa (AN), bulimia nervosa (BN), and Eating Disorder Not Otherwise Specified (ED NOS).

METHODS:  The data source is a 100% sample of Wellmark Blue Cross Blue Shield claims data, 1996–2001. A subject was identified as having an ED if she had a hospitalization, psychiatrist visit, or two outpatient visits with a primary/secondary ICD-code for AN, BN, or ED NOS. Controls were randomly selected, age-matched women without claims for ED. A comorbidity was counted if it occurred during any hospitalization, or at least twice in outpatient claims for a period of greater than 30 days. Odds ratios with 95% confidence intervals were calculated.

RESULTS:  604 women (mean age = 23 yrs) met criteria for AN, 552 (mean age = 25 yrs) met criteria for BN, and 231 (mean age = 28 yrs) met criteria for ED NOS. The average number of comorbidities was higher among women with ED compared to controls: AN = 0.46 vs 0.14 (P < .001); BN = 0.42 vs. 0.17 (P < .001); and ED NOS = 0.65 vs. 0.22 (P < .001). Women with AN were significantly more likely to have arrhythmias (OR = 5.6), valvular disease (OR = 3.6), neurological disorders (OR = 3.6), chronic pulmonary disease (OR = 2.2), peptic ulcer disease (OR = 6.0), hypothyroidism (OR = 2.7), fluid/electrolyte disorders (OR = 14.4), blood loss anemia (OR = 15.1), and deficiency anemias (OR = 4.4). Women with BN were more likely to have chronic pulmonary disease (OR = 2.1), diabetes without complications (OR = 3.0), hypothyroidism (OR = 1.9), peptic ulcer disease (OR = 15.1), obesity (OR = 8.7), hypertension (OR = 2.4), fluid/electrolyte disorders (OR = 3.8), and deficiency anemias (OR = 2.2). Women with ED NOS were more likely than controls to have claims for arrhythmias (OR = 3.5), hypertension (OR = 2.9), hypothyroidism (OR = 2.7), rheumatoid arthritis (OR = 4.3), obesity (OR = 53.5), fluid/electrolyte disorders (OR = 3.1), and deficiency anemias (OR = 2.8).

CONCLUSION:  Young women with ED have considerable medical comorbidity, well in excess of controls. Although an elevated risk for electrolyte disorders and arrhythmias has been reported, an increased likelihood for anemias, hypothyroidism, and chronic pulmonary conditions is not well described. These results substantiate the need for thorough medical assessments in ED patients, and support early intervention to prevent chronic medical complications. The data implicate that insurers be aware of the significant risk for medical comorbidity and provide for treatment programs that address the psychological and physical needs of these patients.

BODY-MASS INDEX IN RELATION TO OVARIAN CANCER RISK IN A POPULATION-BASED CASE-CONTROL STUDY.N.B. Peterson1; A. Trentham-Dietz2; P.A. Newcomb3; T. Gabretsadik1; K.M. Egan1. 1Vanderbilt University, Nashville, TN; 2University of Wisconsin-Madison, Madison, WI; 3University of Washington, Seattle, WA. (Tracking ID #115855)

BACKGROUND:  The relation of body weight to ovarian cancer risk is still uncertain. A few previous studies suggest that a heavier body weight is most related to ovarian cancer risk in premenopausal women.

METHODS:  We analyzed data from a population-based study of 757 cases of ovarian cancer and 5943 controls from Massachusetts and Wisconsin conducted from 1992–2001. In a telephone interview, information was ascertained on height, weight, menopausal status, and other ovarian cancer risk factors. We categorized all women into one of 5 body-mass index (BMI) groups. Logistic regression was used to estimate multivariate-adjusted odds ratios and 95% confidence intervals.

RESULTS:  The mean age for cases and controls was 57 and 60, respectively. Over 96% of the participants were white. 68% of the cases 71% of the controls were postmenopausal. The median BMI (kg/m2) was 26 for cases and 25 for controls. There was a trend of increasing ovarian cancer risk with increasing BMI in premenopausal (p trend 0.04 for ordinal BMI groups) but not postmenopausal (p trend 0.97) women after adjusting for age, state, parity, duration of oral contraceptive use, family history of breast or ovarian cancer, history of tubal ligation, and education (see table).

inline image

CONCLUSION:  Overall, body-mass index was unrelated to ovarian cancer risk in these data. However, BMI was positively related to ovarian cancer risk among premenopausal women, a result consistent with several previous studies. Further research is needed to determine the potential hormonal/endocrine mediators of the effect of premenopausal weight on ovarian cancer risk.


BACKGROUND:  Medical schools remained the last major medical institution in this country to racially integrate and thus, delayed the final steps in eliminating overt discrimination against minorities until the late-1960s, early-1970s. This paper explores strategies used by African American physicians to obtain faculty appointments at predominantly white medical schools and admitting privileges at hospitals in one northern and two southern communities.

METHODS:  Primary research sources include manuscript collections of physicians and medical school and university administrators; medical school and university documents; oral history interviews; Federal documents and court cases; and articles from the Journal of the National Medical Association.

RESULTS:  African American physicians and concerned colleagues pursued three strategies in the 1950s and 1960s to secure faculty appointments at predominantly white medical schools. These included: 1) legal action combined with local pressure; 2) legal action combined with Federal policy; and 3) Federal policy combined with quiet negotiation. The first model, legal action combined with local pressure, is exemplified in the racial integration of Chicago medical schools and teaching hospitals. Denied hospital staff privileges, the African American physicians in Chicago brought suit against all the non-profit hospitals for violation of city and state laws and discrimination under the United States Constitution. This action was followed by pressure applied through the Chicago Commission on Human Rights ultimately resulting in the racial integration of both the city's hospitals and medical schools. The second model, legal action combined with Federal policy, is illustrated by the experience of the University of Alabama in Birmingham. While DHEW officials praised UAB as a model of racial integration of clinical services, its black employees filed suit against the administration for violation of Title VI of the 1964 Civil Rights Act. Fearing loss of Federal funds, the administration moved quickly to admit an African American physician onto the faculty and open education programs to minority applicants. The third model, Federal policy combined with negotiation, is described in the case of Duke University School of Medicine when the administration admitted the first group of African American physicians onto the faculty in 1969 primarily in reponse to Federal policy and pressure from the local Black leadership.

CONCLUSION:  Federal policy combined with legal action and negotiation proved successful in the effort to racially integrate medical schools in the United States. Use of the court system by African American physicians and concerned citizens proved essential in the movement to eliminate race discrimination in medical education, and helped shape substantially Federal policy.

BRIEF INTERVENTION FOR THE SPECTRUM OF ALCOHOL PROBLEMS ON A HOSPITAL MEDICAL SERVICE: PRELIMINARY RESULTS FROM A RANDOMIZED TRIAL.R. Saitz1; T. Palfai1; N.J. Horton2; D.M. Cheng1; N. Freedner3; N. Tibbetts4; K.L. Kraemer5; M.S. Roberts5; J.H. Samet1. 1Boston University, Boston, MA; 2Smith College, Northampton, MA; 3Boston Medical Center, Boston, MA; 4DM-Stat, Medford, MA; 5University of Pittsburgh, Pittsburgh, PA. (Tracking ID #115960)

BACKGROUND:  Brief counseling interventions are efficacious in outpatient settings for problem drinkers, but have not been tested in randomized trials for the spectrum of alcohol problems in hospitalized medical patients.

METHODS:  Medical inpatients in an urban general hospital identified by alcohol screening were randomized (stratified by Alcohol Use Disorders Identification Test score ≥12) to a brief motivational intervention or to usual care. The main eligibility criterion was risky drinking amounts defined by age- and gender-specific quantities and frequencies. Mental status impairment and <2 contacts to assist with follow-up were exclusions; alcohol dependence was not. Primary outcomes were: 1) self-reported linkage with alcoholism treatment by 3 months for patients with alcohol dependence, and 2) changes in alcohol use at 12 months. Intention-to-treat analyses were multivariable logistic and linear regression models adjusting for imbalances in randomized groups.

RESULTS:  We approached 10,273 hospital admissions representing 7,824 individuals, of whom 5,813 were screened for risky drinking; 986 (17%) were drinking risky amounts; 524 were eligible and 341 enrolled. For these interim analyses, 265 (78%) completed 3-month, and 202 (59%) 12-month follow-up. Most were men (71%), black (45%), with mean age 44 years. For those with alcohol dependence (77%) the effect of intervention on 3-month linkage differed by gender: Among women, but not men, those in the intervention group were more likely to link with alcoholism treatment (for women, adjusted odds ratio 11, 95% CI 1.6–78; for men AOR 0.6, 95% CI 0.2–1.4). There was no significant effect of the intervention on binges or drinks/day at 12 months for alcohol dependent patients (n = 163). The intervention did significantly influence drinking among non-dependent drinkers (n = 39): Intervention subjects reported a decrease in the number of binges and drinks/day from baseline to 12 months, while controls reported increases (adjusted mean change in binges –3.3 for intervention vs. +2.9 for controls, P = .06; adjusted mean change in drinks/day −1.8 for intervention vs. +1.5 for controls, P = .02).

CONCLUSION:  In hospitalized patients, brief intervention shows some promise for linking alcohol dependent women with alcoholism treatment, and for decreasing drinking in non-dependent risky drinkers. The prevalence of risky drinking is high but unlike outpatient settings, the majority of patients detected in the hospital have alcohol dependence. The fact that the intervention was not equally effective across the spectrum of alcohol-involved patients suggests that hospital interventions may need to be tailored to the level of alcohol problem severity.

BRIEF SCREENS FOR DEPRESSION AND SUICIDALITY IN PRIMARY CARE: TWO ITEMS ARE BETTER THAN ONE.K. Corson1; M. Gerrity1; S. Dobscha1. 1Portland VA Medical Center, Portland, OR. (Tracking ID #115647)

BACKGROUND:  Background: Very brief measures of depression (e.g., 1 item assessing depressed mood over much of the past year, and 2 items assessing depressed mood and anhedonia over the past month) used for screening have been shown to have good sensitivity and fair specificity. In 1999, Portland Veterans Affairs Medical Center (PVAMC) primary care clinics initiated screening for depression using a 1-item measure. The primary objectives of this study were: 1) to compare the test characteristics of a 1-item measure currently used for annual depression screening to validated scoring algorithms of the Patient Health Questionnaire (PHQ) and 2) to estimate the proportion of primary care patients not currently receiving mental health therapy who screen positive for depression.

METHODS:  Patients due to be seen in the PVAMC primary care clinics were identified using appointment recall lists. Of 1,447 patients who met inclusion criteria for a randomized trial, “Improving Outcomes of Depression in Primary Care,” 1,226 completed the telephone screening. We administered the PVAMC 1-item depression screen (“Have you felt depressed or sad most of the time in the past year?”), the PHQ, and two items devised to follow-up a positive response to the PHQ item assessing thoughts of death/suicide. We compared scores based on the PHQ-9, which encompasses the DSM-IV criteria for major depression, the PHQ-8, which excludes the suicide item, and the PHQ-2, which includes only the anhedonia and depressed mood items.

RESULTS:  Using the PHQ-9 cutpoint for moderate depression as the reference standard, the 1-item screen was specific (93%) but less sensitive (68%), whereas the PHQ-2 demonstrates specificity (88%) and sensitivity (95%). For case-finding, PHQ-8 was virtually equivalent to PHQ-9. Across methods, approximately 20% of patients had scores suggesting moderate or greater depression severity. Of 84 patients endorsing the PHQ-9 suicide item, 28 (33%) reported thoughts of harming themselves, and 16 (19%) had a specific plan.

CONCLUSION:  The goal of screening places sensitivity at a premium. Administering two items rather than one improves performance with minimal added time investment. Assessing suicidal ideation as part of screening did not improve case-finding; however, 1/3 patients who endorsed this item reported recent active suicidal ideation. A substantial proportion of primary care patients experiences depressive symptoms that require further clinician assessment.

CAN BREAST AND CERVICAL CANCER SCREENING VISITS BE USED TO IMPROVE COLORECTAL CANCER SCREENING?R.C. Carlos1; A.M. Fendrick1; S.J. Bernstein2. 1University of Michigan, Ann Arbor, MI; 2Ann Arbor VA, Ann Arbor, MI. (Tracking ID #116711)

BACKGROUND:  Despite high acceptance levels of mammography and cervical cancer screening by U.S. women, adherence with colorectal cancer (CRC) screening remains sub-optimal. A better understanding of the relationship among cancer screening behaviors by women may provide insight into interventions to enhance CRC screening.

METHODS:  Women 50 years and older who participated in the 2001 Behavioral Risk Factors Surveillance Survey (BRFSS) were queried regarding cancer screening patterns. Predictors of CRC screening were determined using multivariate analysis from sociodemographic data and non-CRC screening adherence rates based on American Cancer Society guidelines.

RESULTS:  Among the 52,478 CRC module respondents, cancer screening adherence was significantly less for CRC (46.1%) compared to cervical cancer (69.5%) or breast cancer (82.4%). In multivariate analysis, self-reported good health (adjusted OR 0.81, P < .01) and current smoking (adjusted OR 0.78, P < .01) independently predicted decreased CRC screening adherence; while increasing age (adjusted OR 1.03, P < .01), having health insurance (adjusted OR 1.59, P < .01), having a personal physician (adjusted OR 1.61, P < .01), adherence with cervical cancer screening (adjusted OR 1.85, P < .01) and adherence with breast cancer screening (adjusted OR 2.30, P < .01) were independent positive predictors of CRC screening adherence. Participants who adhered to both mammography and Pap smear guidelines were significantly more likely to adhere to CRC screening (51.5% CRC screening adherence) when compared to women who adhered to neither screening test (8.2% CRC screening adherence), with an adjusted OR 5.33 (P < .001). Participants who adhered to both mammography and Pap smear guidelines were significantly more likely to adhere to CRC screening compared to women who adhered to either screening test (38.0% CRC screening adherence) with an adjusted OR 2.00 (P < .001).

CONCLUSION:  Women adherent with mammography and cervical cancer screening guidelines were significantly more likely to undergo CRC screening than those who were not adherent, although CRC acceptance in the adherent group was still sub-optimal. Non-colon cancer screening visits represent a “teachable moment” for education and behavior-related intervention aimed to reduce the burden of CRC.

CAN INTERVENING TO RAISE EXPECTATIONS FOR AGING CHANGE QUALITY OF LIFE AMONG OLDER ADULTS?C.A. Sarkisian1; C.L. Davis2; B. Weiner1. 1University of California, Los Angeles, Los Angeles, CA; 2W.A. MacColl Institute for Healthcare Innovation, Seattle, WA. (Tracking ID #117006)

BACKGROUND:  Among older adults, attributing age-associated problems to “old age” is a common phenomenon. We set out to examine whether an intervention aimed at raising expectations for aging influences quality of life among older adults.

METHODS:  We recruited 51 adults aged ≥65 years to participate in a pre-post trial at 3 senior centers. The intervention consisted of 4 weekly 2-hour group sessions of 8–14 participants. During the first hour of each session a trained facilitator applied “attribution retraining” using a standardized curriculum developed by our multidisciplinary team: facilitators taught that many age-associated health problems should be attributed to mutable causes rather than “old age.” Each discussion session was followed by a 1-hour exercise class modeled after the Lifetime Fitness Program© during which the attribution retraining session principles were verbally reinforced. In-person interviews were conducted at baseline and 7-week follow-up. Quality of life was measured: 1) using the SF-12 to compute Physical (PCS-12) and Mental (MCS-12) Component Summary scores using standardized weights, with higher scores indicating better quality of life; and 2) by counting the #activities of daily living (ADLs, 13 total) participants could not do without difficulty. We measured expectations for aging using the ERA-38, a previously tested instrument in which higher scores indicate expecting higher functioning with aging. Change in scores from baseline to 7-week follow-up were examined using 2-sided t tests.

RESULTS:  The study was completed by 46 seniors (90% of enrollees). Mean participant age was 77 years; 89% were female; 65% were Latino; 30% were non-Latino white. Results are shown in the table below. All completers reported improved mood; most reported decreased pain (61%), improved energy (85%); and improved sleep (54%).

CONCLUSION:  In this small pre-post study, our attribution retraining intervention raised self-reported mental quality of life and decreased ADL impairment among older adults. Efficacy of this intervention should be tested in a randomized trial.

Table 2. Effect of Attribution Retraining Intervention on Quality of Life and Expectations for Aging:
 Mean at BaselineMean at 7-weeksEffect Sizep-value
MCS12 Score52.155.311%0.049
PCS12 Score43.644.2 2%0.671
ADL imprmnts 0.93 0.59n/a0.041
ERA-38 Score30.840.130%0.0003

CANCER RISK PERCEPTIONS AND SCREENING PRACTICES AMONG BLACK AND WHITE WOMEN.T.A. Battaglia1; K.S. Roloff1; J.L. Speckman1; F.A. Farraye1; K.M. Freund1. 1Boston University, Boston, MA. (Tracking ID #117187)

BACKGROUND:  Risk perceptions have been found to influence screening practices, but little is known about racial differences. We compare Black and White women on their perception of developing breast and colorectal cancer (CRC) and examine for an association with screening practices.

METHODS:  We surveyed English speaking women, 40 years and older presenting to an inner-city, hospital-based primary care practice for women. A self-administered survey assessed demographics, screening practices and measures of breast and CRC risk perceptions: 1) Quantitative: “My chances of developing cancer in the next 20 years/lifetime is __in 1000”; 2) Categorical: “Compared to other women my age, my chances of developing cancer are: Above/Average/Below”; and 3) a validated susceptibility scale. Numeracy was assessed based on previous criteria, including a test question of number of heads with 1000 coin tosses. Age and race-specific cancer probabilities were calculated using the Modified Gail Model and Harvard Center for Cancer Prevention Model. Each woman was categorized as an under/correct/over-estimator based on this calculated probability compared with her categorical risk perception.

RESULTS:  At interim 93 women completed the survey (76% response): 38% Black, 52% White; mean age 54; 67% had some college education; 31% received public assistance or insurance. 80% of women were up to date with mammography and 71% of women over age 50 were up-to-date with CRC screening. No racial differences were found in screening practices (breast P = .80, colorectal P = .72). Only 39% of women were considered numerate. Categorical risk perceptions correlated well with susceptibility scores (breast and colon both P = .001). Each woman was categorized on her ability to estimate her calculated cancer risk (see Table 1). There was no significant difference between black and white women in their ability to estimate categorical relative risk (breast P = .11, colon P = .50) (see Table 2). There was no significant difference between women by their screening status (up-to-date or not) (breast P = .81, colon P = .54), or by numeracy (numerate or not) (breast P = .92, colon P = .67) in their ability to estimate risk.

Table 1. Ability to Estimate Calculated Risk
 UnderestimateCorrect EstimateOverestimate
Breast Cancer Risk25 (30%)34 (41%)24 (29%)
Colorectal Cancer Risk 7 (8%35 (41%)44 (51%)
Table 2. Ability to Estimate Calculated Risk by Race
  UnderestimateCorrect EstimateOverestimate
Breast Cancer RiskBlack10 (32%) 8 (26%)13 (42%)
 White13 (30%)21 (49%) 9 (21%)
Colorectal Cancer RiskBlack 4 (13%)14 (44%)14 (44%)
 White 3 (7%)17 (38%)25 (55%)

CONCLUSION:  Even in this educated group of women with high screening rates, there was a low level of numeracy. Racial differences in risk perception and screening behavior were not identified. Greater education on risk assessment should concentrate on non-numeric categories.

CANCER SCREENING IN OLDER WOMEN: WHAT ARE WE DOING?R. Salazar1; G. Gildengorin1; E.J. Perez-Stable1; J.M. Walsh1. 1University of California, San Francisco, San Francisco, CA. (Tracking ID #116154)

BACKGROUND:  The appropriate age at which screening for breast and cervical cancer should be discontinued is unclear although evidence and guidelines question routine screening after age 70.

METHODS:  We performed a retrospective review of computerized medical records from primary care practices in an academic medical center to evaluate rates and predictors of cervical and breast cancer screening in women over age 69. Women with at least one visit between 1/1/01 and 1/1/03 were included. Women with a history of AIDS, dementia, hysterectomy, cervical cancer or breast cancer were excluded.

RESULTS:  Data regarding cervical cancer screening was collected for 12,259 women age 18 including 1,150 (9.4%) women over age 69. 60.9% of women over age 69 received a Papinocolau smear (Pap) in the preceding 3 years compared with a rate of 88.6% (P < .001) in women age 18–69. Although the percentage of those women receiving a Pap decreased with increasing age, many women were still screened (age 70–79: 66.5%, 80–89: 52.3%, 90+: 39.4%; P < .001). More women over the age of 69 receiving their care in OB-Gyn had a Pap compared with women seen in an Internal Medicine (IM) or Family Practice (FP) setting (82.3% vs. 53.7% and 58.5%, respectively; P < .001). Receiving care in an Ob-Gyn clinic was associated with a greater likelihood of having a Pap (O.R. 5.3; 95% CI 3.42–8.26). Information regarding mammography was obtained for 6,000 women age 40 and older, including 1,236 (20.6%) women over age 69. 83.4% of women age 70 and older had a mammography in the preceding 2 years compared with 54.3 % of women age 40 to 49 and 84.0% of women age 50–69 (P < .001). Rates of mammography use decreased with increasing age (age 70–79: 86.2%, age 80–89: 79.0%, age 90+: 71.9%; P < .001). Older white women were screened less frequently than Asian, African American or Latina women (79.8% vs. 84.8%, 87.8% and 87.0%, respectively; P = .01). Older women with more than 5 visits per year to were more likely to have had mammography (O.R. 1.2; 95% CI 1.12–1.29).

CONCLUSION:  A substantial number of older women continue to be screened for cervical cancer, especially in Ob-Gyn practices. Furthermore, a significant number of these women continue to receive mammograms despite unclear efficacy. Continued research is needed to futher understand patient and clinician factors that lead to these screening practices.

CARE OF PATIENTS UNDERGOING ANGIOPLASTY IN SPECIALTY AND NON-SPECIALTY HEART HOSPITALS: CHERRY PICKING AND/OR IMPROVED OUTCOMES?P. Cram1; G.E. Rosenthal2; M.V. Sarrazin3. 1University of Iowa, Iowa City, IA; 2None Given, Iowa City, IA; 3VAMC Iowa City, Iowa City, IA. (Tracking ID #116831)

BACKGROUND:  There has been little rigorous research addressing the issues of “cherry picking” and health outcomes among patients receiving care in specialty and non-specialty hospitals. We conducted a case-control study of patients who underwent percutaneous coronary interventions (PCI) in specialty and non-specialty heart hospitals to determine whether specialty hospitals cared for patients with greater socio-economic resources and/or delivered improved outcomes.

METHODS:  MEDPAR Part A data was used to identify all patients admitted to specialty and non-specialty hospitals within the same hospital referral region (HRR) for PCI (via ICD-9CM code) during 2000-01. Specialty hospitals were defined as hospitals with ratios of Major Diagnostic Category 5 admissions (MDC 5: Diseases of the Circulatory System) to total admission exceeded 50%. Patients were linked to U.S. census data to obtain socio-economic measures. We compared demographics, socio-economic status, and 30-day mortality among patients admitted to specialty and non-specialty hospitals. Logistic regression was used first to compare mortality after adjusting for demographics and comorbidity alone; subsequently, adjustment was made for hospital PCI volume. Sensitivity analysis was performed using alternative thresholds for defining specialty and non-specialty hospitals.

RESULTS:  There were 26,542 admissions to 28 specialty hospitals and 37,950 admissions to 133 non-specialty hospitals. Demographic characteristics were similar for specialty and non-specialty hospital admissions; likewise, mean per-capita income did not differ among patients admitted to specialty and non-specialty hospitals ($23,300 vs. $23,700) nor median home values ($145,000 vs. $149,000). Specialty hospitals had significantly higher PCI volume than their local competitors (477 vs. 162). Unadjusted mortality was lower for specialty than non-specialty hospital admissions (2.9% vs. 2.0%; P < .001). Logistic regression adjusting for demographics and comorbidity revealed lower mortality for patients who received care in specialty hospitals (OR: 0.74: 95% CI: 0.67–0.83); addition of hospital PCI volume to the model eliminated this difference (OR: 1.02: 95% CI: 0.90–1.16). Using alternative methods for defining specialty hospitals did not alter our findings.

CONCLUSION:  Specialty hospitals do not appear to preferentially admit patients with greater economic resources. Specialty hospitals do have lower mortality rates, and this is largely a function of their higher PCI volume.

CAREGIVER CHARACTERISTICS ASSOCIATED WITH DEMENTIA-RELATED BEHAVIORS.K.M. Sink1; K.E. Covinsky1; R.J. Newcomer1; K. Yaffe1. 1University of California, San Francisco, San Francisco, CA. (Tracking ID #116745)

BACKGROUND:  Among patients with dementia, dementia-related behaviors (DRB) are common and associated with negative outcomes for both patients and caregivers. Though it has been observed that some patient-caregiver interactions may increase DRB, few studies have examined the extent to which caregiver characteristics influence the report of DRB. Our objective was to determine if caregiver characteristics are independently associated with DRB after accounting for patient characteristics.

METHODS:  We conducted a cross-sectional study of 5788 community dwelling patients with dementia and their informal caregivers enrolled in a case-management intervention at 8 sites across the U.S. Caregivers were asked about the presence of 12 DRB (for example: hallucinations, constantly restless, combative, wanders, and wakes caregiver). Predictors included caregiver characteristics such as demographics, relationship to the patient, hours/week spent caregiving, self-rated health, functional status, depression, and burden. We used multivariate linear regression to determine which caregiver characteristics were independently associated with higher numbers of DRB, controlling for patient age, gender, dementia severity, functional status, and dementia type.

RESULTS:  The median age of patients was 79, 60% were female, and their mean MMSE = 14. Median caregiver age was 65, 72% were female, 49% were spouses, and 31% daughters. The median hrs/wk spent caregiving was 89.5 and 32% of caregivers were depressed. The mean number of DRB was 4.8. Bivariate results are presented in the table. After adjustment for patient characteristics, caregiver characteristics that remained independently associated with higher numbers of DRB were younger age, less education, being the daughter, more weekly hours caregiving, greater depression, and greater burden (all P < .005).

CONCLUSION:  Caregiver characteristics are associated with increased reports of DRB in patients with dementia. Clinicians should consider the dynamic between patient and caregiver characteristics as a factor in the presence of DRB. Whether higher numbers of reported DRB represent actual greater DRB vs. differential reporting, needs to be examined.

Table 5. Mean Number of DRB by Caregiver Characteristics (all P < .001 except education)
AgeEducationRelation to PatientHours/weekDepressedBurden
<65: 5.1<9yrs: 4.8Dtr: 5.2<89.5: 4.5No: 4.5<median: 4.1
65: 4.59yrs: 4.8Other: 4.6 89.5: 5.2Yes: 5.5median: 5.6

CAREGIVING IN THE FINAL MONTH OF LIFE: ARE MEN LESS DISTRESSED OR DOING LESS?E.K. Fromme1; P. Ebert1; L.L. Drach1; V.P. Tilden2; S.W. Tolle1. 1Oregon Health & Science University, Portland, OR; 2University of Nebraska Medical Center, Omaha, NE. (Tracking ID #117545)

BACKGROUND:  As more patients die in non-hospital settings, families bear greater responsibility for end-of-life caregiving. Gender and generation (i.e. spouse vs. adult child) issues are important in understanding the different challenges (e.g. employment) that caregivers must face. We investigated how gender and generation issues related to how caregivers perceived their own distress and that of their dying family members.

METHODS:  Using a random sample of Oregon death certificates, we identified decedents who died between June 2000 to March 2002. We conducted a structured telephone interview with the decedent's main family caregiver. The caregivers reported their own distress (physical, emotional, sleep disruption, confinement) using 8 items from Robinson's Caregiver Strain Index. They also reported their perception of decedents’ distress using the Family Memorial Global Distress Index (GDI), a 26-item validated inventory of 10 common end-of-life symptoms and symptom-related distress. We assessed univariate relationships using independent t tests, and constructed a stepwise linear regression with caregiver distress as the dependent variable.

RESULTS:  1,384 of 1,825 (76%) eligible caregivers who were located completed interviews. We excluded 298 caregivers who were not spouses or adult children, leaving n = 1086 (122 husbands, 274 wives, 203 sons, and 489 daughters). Decedent's mean age was 79 years, 83% were white, 67% were enrolled in hospice, 99% had health coverage, and 52% died at home. For caregivers’ personal distress, husbands (µ = 5.31,SD = 4.2) and sons (µ = 5.29,SD = 3.9) were equivalent (P = .98). They reported less personal distress (P = .01) than wives (µ = 6.48, SD = 4.5) and daughters (µ = 7.56,SD = 4.3). GDI scores showed the same pattern: husbands (µ = .77, SD = .63) and sons (µ = .91,SD = .71) did not report significantly different levels of perceived decedent distress (P = .10), but reported lower levels (P = .001) than wives (µ = 1.02,SD = .8) or daughters (µ = 1.16,SD = .82). Daughters reported higher levels of distress than wives (and husbands and sons), both personally (P = .004) and for decedents (P = .01). In the regression model, the GDI was twice as powerful a predictor of caregiver distress (β = .31) as gender (β = .14), and 4 times as powerful as generation (β = .08). Many demographic variables were excluded because they did not contribute significantly to the predictive value of the model. (R2 = .139 (F[3,820] = 45.8, P < .001).

CONCLUSION:  Generation was important for women, but not men, while gender was important regardless of generation. Daughters reported the highest levels of both personal distress and perceived decedent distress. A more powerful predictor of caregiver distress, however, was the perception that the decedent was in distress due to symptoms.

CERVICAL CANCER SCREENING AMONG WOMEN IN CALIFORNIA: ARE WE CLOSING THE RACIAL/ETHNIC GAP?I. De Alba1; Q. Ngo-Metzger1; J. Sweningson1; F.A. Hubbell1. 1University of California, Irvine, Irvine, CA. (Tracking ID #117157)

BACKGROUND:  Nearly 14 percent of all new cases of cervical cancer nationwide are diagnosed in women residing in California. Minority population groups at high risk for cervical cancer may be failing to fully comply with screening recommendations.

METHODS:  We analyzed data from the 2001 California Health Interview Survey (CHIS), a population based telephone survey. The outcome measure was having a recent Pap smear (in the past 3 years). We performed a main analysis that included all women age 18 or older without a hysterectomy and two sub-analyses that included Hispanic or Asian women only. Logistic regression was used to determine the predictors of recent Pap smear use in each group. The ethnic group with the lowest bivariate screening rate was used as the reference group in the models.

RESULTS:  The mean age for all women was 41 years and 58% had more than a high school education, a usual source of health care (88%) and health insurance (84%). Half were White, 30% were Hispanic, 11% were Asian, 5% were Black and 4% were classified as other race. In the main analysis, Asians had the lowest unadjusted screening rates (70.9%) of all groups were subsequently selected as the reference group. In the adjusted model, Asians were the most disadvantaged with Blacks (aOR 4.28, 95% CI 2.96–6.18), Hispanics (aOR 4.30, 95% CI 3.47–5.31), Whites (aOR 3.02, 95% CI 2.59–3.52) and other races (aOR 2.17, 95% CI 1.43–3.28) all being significantly more likely to report having a recent Pap smear. In the Asian subanalysis, Vietnamese were the referent group due to having the lowest unadjusted screening rates (60.3%). In the adjusted model, Filipino, Korean, Japanese, Chinese and other Asians were not significantly different than Vietnamese. In the subanalysis of Hispanics, Mexicans had the lowest unadjusted screening rates (85.5%) and thus were selected as referent. Other Hispanics were just as likely as the Mexican group to report a recent screening; South Americans and Central American (aOR 1.59, 95% CI 1.05-2.43) were more likely although only the aOR for Central Americans was significant. Although the three models varied slightly, having a usual source of care, being insured, and higher SES were significant predictors of recent screening in all three models.

CONCLUSION:  In California, Black and Hispanic women are the most likely to report a Pap smear in the past three years as compared to any other racial/ethnic group, including Whites. Asians were the least likely to report recent Pap smear despite a more favorable sociodemographic profile. The increasing use of Pap smears among Hispanics and Blacks is encouraging; however, increased efforts are needed to eliminate continuing racial/ethnic disparities in cervical cancer burden.

CHANGE IN CONDITION AND PROGNOSIS IN CHRONIC HEART FAILURE.S.E. Hardy1; P. Peduzzi2; H.M. Krumholz1. 1Yale University, New Haven, CT; 2Cooperative Studies Coordinating Center, VA Connecticut Healthcare System, West Haven, CT. (Tracking ID #116956)

BACKGROUND:  Clinicians’ ability to prognosticate for chronic conditions such as heart failure is poor. Change in condition over time is rarely considered in research on prognosis in heart failure. The objective of this case study is to determine whether adding change-in-state variables significantly improves the fit and discrimination of prognostic models for six-month all-cause mortality in chronic heart failure.

METHODS:  This nested case-control study used data from the Veterans Affairs Vasodilators in Heart Failure Trial (VHeFT II), which included men with clinical heart failure and cardiac dysfunction confirmed by imaging. Cases were selected from those VHeFT II subjects who died during follow-up. Up to three matched controls were randomly selected for each case from those subjects who survived the six-month interval during which the case died. A set of candidate predictive factors was selected based on the literature, and the independent predictors of six-month mortality were identified using conditional logistic regression with backwards elimination. For each independent predictor that changed over time, the change in that factor over the prior six months was added to the model. The fit and discrimination of the models with and without change were calculated and compared.

RESULTS:  200 cases and 596 matched controls were identified from the 804 trial participants. Treatment group (enalapril versus hydralazine and isosorbide), New York Heart Association class, serum sodium, ejection fraction, and peak oxygen consumption were identified as independent predictors of six-month mortality. When the change in these factors was added to the model, change in serum sodium and change in ejection fraction were both independent predictors of mortality, with odds ratios (95% confidence interval) of 1.07 (1.01–1.14) per 1 mEq/L increase and 0.96 (0.93–0.99) per 1% increase, respectively. Inclusion of these two change variables increased the proportion of variance explained by 13%, and increased discrimination (measured by the c-statistic) from 0.646 to 0.768.

CONCLUSION:  Considering change in clinical factors over time in addition to the current clinical state significantly improves a prognostic model for six-month all-cause mortality among men with chronic heart failure, suggesting that incorporation of change in condition over time into prognostic models would improve prognostication in heart failure, and perhaps in other chronic conditions as well.

CHANGES IN NATIONAL PATTERNS OF TREATMENT FOR DIABETES MELLITUS, 1992–2001.N.L. Sehgal1; J. Ma1; R.S. Stafford1. 1Stanford Prevention Research Center, Stanford, CA. (Tracking ID #116495)

BACKGROUND:  Diabetes mellitus (DM) is reported to affect 12 million Americans with an additional 12 million classified as having prediabetes. Medication treatment for DM accounts for $12 billion in annual expenditures. Using national data, we assessed the impact that newly expanded therapeutic options for DM have had on patterns of DM treatment.

METHODS:  We used 1992–2001 National Ambulatory Medical Care Survey (NAMCS) and National Hospital Ambulatory Medical Care Survey (NHAMCS) to estimate patterns of DM medication use for patients visiting private physicians and hospital outpatient departments. These nationally representative surveys allow assessment of national practice patterns. The main outcome measures were trends in number of DM patient visits, the percentage of visits where any DM drug was prescribed, and the drugs and drug classes selected for treatment. Odds ratios for the reported use of DM medications by various predictors were obtained via multivariate logistic regression using SUDAAN.

RESULTS:  Patient visits for DM increased significantly over the study period while a shift occurred in therapeutic choices. Patient visits increased by 73% from an estimated 27.7 million (M) in 1992 to 47.9 M in 2001. The proportion of patient visits where DM medication use was reported remained stable between 49–55%. Insulin use, however, declined by 38% from its peak use in 26% (99% CI: 20–31%) of patient visits in 1993 to 16% (12–20%) in 2001. In contrast, oral agents experienced a 72% increase in use from 30% (25–35%) of patient visits in 1992 to 43% (38–48%) in 2000. Among specific medications, metformin use increased significantly from 2% of patient visits in 1995 (year of FDA approval) to 16% in 2001. Similarly, thiazolidinedione use was 3% in 1997 (year of FDA approval) and increased to 13% in 2001. Overall, greater likelihood of DM medication use at visits was independently associated with younger patient age and private insurance coverage, but no associations were noted by gender, race, region of practice, or among primary care specialists. Greater insulin use was associated only with younger patient age and no significant associations were noted with respect to oral agent use.

CONCLUSION:  At the same time that the volume of DM care has increased, the development of new therapies has contributed to a shift in DM management towards more expensive therapies. The continued introduction of new medications within existing classes and the growth of combination therapy will continue to impact treatment patterns and, by extension, treatment costs. Future clinical guidelines for DM should consider the cost effectiveness of these new medications in determining final recommendations to providers.

CHARACTERISTICS OF PRIMARY CARE PHYSICIANS WHO TREAT WHITES AND BLACKS IN THE UNITED STATES. P.B. Bach1; H.H. Pham2; D. Schrag3; R. Tate1; J. Hargraves2. 1Memorial Sloan Kettering Cancer Center, New York, NY; 2Center for Studying Health System Change, Washington, DC; 3Sloan-Kettering Institute for Cancer Research, New York, NY. (Tracking ID #101930)

BACKGROUND:  African Americans receive inferior quality healthcare compared to whites. One plausible explanation for this disparity is that African American patients receive their care from a subset of US physicians who have inferior qualifications or resources than those of physicians who treat whites.

METHODS:  Cross-sectional analysis of Medicare claims for 150,391 visits of African American and white beneficiaries aged 65 and older for “evaluation and management” with 4,355 United States primary care physicians who participated in the 2000–2001 Community Tracking Study, a survey of a nationally representative sample of physicians involved in direct patient care.

RESULTS:  African American Medicare beneficiaries received primary care from a small group of physicians (80% of visits provided by 25% of physicians). Compared to their white counterparts, African American Medicare beneficiaries were more likely to have visits with physicians who were black (22.4% vs. 0.7%, P < .001), not board certified in their primary specialty (22.6% vs. 13.9%, P = .02), and who derived a greater percentage of their practice revenue from Medicaid (13.4% vs. 9.3% of practice revenue, P < .001). African American patients also had a higher proportion of visits with physicians who reported that they were not able to consistently provide high quality care (27.8% vs. 19.3%, P = .005), or secure access for their patients to high quality ancillary services (37.7% vs. 27.8%, P = .02), high quality diagnostic imaging (25.2% vs. 17.0%, P = .01), or elective admission to the hospital (47.8% vs. 36.1%, P < .001). These differences in qualifications and resources persisted in adjusted analyses controlling for the physician's practice setting and payment mix.

CONCLUSION:  Primary care for African Americans is provided by an identifiable subset of physicians, who, relative to physicians involved in care of white patients, are less well credentialed and face greater obstacles to the delivery of high quality care. Interventions to reduce racial disparities should target the subset of providers who assume most responsibility for treating African Americans, and should be designed to address potential gaps in their training and access to resources.

CHIEF RESIDENTS’ SKILLS WITH EVIDENCE-BASED MEDICINE: STILL HAZY AFTER ALL THESE YEARS.R. Watkins1; M.C. Wilson2; S. Richardson3. 1Wake Forest University, Winston-Salem, NC; 2University of Iowa Hospitals and Clinics, Iowa City, IA; 3Wright State University, Dayton, OH. (Tracking ID #117412)

BACKGROUND:  Chief Residents (CRs) represent the best outcomes of our residency programs, and they will serve in a variety of clinical and teaching settings. We wondered if educators have made progress in preparing these residents to practice and teach EBM.

METHODS:  At the 1999 and 2003 National Chief Resident Workshop of the Association of Program Directors in Internal Medicine, we surveyed attendees at 2 EBM seminars about their skills and confidence to assume teaching responsibilities in this area. Participants (n = 288: 169 in 1999, 119 in 2003) completed a 2 page questionnaire composed of both open-ended and 5-point ordinal scale responses; 88% would serve as CR during the PGY-4 year, 61% were university-based, and 46% were female.

RESULTS:  Although an increase had occurred, only 39% of 2003 respondents compared to one-quarter of the 1999 sample felt confident to use an evidence-based approach in their clinical practice. Little change had occurred in their confidence to teach others how to make evidence-based decisions (25% of the 2003 sample and 18% of the 1999 respondents). Small minorities of either sample (20% in 2003; 10% in 1999) felt that their programs had prepared them to teach EBM, and less than 21% of all respondents were comfortable applying JAMA Users’ Guides to the literature they read. Less than 10% of our total sample felt comfortable teaching others to do so. When stratified by gender or training site (community vs. university-based), no significant differences emerged.

CONCLUSION:  During this 5-year window, this large cross-sectional survey raises significant concerns that many rising CRs have not been adequately prepared by their residency programs to practice—much less teach—in an evidence-based fashion. Even more concerning is that this self-reported survey likely represents an overestimation of their actual skills, and we can only speculate as to where the EBM skills of all the other resident graduates would fall. Residency program directors should assess the state of their program's educational efforts to foster evidence-based practice and consider adopting multiple complementary opportunities for residents to hone and reinforce their EBM skills.

CHILDHOOD SOCIOECONOMIC STATUS PREDICTS HEALTH DECLINE IN OLDER ADULTS.S. Moody-Ayers1; K. Lindquist1; K. Covinsky1. 1UCSF/VAMC, San Francisco, CA. (Tracking ID #116567)

BACKGROUND:  Recent evidence indicates that childhood socioeconomic factors (SES) acting over a lifetime influence health outcomes in adults. However, few studies have examined the influence of these factors on the health status of older adults in the U.S. Our goal was to examine the influence of childhood SES (between birth and 16 yrs) on health status in later life.

METHODS:  We studied 20,566 community-dwelling adults aged 50 and older from the 1998 Health and Retirement Survey, a national cohort study. Our outcome, health decline, was defined as the occurrence of any of the following at 2 yrs: increased difficulty in basic activities of daily living or mobility, nursing home placement, or death. Childhood SES measures consisted of family financial status (described as well off, average, or poor), absence of a father at home, family hardship (family needed financial help or had to move because of financial difficulty), and parental education. We used logistic regression to examine the impact of childhood SES after first adjusting for age, gender, and race; and then also adjusting for the subject's current SES (education, income, and net worth).

RESULTS:  The mean age was 66.8 yrs; 57% were female; 70% were white, 14% black, and 8% Latino. 29.1% declined in health over 2 yrs. Rates of health decline were highest in those with poor childhood financial status (33.4% vs. 28.4%), family hardship (31.7% vs. 29.5%), absent father (34.4% vs. 29.2%), and low parental education (40.2% vs. 26%). As shown in the table, all measures of low childhood SES persisted after adjusting for age, gender, and race (P < .05 for all). After further adjusting for current SES, the association persisted only for family hardship.

CONCLUSION:  Childhood SES, particularly family hardship, appears to have a lifelong impact on health outcomes even when age, gender and race are considered. However, our results indicate that family disadvantages during childhood may be overcome if future opportunities for educational and financial achievement are attained.

Table 6. Childhood Family SES and Health Decline Over 2 Years, OR (Confidence Interval)
Childhood Family SESAdjusted for Age, Gender, RaceAdjusted Also for Current SES
Poor Financial Status1.23 (1.05–1.43)0.92 (0.79–1.08)
Family Hardship1.17 (1.08–1.27)1.12 (1.03–1.22)
Absent Father1.14 (1.01–1.29)1.05 (0.93–1.19)
Parental Education: 7 yrs1.48 (1.33–1.64)1.10 (0.98–1.22)

CLINICAL UTILITY OF FOLIC ACID TESTING FOR ANEMIA AND DEMENTIA SCREEN.M.J. Ashraf1; M. Goyal1; K. Hinchey2; J.R. Cook3. 1Baystate Medical Center/ Tufts University, Springfield, MA; 2Baystate Medical Center, Springfield, MA; 3Tufts University, Springfield, MA. (Tracking ID #116770)

BACKGROUND:  Folic acid tests are routinely ordered by physicians for the evaluation of anemia irrespective of MCV value. Folate testing is also considered as part of dementia screen and altered mental status. However, we feel that folic acid deficiency in Americans is extremely rare especially due to fortification of breakfast cereals and other dietary foods. HYPOTHESIS: Folic acid deficiency will not be a cause of anemia and dementia in USA population and there will be no correlation between folic acid testing and MCV values.

METHODS:  We designed a retrospective study to review all the serum folic acid tests done at our laboratory during one year. So far we have reviewed 1007 tests done in four months. We determined the frequency of low folate levels. We also recorded mean corpuscular volume (MCV), hemoglobin and hematocrit of all patients. These values were collected from the complete blood count (CBC) performed immediately before ordering the folic acid test. We also recorded B12 and TSH if done at the same time of folic acid test. Cost of folic acid testing was also analyzed.

RESULTS:  1007 folic acid tests were done on 980 patients. 27 patients had more than one test. The average age was 63.80 years. 66% of the tests were from out-patient facilities including clinics and nursing homes. Out of total 1007 tests, only one folic acid level was low. Overall 58% patients were not anemic. 430 patients (43.4%) were over the age of 70. More than half of the older patients had hematocrit greater than 34%. All the patients without anemia had folic acid tests done as part of dementia screen or due to altered mental status. Folic acid level was normal in all these patients. MCV was normal in 80% of patients while 11.6% had MCV greater than 100FL. None of the patients with macrocytosis had low folate level. There was no correlation between MCV and folic acid or between anemia and folic acid level with coefficient values(r) 0.0060 and 0.1827 respectively. No association observed between dementia and low folic acid level.

CONCLUSION:  Folic acid deficiency is extremely rare in Americans. There is no correlation between MCV and folic acid testing. Other causes of anemia and macrocytosis should be considered before ordering folic acid test. Folic acid deficiency does not appear to be a cause of dementia or altered mental status and routine ordering of folic acid test is not cost effective.

CLINICAL VIGNETTES TO EXAMINE THE EFFECT OF FATIGUE ON QUALITY OF CARE.J.B. Sussman1; S. Jain1; A.K. Jha2; J. Luck3; J.W. Peabody1. 1University of California, San Francisco, San Francisco, CA; 2Brigham and Women's Hospital, Boston, MA; 3Veterans Affairs Greater Los Angeles Healthcare System, Los Angeles, CA. (Tracking ID #116915)

BACKGROUND:  The exact effects of extended duty hours on medical residents and patients has gained new importance since the Accreditation Council on Graduate Medical Education's (ACGME) recent requirements limiting residents’ work hours. In spite of the ACGME regulations, questions remain about the effect of fatigue on medical care. One piece of evidence that is lacking in the research is examination of the effect of overnight call on the quality of the patient encounters that constitute the daily practice of the resident physician. The goal of this study was to examine the effect of a single night of call on the quality of the medical encounter using computerized clinical vignettes. Vignettes have been well-validated in other research and shown to simulate doctor's performance.

METHODS:  17 randomly selected internal medicine residents at an academic Veterans Affairs Medical Center were provided with 4 different computer-based clinical vignettes at each of 2 sittings—once after a regular work-day (baseline) and another at the end of a complete work-day after a complete night on-call (post-call). The scorer was blinded to call status and participant identification. Participants were aware of the study design. All comparisons between post-call and non-call participants used paired t tests. All data was gathered before the new ACGME regulations went into effect.

RESULTS:  Post-call residents reported significantly fewer hours of sleep the previous night than non-call residents (3.8  1.6 hrs SD vs 7.3  0.37 SD, P < .001) and more subjective fatigue (2.6  0.87 vs 3.5  1 P < .01 on a 5-pt scale). They expressed the same subjective mood (3.94  0.23 SD vs. 3.71  0.21, P = .22 on a 5-pt scale). Non-call and post-call scores were correlated with an r = 0.57. Call status did not predict overall performance on the vignettes (68.3  1.7 baseline vs 67.3  1.7 post-call, P = .28). Sub-analyses showed no effect of call status on test performance by case complexity, by specific medical skill domain (history, physical exam, lab ordering, diagnostic accuracy, treatment accuracy), or by self-reported amount of sleep either of the prior 2 nights. Residents’ performances on the vignettes did not improve with level of training.

CONCLUSION:  If a single night of call has an effect on the patient encounter, it is not very large. This study is one of the first to examine how fatigue affects the individual domains of medical practice and one of few studies to examine the effects of fatigue using methods that have been validated to approximate actual medical care. In this study, however, small sample size limits our ability to measure smaller night call effects. Future work, including longer-term studies, may be useful to see if our studies are robust.

CLUSTER-RANDOMIZED CONTROLLED TRIAL OF THREE DIFFERENT INTERVENTIONS TO IMPROVE ANTIHYPERTENSIVE PRESCRIBING IN PRIMARY CARE.S.R. Simon1; S.R. Majumdar2; K.P. Kleinman1; S. Salem-Schatz3; C. Warner4; L. Prosser1; I. Miroshnik1; S.B. Soumerai1. 1Department of Ambulatory Care and Prevention, Harvard Medical School, Boston, MA; 2University of Alberta, Edmonton, Alberta; 3HealthCare Quality Initiatives, Newton, MA; 4Harvard Vanguard Medical Associates, Medford, MA. (Tracking ID #116957)

BACKGROUND:  Academic detailing, also called educational outreach, has been shown to improve individual physicians’ prescribing practices but is perceived to be costly and labor-intensive and is therefore not widely used. Therefore, we compared traditional one-on-one individual academic detailing (IAD) with group academic detailing (GAD), and compared these strategies to a typical mailed guideline dissemination program (MG) within one large managed care organization to improve the use of antihypertensive medications in primary care.

METHODS:  We conducted a cluster-randomized controlled trial, allocating 3 practice sites to IAD (N = 235 prescribers and 2478 patients), 3 to GAD (N = 227 and 2352), and 3 to MG (N = 319 and 3575). The goal of the intervention was to increase the use of diuretics and beta-blockers (DIUR-BB), the guideline-recommended first-line agents. The IAD intervention consisted of a single physician visit following established principles of academic detailing; the GAD intervention was a single group session incorporating those principles, also led by a trained physician. With an intention-to-treat analysis and the individual patient as the unit of analysis, we used generalized estimating equations (GEE) to account for clustering at the level of the prescriber.

RESULTS:  At baseline, sociodemographic characteristics and rates of prescribing DIUR-BB among newly diagnosed and treated patients with hypertension were almost identical in the three experimental groups: IAD = 57.6%, GAD = 59.1%, and MG = 57.6%. In the 9-month period following the intervention, rates of DIUR-BB use increased by 12.5% in IAD, 13.2% in GAD, and 6.2% in MG. As compared with MG patients, DIUR-BB use among patients with newly diagnosed hypertension was more likely in both IAD (OR 1.40; 95% CI, 1.07–1.84) and GAD (OR 1.30; 95% CI, 0.89–1.90). The effects of IAD and GAD were of similar magnitude. There was no apparent effect of the intervention on rates of switching patients previously treated with medications other than DIUR-BB to the guideline-recommended agents.

CONCLUSION:  Academic detailing, whether one-on-one or with small groups, improves antihypertensive prescribing over and above the creation and dissemination of guidelines. If GAD is found to be as effective and less expensive than IAD, it may represent a more attractive option for improving practice.

COMPARING PRIMARY CARE CLINICIAN AND PHYSICIAN MANAGERS USES AND VIEWS’ OF EVIDENCE.K. Lorenz1; G. Ryan2; K. Chan2; S.C. Morton3; P.G. Shekelle2. 1VA Greater Los Angeles Healthcare System, Los Angeles, CA; 2RAND, Los Angeles, CA; 3RAND, Santa Monica, CA. (Tracking ID #115605)

BACKGROUND:  In light of the continuing gap between evidence and practice, we aimed to understand how clinicians and managers differ as information clients, and the implications for improving upon the EBM model focused on the clinician end user.

METHODS:  19 primary care physicians with >3 years experience and 16 physician managers within moderate to large organizations in Southern California participated in 2 clinician and 2 manager focus groups. Using examples as a realistic backdrop, the groups discussed 1) why physicians initiate information searches and 2) strategies for information search. Verbatim transcripts were reviewed for text that addressed the 2 major topic areas. Text segments were marked, extracted, and independently coded to identify themes using standard techniques. Team deliberation and consensus determined final theme lists.

RESULTS:  Both groups mentioned common rationales for information search including unfamiliarity, treatment uncertainty, cost, need to justify decisions, desire to make programmatic changes, and personal interest. However, clinicians invoked “non-evidence-based common sense” in typical care decisions. In contrast, managers emphasized cost and a frequent need for information. A manager explained, “As a doctor for an individual patient, I can go to my colleagues and consultants and trust because of experience how to make decisions for that particular patient, but as a manager…Everything has to be verifiable in the literature to make broad decisions for a whole group of people.” Both groups upheld trustworthiness as the most important attribute of information because neither group has time to adequately assess evidence nor do they trust themselves to make informed choices. A clinician explained, “The last thing you want me [to do] as a general internist [is] to have to review every little article on every little thing. I can’t do that. I totally don’t trust myself to do it. I absolutely don’t.” Of lesser importance, participants mentioned themes of statistical simplicity, consistency, accessibility, brevity, and practicality.

CONCLUSION:  1) EBM interventions that target managers and questions of cost are much more likely to effect health outcomes than targeting clinicians and individual patient treatment. 2) Neither group expressed comfort with basic evidence appraisal, and clinicians rarely search, so teaching clinicians to appraise the primary literature is unlikely to improve practice. 3) Both clinicians and managers need trusted, accessible advice. Physician managers are ideally poised to select and implement successful strategies (e.g. guidelines, reminders) because they represent both clinician and system values.

COMPLICATIONS AND PROCEDURES AFTER BILATERAL PROPHYLACTIC MASTECTOMY.C.N. West1; M.B. Barton2; A.I. Liu1; A.M. Geiger1. 1Southern California Permanente Medical Group, Pasadena, CA; 2Harvard University, Boston, MA. (Tracking ID #117121)

BACKGROUND:  Bilateral prophylactic mastectomy is known to be highly efficacious but there is little known about the complications or subsequent procedures needed to rectify complications or cosmetic problems. We therefore studied the complications and procedures occurring after bilateral prophylactic mastectomy.

METHODS:  Automated hospitalization and cancer registry records were used to identify women without breast cancer who underwent bilateral mastectomy at one of six health maintenance organizations between 1979 and 1999. Structured medical record reviews confirmed that bilateral mastectomies were done for prophylactic reasons, identified timing of initial reconstruction and ascertained complications and subsequent procedures.

RESULTS:  During the 20 year study period 270 women underwent bilateral prophylactic mastectomy. The majority of the women (90%) were Caucasian and the median age at surgery was 44 years (range 23 to 74). Most women (179, 66%) had simultaneous reconstruction but 36 (13%) had delayed reconstruction and 55 (20%) had none. After bilateral prophylactic mastectomy 466 complications occurred in 171 (63%) women, with a median of two per woman (range 1 to 13). Over half (55%) required repair, including excessive scarring and implant leakage or rupture. About a third (167, 36%) were temporary, including hematoma, hemorrhage and infection. The remaining 42 (9%) of complications, including lymphedema and depression, were of a permanent or psychological nature. A total of 822 subsequent procedures were performed in 167 (62%) women, with a median of 4 per woman (range 1 to 22). Nearly all (766, 93%) were cosmetic in nature. Complications and subsequent procedures were less common in women with no reconstruction (chi square P = .067 and P = < .001 respectively) but occurred in nearly identical proportions among women with simultaneous or delayed reconstruction (chi square P = .764 and P = .958 respectively).

CONCLUSION:  Women who choose to undergo bilateral prophylactic mastectomy may experience a range of complications and the potential for additional procedures after reconstruction increases. Women and their physicians need to weigh these risks with the potential benefits of bilateral prophylactic mastectomy.

COMPREHENSIVE CARE FOR WOMEN VETERANS: INDICATORS OF DUAL USE OF VA AND NON-VA PROVIDERS.B.A. Bean-Mayberry1; C.H. Chang2; M.A. McNeil2; P.M. Hayes1; S.H. Scholle3. 1VA Center for Health Equity Research and Promotion, VA Pittsburgh Healthcare System, Pittsburgh, PA; 2University of Pittsburgh, Pittsburgh, PA; 3National Committee on Quality Assurance, Washington, DC. (Tracking ID #117326)

BACKGROUND:  We initiated this Veterans’ Administration (VA) study to identify the health care experience factors (e.g., provider gender, gynecological care by provider, overall satisfaction, and use of a VA women's clinic) that influence dual use of VA and non-VA regular providers among women veterans. We hypothesized that women who receive comprehensive (general and gender-specific) care in the VA would be less likely to demonstrate dual use and that lower overall patient satisfaction with primary care would positively influence dual use of VA and non-VA regular providers.

METHODS:  An anonymous survey was mailed to a random sample of women veterans from a regional network of ten VA medical centers to evaluate satisfaction with primary care (Response Rate 61%, N = 1321). A secondary analysis of dual use of providers was performed with the subgroup of women (N = 1051) who identified their VA regular provider. Multiple logistic regression analyses were performed to determine the effect of each health care experience factor on use of a non-VA regular provider with adjustment for patient demographics, perceived health status, VA service connection, and clustering by site.

RESULTS:  In multivariable analyses, provision of routine gynecological care by the VA provider (OR 0.37, 95% CI 0.22, 0.60) and use of VA women's clinics (OR 0.56, 95% CI 0.35, 0.90) were significantly associated with a lower likelihood of dual use. However, overall dissatisfaction (OR 1.88, 95% CI 1.04, 3.41) and income (OR 1.89, 95% CI 1.32, 2.71) were the characteristics significantly associated with an increased likelihood of dual use.

CONCLUSION:  Women veteran's use of VA and non-VA regular providers is not associated with provider gender but is influenced by modifiable factors such as satisfaction or a broader scope of services from a provider or clinic setting. As a result, primary care settings should promote routine gynecological care within primary care or linked with specialized women's clinics to enhance coordination and comprehensiveness and, thus, reduce fragmentation of care or potential overuse of care across health systems.

COMPUTER-GENERATED DECISION SUPPORT CAN PROMOTE BEHAVIORAL CHANGE AND PREVENTIVE SCREENING.N.F. Col1; J.M. Fortin1; L. Ngo2; A.M. O’Connor3; R. Goldberg4. 1Brigham and Women's Hospital, Boston, MA; 2Harvard School of Public Health, Boston, MA; 3University of Ottawa, Ottawa, Ontario; 4University of Massachusetts Medical School (Worcester), Worcester, MA. (Tracking ID #117489)

BACKGROUND:  Most North American women are overweight and sedentary, which increases their risks for many chronic diseases. Time pressures and inadequate training limit clinicians’ ability to effectively counsel patients about behavioral change. We developed a personalized, computer-generated decision support aid (DSA) to help clinicians counsel their patients about behavioral change and appropriate screening. Our objective was to test the impact of the DSA on lifestyle and preventive screening behaviors.

METHODS:  We developed decision support software that provides women with 1) personalized risk estimates for coronary heart disease, breast cancer, and osteoporotic hip fracture, 2) personalized information about their risk factors for common chronic conditions such as colorectal cancer; and 3) strategies for risk reduction and early detection. A randomized control trial was conducted among postmenopausal women in 4 clinical practices comparing a control group (ACP brochure) to receipt of the DSA. Healthy menopausal women between the ages of 45 and 65 were recruited through clinic rosters. Participants completed a questionnaire about their lifestyle and medical history; responses were entered into a software program that generates a 35+ page personalized health report (the DSA) for the patient and a summary page for her clinician. Lifestyle and screening behaviors were measured at baseline and 2 weeks after the clinic visit. Chi square analyses examined intent to change various lifestyle and screening behaviors among those at-risk (exhibiting the risky behavior or not presently undergoing the screening test).

RESULTS:  Of the 145 women included in the trial, 99 had valid responses. The mean age was 52.3, 98% were white, and 58% had completed some college. No differences in intent to change behavior were present at baseline. At 2 weeks, women who received the intervention were more likely than controls to intend to decrease alcohol consumption (50% vs. 0%), eat a low fat diet (82% vs. 50%), eat more fruits and vegetables (87% vs. 62%), exercise more (100% vs. 80%), take daily calcium and vitamin D (60% vs. 19%), and undergo flexible sigmoidoscopy (83% vs. 13%) and fecal occult blood test (79% vs. 33%), although only the latter 3 were statistically significant.

CONCLUSION:  Computer-generated personalized information can promote behavioral change and screening among women, presenting a novel approach to helping clinicians counsel their patients. More work is needed examining the impact of this approach on observed behaviors and on sustained behavior change.

COMPUTERS IN THE CLINIC: IMPACT ON THE DOCTOR-PATIENT RELATIONSHIP.E. Rouf1; K. Felix2; C. Okamora2; M.D. Schwartz2. 1University of Kansas, Kansas City, KS; 2New York University, New York, NY. (Tracking ID #116991)

BACKGROUND:  Physicians are increasingly dependent on computers and Electronic Medical Records (EMR) for clinical care. Although EMR may improve accuracy and efficiency of care, its impact on the doctor-patient relationship as a “3rd entity” in the encounter remains unclear.

METHODS:  This cross-sectional study at a VA Primary Care Clinic included a physician survey, post-visit exit surveys for physicians and patients, and chart reviews. Main measurements were satisfaction with the visit, value of the EMR, perceived time spent interacting with the computer, and computer interference with the doctor-patient relationship.

RESULTS:  23 physicians (13 male, 12 residents, and 11 attendings) and 155 of their patients enrolled in the study (mean 7/physician, range 6–10). Patients were 97% men, 10% Hispanic, 28% Black, mean age of 71 years, with 9 medical problems, and had seen their physician 7 times. Patients were more likely than physicians to rate satisfaction with the visit as excellent, 42.9% vs. 5.3%, P < .001. 84% of patients and 52% of physicians felt that computers improved the quality of care received and given, respectively, P < .001. However, patients, more than physicians, agreed that because of the computer, doctors spent less time than they liked looking at patients (34.4% patients vs. 22.2% physicians, P = .3), talking to patients (24.7% patients vs. 5.0% physicians, P = .05), and examining patients (22.2% patients vs. 0% physician, P < .05). More patients were concerned than physicians that the EMR made the visit less personal 12.3% vs. 0.0%, P = .08). Physicians felt they spent a higher proportion of the visit directly interacting with the computer than did their patients, 37.9% vs. 22.8% of the visit, P = .003. Patients were less satisfied and less likely to recommend the doctor to a friend the more time they thought the physician spent interacting with the computer.

CONCLUSION:  Patient satisfaction was high and both patients and physicians felt the EMR improved quality of care. However, patients were more concerned than their physicians about the computer's negative impact on their relationship and the process of care. Educational interventions to improve communication skills in the EMR era are warranted.

CONCURRENT USE OF PRESCRIPTION DRUGS AND HERB/DIETARY SUPPLEMENTS IN AMBULATORY CARE: PREVALENCE, RISK FACTORS, AND ADVERSE DRUG EVENTS.R.E. Graham1; T.K. Gandhi2; J. Borus2; A. Seger2; E. Burdick2; D.W. Bates2; R.S. Phillips1; S.N. Weingart1. 1Harvard Medical School, Boston, MA; 2Brigham and Women's Hospital, Boston, MA. (Tracking ID #116385)

BACKGROUND:  Despite concerns about possible harmful interactions between prescription drugs and herb/dietary supplements little is known about concurrent use in ambulatory patients. To estimate the potential risk, we determined the prevalence of herbs and dietary supplements use among ambulatory care patients on prescription medications, the factors associated with their use, and their association with adverse drug events (ADEs).

METHODS:  We studied 661 patients who received prescription medications at four primary care practices in metropolitan Boston. Investigators interviewed patients by telephone and reviewed their medical records in order collect information about the use of herbs and dietary supplements, prescription medications, ADEs, and patients’ clinical histories and sociodemographic characteristics. We used bivariable analysis and multivariable logistic regression models to examine factors associated with the use of herbs and supplements. A multivariable model was used to examine whether herb and dietary supplement use was an independent correlate of ADEs.

RESULTS:  Of 661 patients enrolled, the mean age was 53 (range 19–100); 66% were female, 8% were non-English speaking, and 18% were non-white. 102 (15.4%) patients reported using herb and non-herb dietary supplements. The most commonly used herbs were: echinacea (21.7%), ginkgo (13.7%), and St. Johns Wort (6.8%). The most commonly used non-herb dietary supplements were: glucosamine (13.7%), omega-3 fatty acids (12.7%), and garlic (7.8%). Compared to non-users, herb and supplement users were more often white (88% vs. 79%, P = .03), college educated (90% vs. 80%, P = .02), English speaking (98% vs. 91%, P = .02), and had 3 or fewer years of continuous care at the practice site (44% vs. 34%, P = .03). In multivariable analyses adjusted for age, race, gender, and number of prescription medications, herb and supplement use was associated with college education (AOR 2.3, 95% CI [1.1, 4.6]), English speakers (AOR 4.3, 95% CI [1.1, 18.5]), and continuous care (AOR 0.8 for each additional year, 95% CI [0.7, 1.0]). Herb and dietary supplement use conferred no significant increased risk of ADEs in univariable or multivariable models.

CONCLUSION:  Use of herbs and dietary supplements was common in this population, with certain groups being much more likely to use these agents. Although we observed no increased risk of ADEs among patients using herbs and dietary supplements, the possible risk of interaction between these agents with prescription medications merit further study.

CONSISTENCY OF PHYSICIAN DISABILITY ASSESSMENT.E.C. O’Fallon1; S. Hillson2. 1Hennepin County Medical Center, Minneapolis, MN; 2University of Minnesota, Minneapolis, MN. (Tracking ID #117336)

BACKGROUND:  Completing a disability assessment is a common physician task. Our previous research has shown that many physicians find it a particularly uncomfortable responsibility. Interviews with physicians suggested that many felt ill prepared for assessing disability, and also concerned about the format of assessment documents. We hypothesize that these issues may lead to inconsistent disability assessments from physicians, further complicating the disability assessment process.

METHODS:  We created two different hypothetical patient scenarios requesting disability assessment. The survey included a written description of each history and exam and asked participants to choose a level of disability assessment for each patient from four options. The two cases were similar in age, gender and normal physical exam, but differed in the duration of symptoms. Case 1 represents an acute injury to an otherwise healthy adult. Case 2 describes a patient with many years of pain with associated depression and unemployment. The assessment options we included were adapted from a commonly used county government assessment form. Options ranged from no disability to complete disability, and offered space to describe specific temporary or permanent limitations, and the anticipated duration of the disability. The survey was distributed to resident and staff physicians at a large urban county hospital where disability assessments are frequently performed. Surveys were collected anonymously, coded and scored by the research team.

RESULTS:  35 surveys were returned. 91% of respondents found Case 1 as qualified for limited employment, but varied widely on the types of limitations. Responses to Case 2 were divided, 36% found no disability, 39% allowed for limited employment, 25% chose full disability until a specified date or until reassessed.

CONCLUSION:  Disability assessments are a challenging physician task. Although our survey used a written description of a patient scenario, rather than a real life assessment, our consistent results for Case 1 suggest a reliable physician response is possible. However, the variability in assessments of Case 2 reveals a wide discrepancy in physician opinions. These results may suggest increased physician variation in situations of more chronic disability. Alternatively the variability may stem from flaws in the currently used assessment forms, or in physician training for using them. Further research may help understand the variation in physician disability assessment and evaluate a potential underlying lack of validity in our current disability assessment system.

CONTENT OF PROSTATE CANCER SCREENING DISCUSSION AFFECTS PSA TESTING RATE IN A LOW-LITERACY POPULATION.J. Sharma1; E. Justice1; J. Justice1; S. Kripalani1; C.A. Spiker2; T. Jacobson1; L.E. Laufman2; A.D. Weinberg2. 1Emory University, Atlanta, GA; 2Baylor College of Medicine, Houston, TX. (Tracking ID #115711)

BACKGROUND:  Experts recommend that physicians counsel patients on the risks and benefits of prostate cancer screening, so that patients may make an informed decision about testing. Men with limited literacy skills, who suffer a disproportionate burden of prostate cancer, may have difficulty with such complex decision-making. This study examined the content of physician-patient discussions of prostate cancer screening and their relation with subsequent prostate-specific antigen (PSA) testing in an inner-city primary care clinic.

METHODS:  Subjects were men, age 45–70, participating in a randomized controlled trial to increase physician-patient discussion of prostate cancer screening. The two study interventions (a patient education handout and a flyer instructing patients to talk to their doctor) were given to patients in the waiting room before the physician appointment. They did not advocate for or against screening. After the appointment, patients completed a brief interview and the Rapid Estimate of Adult Literacy in Medicine (REALM), and we reviewed the clinic chart.

RESULTS:  Most of the 250 subjects (91%) were African-American, and 78.8% read below a 9th grade level by the REALM. Nearly half (48.4%, N = 121) reported discussing prostate cancer with their physician. Such conversations were more common among patients receiving either intervention (53.9% vs. 37.4% for control, P < .05), Black men (50.4% vs. 25.4% for White men, P < .05), men over age 60 (59.8% vs. 42.3%, P < .01), and patients with low health literacy (56.4% for reading level below 4th grade vs. 34.0% for high school reading level, P < .05). Of the 121 subjects who talked to their physician about prostate cancer, 47 (38.8%) reported discussing digital rectal examination (DRE), and 59 (48.8%) discussed the PSA test. Most patients who discussed PSA (88.5%) recalled that their physician said it was important to get the test, while few reported talking about the benefits and risks (29.5%) or that additional tests may be necessary (45.9%). There was a trend toward less PSA testing when the benefits and risks were discussed (11.1% vs. 25.6%). PSA testing was more common when the physician said it was important (24.5% vs. 0%).

CONCLUSION:  Although experts recommend a process of informed decision-making for prostate cancer screening, this occurred infrequently in an inner-city clinic. Physicians appeared to influence their patients’ PSA testing through the content of the discussions.

CONTINUED SUBOPTIMAL USE OF ASPIRIN IN PATIENTS AT HIGH AND MODERATE RISK FOR CHD EVENTS.R. Stafford1; J. Ma1. 1Stanford Prevention Research Center, Program on Prevention Outcomes and Practices, Stanford, CA. (Tracking ID #116585)

BACKGROUND:  Despite the benefits of aspirin therapy in the primary and secondary prevention of coronary heart disease (CHD) events, studies indicate that aspirin use falls short of current practice recommendations. To better understand these treatment gaps and identify patient and physician contributors to shortfalls in aspirin use, we studied national trends in aspirin use by CHD risk, focusing on moderate-to-high risk patients.

METHODS:  We used 1992–2001 National Ambulatory Medical Care Survey (NAMCS) and National Hospital Ambulatory Medical Care Survey (NHAMCS) to estimate the likelihood of reported aspirin use by CHD risk in patients visits to private physicians and hospital outpatient departments. The presence of CHD or other atherosclerotic diseases defined high CHD risk. Moderate risk included two categories: 1) a diagnosis of diabetes mellitus and 2) 2 or more major CHD risk factors among men <45 and women <55 or 1 or more risk factors among men >45 and women >55. Independent predictors of aspirin use were evaluated via multivariate logistic regression analyses using SUDAAN.

RESULTS:  Reported aspirin use grew only modestly from 21% (95% CI 18–23%) in 1992–93 to 24% (20–27%) in 2000–01 during visits by high-risk patients and from 4% (2%–5%) to 6% (4%–7%) among visits by patients with diabetes. Among moderate-risk patient visits defined by multiple risk factors, aspirin increased significantly from 3% (3%–4%) in 1992–93 to 7% (5%–8%) in 2000–01. Aspirin use was never higher than 2% among low risk patient visits. Furthermore, among visits where a statin was prescribed (an indicator of ongoing CHD risk reduction activity), aspirin use was reported in only 31% (26%–36%) of high-risk patients, 16% (9%–23%) of moderate-risk patients, and 11% (7%–14%) of low-risk patients. Higher CHD risk was independently predictive of aspirin use. In addition to CHD risk, greater aspirin use also was independently associated with advanced age (OR 3.3, 95% CI 2–5 for age 80+ yrs. vs. <45 yrs), male gender (OR 1.2, 1.1–1.3 vs. female) and cardiologist care (OR 3.2, 2–5 vs. internal medicine).

CONCLUSION:  Improvements in reported aspirin use for cardiovascular prevention were modest in the past decade and substantial treatment gaps have persisted, particularly for moderate-risk patients, including individuals with diabetes. Marked changes in clinical practice are unlikely to occur unless more aggressive, innovative means are implemented to enhance physician and patient adherence to consensus guidelines on aspirin therapy to prevent new or recurrent CHD events. In particular, targeted interventions may be warranted in patients with diabetes and/or multiple major CHD risk factors, as well as in at-risk subpopulations where aspirin use is lower than average.


BACKGROUND:  Randomized controlled trials have found that ximelagatran is as effective as warfarin at preventing strokes in patients with atrial fibrillation (AF). Because ximelagatran can be taken in a fixed, oral dose without INR monitoring, it has the potential to improve quality-adjusted survival. We assessed the hypothesis that ximelagatran offers a cost-effective alternative to warfarin and aspirin for stroke prophylaxis in patients with AF.

METHODS:  We designed a Markov decision model to simulate long-term health states, quality-adjusted survival, and economic outcomes in a hypothetical cohort of 70 year-old patients with AF, who had varying risk of stroke and were good candidates for anticoagulation. We compared ximelagatran, adjusted-dose warfarin, and aspirin therapy. Probabilities for stroke, hemorrhage, and death were obtained from meta-analysis of randomized controlled trials and observational studies. Quality-of-life estimates were obtained from previously published patient interviews. Costs were estimated from Medicare data and literature review. In the base case, the market price of ximelagatran was estimated at $1200 per year, based on the cost of clopidogrel.

RESULTS:  In our meta-analysis of 4 randomized controlled trials, ximelagatran reduced the relative risk of major hemorrhage by 26%. For patients with AF and no additional risk factors for stroke, stroke rates were low (<2%/yr of aspirin) and either warfarin or ximelagatran would cost >$50,000 per QALY compared with aspirin. At typical stroke risks (4%/yr), warfarin would cost <$50,000 per QALY vs. aspirin therapy, but ximelagatran would cost $71,000 per QALY vs. warfarin. For ximelagatran to cost <$50,000 per QALY in a typical patient, it would have to sell for <$1,000 per year or be prescribed only in patients at high risk of bleeding (4%/yr warfarin). The annual price of ximelagatran that maximized sales for AF occurred at an annual price of $650/yr. Projected US sales at this price were $1.2 billion/yr.

CONCLUSION:  At a price of $1,200 per year, ximelagatran would not be cost effective compared with warfarin except in patients at high risk of bleeding. At a market price of $650 per year, ximelagatran would be cost-effective for AF patients with additional risk factors for stroke.

COST-EFFECTIVENESS OF AUTOMATED EXTERNAL DEFIBRILLATORS: THERE's NO PLACE LIKE HOME?P. Cram1; S. Vijan2; A.M. Fendrick2. 1University of Iowa, Iowa City, IA; 2University of Michigan, Ann Arbor, MI. (Tracking ID #116493)

BACKGROUND:  A majority of out-of-hospital cardiac arrests occur at home in individuals who do not meet criteria for implantable defibrillators. In an effort to reduce mortality, the Federal Drug Administration recently approved the sale of automated external defibrillators (AEDs) for in-home use. We sought to evaluate the cost-effectiveness of in-home AED deployment for individuals with risk factors for cardiac arrest and explore the relationship between the probability of suffering an in-home cardiac arrest and the cost-effectiveness of AED deployment.

METHODS:  A Markov Decision Model was created to evaluate the cost-effectiveness of two alternative strategies: Strategy 1-individuals experiencing in-home cardiac arrest are treated by Emergency Medical Services equipped with AEDs (EMS-D); Strategy 2- individuals are initially treated with their in-home AED followed immediately by EMS (AED). Strategies differ only in the initial availability of an AED and its impact on cardiac arrest survival (15% absolute increase in survival) and cost ($1,500). The base-case annual rate of cardiac arrest was assumed to be 1%, derived from the incidence of sudden cardiac death in adults with multiple risk factors for coronary disease. 50% of these arrests were estimated to occur at home and the AED was assumed to be used on 50% of the arrest victims. In sensitivity analysis, the annual incidence of cardiac arrest was varied from 0.1% (all adults) to 5% (adults with prior cardiac events) to reflect the probability of cardiac arrest among different patient populations.

RESULTS:  Under the base-case assumption (1% annual risk of cardiac arrest) the cost per quality adjusted life-year (QALY) gained from in home AED was $320,000. As the annual probability of cardiac arrest increased from 0.1% to 5%, the cost per QALY gained declined from $3.2 million to $65,000 respectively. The cost-effectiveness of AED deployment was also sensitive to the impact of the AED on arrest survival, the probability that the AED was used on the arrest victim, and the likelihood that the AED was used on any person other than the individual for whom it was purchased.

CONCLUSION:  Targeted deployment of in-home AEDs for individuals with multiple cardiac risk factors or prior cardiac events appears to be relatively cost-effective and should be considered for coverage by third-party payers. Alternatively, universal deployment of in-home AEDs is expensive relative to commonly accepted cost-effectiveness thresholds.

COST-RELATED MEDICATION UNDER-USE: AN ANALYSIS OF TOUGH CHOICES BY CHRONICALLY-ILL ADULTS.J.D. Piette1; M. Heisler1; T.H. Wagner2. 1Ann Arbor VAMC and the University of Michigan, Ann Arbor, MI; 2Palo Alto VAMC and Stanford University, Menlo Park, CA. (Tracking ID #101798)

BACKGROUND:  Many chronically ill patients under-use prescription drugs due to out-of-pocket costs, yet little is known about the characteristics of medications that patients weigh when making these decisions. Using a survey of a nationwide sample of chronically-ill adults, we identified the extent to which patients are more likely to under-use symptom-relief medications (e.g., analgesics) versus “preventive” medications (e.g., antihypertensives) when facing similar cost pressures.

METHODS:  4,055 individuals 50 years or older and using prescription medication for at least one of five chronic health conditions were identified from a national community sample. Participants reported cost-related under-use for symptom-relief and preventive medications treating 16 chronic illnesses and the extent to which they cut-back on other basic necessities to pay medication costs. Prescription-level analyses were conducted controlling for patient-level covariates using hierarchical models.

RESULTS:  Eighteen percent of respondents reported one or more episodes of cost-related medication under-use. Rates of cost-related under-use were higher for medications treating symptoms, such as back pain (23%) or migraine headaches (21%), than for medications treating asymptomatic conditions such as hypertension (9%) or high cholesterol (12%). Controlling for out-of-pocket costs, income, and other sociodemographic and clinical characteristics, patients were more likely to restrict their use of symptom-relief than preventive medications due to cost (adjusted odds ratio or AOR: 1.6, 95% CI = 1.5–1.9). Patients using both medication types (N = 2,839) reported significantly higher odds of cost-related under-use for their symptom relief than preventive drugs (AOR = 1.5, CI = 1.4–1.7). Analyses indicate that patients taking both medication types not only were more likely to choose symptom-relief medications to forego, but often cut-back on food, heat, or other necessities before restricting their use of preventive treatments.

CONCLUSION:  Medication characteristics beyond cost alone influence chronically ill patients’ decisions to cut back on treatment in response to out-of-pocket medication costs. Contrary to many clinicians’ expectations, patients may value preventive medications more highly than symptom-relief medications and often forego basic life necessities rather than cut-back on preventive medication use. Given the limited prospects for medication insurance reforms, physicians must explore more fully patients’ medication choices and collaborate with them to develop effective regimens consistent with their preferences, ability-to-pay, and health needs.

COUNSELING OVERWEIGHT AND OBESE PATIENTS IN PRIMARY CARE: DESCRIPTION AND OUTCOME AT ONE YEAR OF FOLLOW-UP.N. Rodondi1; J. Humair2; W.A. Ghali3; C. Ruffieux4; R. Stoianov5; L. Seematter Bagnoud6; H. Stalder7; A. Pecoud8; J. Cornuz9. 1Dpt of Epidemiology and Biostatistics, UCSF, San Francisco, CA; 2Geneva University Hospital, Geneva, ; 3University of Calgary, Calgary, Alberta; 4University Institute of Social and Preventive Medicine, Lausanne, ; 5University Hospital, Lausanne, ; 6Lausanne University Hospital, Lausanne, ; 7Medical school of Geneva, CH-1211 Geneve 14, ; 8University Outpatient Clinic of Lausanne, Lausanne, ; 9Department of Medicine, Lausanne University Hospital, Lausanne, . (Tracking ID #116394)

BACKGROUND:  Primary care physicians are well placed to provide counseling for obese and overweight patients, but many believe that their counseling is not effective. No study has assessed counseling overweight and obese patients in current primary care practice with a prospective follow-up. We aimed to determine physician counseling received by those patients in primary care and to analyze the relationship between counseling and weight change over 1 year, as well as patients’ behavior to control weight.

METHODS:  Our prospective cohort study included 523 consecutive overweight and obese patients cared for by the 35 internal medicine physicians of two Swiss academic primary care clinics. The physicians were blinded to the study aims and did not receive specific training. We determined counseling received by patients based on 10 pre-defined counseling strategies for weight reduction and a score corresponding to the number of applied strategies. At one-year, we compared patients who received any form of weight counseling versus those who received no counseling. We used t test and multiple linear regression to compare weight change between groups, and chi-square tests to compare patients’ behavior to control weight.

RESULTS:  35% of patients received no counseling at all and the mean overall counseling score was 2.6 (range: 0–10). After 1 year, patients who had any form of counseling achieved a mean ± SD weight reduction of –1.0 ± 5.0 kg while those who received no counseling had a mean weight increase of +0.3 ± 5.0 kg (P = .02). A multiple linear regression demonstrated a pattern of progressive decrease of 0.2 kg for every counseling strategy provided, after adjusting for potential confounders. Patients counseled by their physician were more likely than those who received no counseling to change their behavior to control weight, such as setting a target weight (56% vs. 36%, P < .01) or visiting a dietician (23% vs. 10%, P < .01).

CONCLUSION:  Overweight and obese patients who received counseling from their primary care physicians had better weight outcomes at one year follow-up than did those who received no counseling. However, weight counseling by primary care physicians was sub-optimal.

CULTURAL COMPETENCE: A SYSTEMATIC REVIEW OF HEALTH CARE PROVIDER EDUCATIONAL INTERVENTIONS.M.C. Beach1; E.G. Price1; T.L. Gary1; K.A. Robinson1; A. Gozu1; A.M. Palacio1; C. Smarth1; M.W. Jenckes1; C. Feuerstein1; E.B. Bass1; N.R. Powe1; L.A. Cooper1. 1Johns Hopkins University, Baltimore, MD. (Tracking ID #117015)

BACKGROUND:  To reduce racial and ethnic disparities in care, efforts have been made to improve the cultural competence of health professionals. We synthesized the findings of studies evaluating interventions to improve cultural competence.

METHODS:  We performed a systematic review of literature from 1980 through June 2003 using electronic and hand searches to identify studies that evaluated interventions designed to improve the cultural competence of health professionals. Eligible studies were in English, with original data, and described an intervention evaluated with a pre- and a post-test or by comparison to a control arm. Two reviewers abstracted data on the effectiveness and cost of cultural competence training, and graded the strength of the evidence as excellent, good, fair or poor using predetermined criteria.

RESULTS:  Thirty-four studies were included in our review. Four studies were published between 1980–1989, 14 between 1990–1999, and 18 between 2000–2003. Targeted learners were most often nurses (n = 17) and physicians (n = 18). Most curricula focused on specific cultural content (n = 26), general concepts of culture (n = 19), language (n = 10), and patient-provider interaction (n = 8); fewer focused on health care access (n = 3), racism (n = 2), and socio-economic status (n = 2). Most curricular interventions used more than one training method. The most common training methods were group discussion (n = 17), lectures (n = 17), case scenarios (n = 12), clinical experiences (n = 10), presentations by members of another culture (n = 9), small group work (n = 9), and cultural immersion (n = 8). Evidence is excellent that training improves the knowledge of health professionals (17 of 19 studies demonstrated a beneficial effect). Evidence is good that training improves the attitudes and skills of health professionals (21 of 25 studies demonstrated a beneficial effect and 14 of 14 studies demonstrated a beneficial effect, respectively). Evidence is good that training impacts patient satisfaction (3 of 3 studies demonstrated a beneficial effect) and poor that training impacts patient adherence (although the one study designed to do this demonstrated a beneficial effect). No studies have evaluated patient health status outcomes. There is insufficient evidence to determine the costs of cultural competence training.

CONCLUSION:  Cultural competence training shows promise as a strategy for improving the knowledge, attitudes and skills of health professionals. However, evidence is lacking that it improves patient adherence, health outcomes and equity of services across racial and ethnic groups. Future research should focus on these outcomes and determine which teaching methods and content are most effective.

CURRENT CONFLICT OF INTEREST DISCLOSURE POLICIES OF MAJOR MEDICAL JOURNALS.A. Riley1; E. Teeter1; P. Watson2; J. Musial2; A. Khandelwal2; J.D. Buckley2. 1Henry Ford Hospital Detroit/Wayne State University, Detroit, MI; 2Henry Ford Hospital Detroit, Detroit, MI. (Tracking ID #117023)

BACKGROUND:  The influence of the pharmaceutical and medical device industry on scientific research and publications has been well documented. Conflict of interest disclosure in publications has been encouraged by the scientific community. It is unclear, however, what types of disclosure policies exist in major medical journals.

METHODS:  In July 2003, the scientific journals of the Abridged Index Medicus (AIM) were evaluated for the presence and characteristics of their conflict of interest disclosure policies. Information was primarily collected through individual journal websites, by examining the actual journals, and when necessary, by directly contacting the journal editorial staff.

RESULTS:  Of the 118 existing AIM journals, 108 (92%) had an identifiable disclosure policy. Ninety-two journals (78%) required disclosure of financial relationships related to the published subject matter, while only one journal asked that all financial relationships be disclosed, regardless of subject matter. Sixty-eight journals (58%) required the authors to disclose specific items that may pose a potential conflict of interest. Journals more commonly required authors to disclose consultant fees (53%) and stock ownership (47%), while reporting of honoraria (28%), and royalties (14%) were less frequent. However, nearly all the journals studied (96%) did not ask for an individual dollar amount of any of the disclosed items. Only 31 journals (26%) had policies requiring all financial relationships of the authors to be disclosed to the reader, while less than half of the journals (42%) had a policy specifically requiring disclosure information to be placed in the article.

CONCLUSION:  While many major medical journals have some type of disclosure policy, there is considerable variability in what authors are asked to disclose and in what journals actually reveal to the reader. This may suggest that authors should be provided with more uniform and clearly defined standards to inform readers of potential conflicts of interest in published research.

D-DIMER AND INFLAMMATORY BLOOD MARKERS PREDICT FUNCTIONAL DECLINE IN MEN AND WOMEN WITH AND WITHOUT PERIPHERAL ARTERIAL DISEASE.M.M. McDermott1; L. Ferrucci2; K. Liu1; C. Chan1; P. Greenland1; W.H. Pearce1; M.H. Criqui3; G.J. Martin1. 1Northwestern University's Feinberg School of Medicine, Chicago, IL; 2National Institutes on Aging, Bethesda, MD; 3University of California, San Diego, La Jolla, CA. (Tracking ID #115920)

BACKGROUND:  We determined whether higher baseline levels of D-dimer and inflammatory factors predict greater functional decline at 2-year follow-up in patients with and without peripheral arterial disease (PAD).

METHODS:  Participants were 552 men and women age 55 and older (336 with PAD). Measures of lower extremity functioning (six-minute walk distance and the summary performance score (SPS)) were assessed at baseline and annually for two years. The SPS, a comprehensive measure of functioning, combines performance for walking speed, standing balance, and time for five repeated chair rises into an ordinal score ranging from 0–12 (12 = best). Levels of D-dimer, high sensitivity C-reactive protein (hsCRP), serum amyloid A (SAA), and fibrinogen were obtained at baseline. Participants were categorized into one of three groups: Group 1: Three or more blood factors within the lowest tertile for each blood factor; Group 2: Fewer than three blood factor levels within either the highest or lowest tertiles, respectively; Group 3: Three or more blood factors within the highest tertile of each blood factor.

RESULTS:  The table shows results adjusting for age, sex, race, the ankle brachial index, comorbidities, body mass index, cigarette smoking, and patterns of missing data.

CONCLUSION:  Higher baseline levels of inflammatory markers and D-dimer are associated with significantly greater average annual decline in lower extremity functioning, as measured by the SPS. Future study is needed to determine whether therapies that reduce inflammation and D-dimer levels are associated with lesser declines in functioning.

Table 7. Adjusted associations between groups of inflammatory blood markers and D-dimer with functional decline at 2-year follow-up
 Group 1 (n = 69)Group 2 (n = 401)Group 3 (n = 82)Trend P value
Summary Performance Score–0.041/year–0.272/year–0.502/year.014
Six-minute walk–11.6 feet/year–36.0 feet/year–53.0 feet/year.306

DEFICITS IN INFORMATION TRANSFER FROM INPATIENT TO OUTPATIENT PHYSICIANS AT HOSPITAL DISCHARGE: A SYSTEMATIC REVIEW.S. Kripalani1; C. Phillips2; P. Basaviah3; M.V. Williams1; S.K. Saint4; D. Baker5. 1Emory University, Atlanta, GA; 2Brigham and Women's Hospital, Boston, MA; 3University of California, San Francisco, San Francisco, CA; 4Ann Arbor VA Medical Center, Ann Arbor, MI; 5Northwestern University, Chicago, IL. (Tracking ID #115953)

BACKGROUND:  Transfer of patient care between hospital-based and primary care physicians is increasingly common. Recent evidence suggests that ineffective communication between physicians at hospital discharge is a leading contributor to adverse events in the post-discharge period. We performed a systematic review to determine the type and frequency of reported problems with information transfer from inpatient to outpatient physician at hospital discharge.

METHODS:  We searched MEDLINE (1966-April 2003), the Cochrane Database of Systematic Reviews, abstracts from selected national meetings (2001–2003), and references of relevant articles. We included English language publications pertaining to the fields of adult primary care and hospital medicine (i.e., internal medicine, family medicine, and general practice).

RESULTS:  Of a total of 111 articles retrieved and reviewed independently by 2 investigators, 35 articles met our pre-specified inclusion criteria. Discharge information was communicated via structured discharge summaries (most common), informal discharge letters, telephone contact, fax, and electronic mail. The likelihood of the primary care physician receiving a discharge communication ranged from 75% to 99%. Letters generally arrived within 5 days, while discharge summaries took 2 to 4 weeks on average. Many discharged patients (16–53%) contacted the primary physician before receipt of any discharge information such as a preliminary letter or phone call, and 66–92% of patients saw the physician before a complete discharge summary had arrived. Outpatient management was sometimes affected by delays in receipt of information (10%) or lack of appropriate detail (14%). Discharge summaries often did not identify the responsible hospital physician (16–27%), main diagnosis (2–27%), other diagnoses (7–87%), results of abnormal investigations (20–75%), hospital course (28–55%), discharge medications (2–41%), follow-up plans (6–92%), patient or family counseling (90–97%), and tests pending at discharge (12%). Outpatient physicians were more satisfied with summaries that arrived within 1 week of discharge and contained the above information in a structured, legible format.

CONCLUSION:  Unfortunately, information transfer at hospital discharge is commonly delayed and incomplete. Hospitalists can enhance patient safety and outpatient physician satisfaction by providing structured discharge summaries which arrive within 1 week, and emphasize discharge medications, diagnoses, important test results, and follow-up needs.

DEFINING, NAVIGATING AND NEGOTIATING SUCCESS: WHAT MID-CAREER ROBERT WOOD JOHNSON CLINICAL SCHOLAR WOMEN SAY.A.L. Kalet1; K.E. Fletcher2; D. Ferdman1; N. Bickell3. 1New York University, New York, NY; 2Milwaukee VAMC, Milwaukee, WI; 3Mount Sinai School of Medicine, New York, NY. (Tracking ID #116054)

BACKGROUND:  Over the past decade, others have identified specific institutional strategies to address gender inequity in career advancement that benefit both women and men academic medicine faculty. Despite this knowledge, there remains a dearth of mid-career women role models. In a 1989 survey of women 1984–89 Clinical Scholar Program graduates (response: 21/39 (54%) we received practical advice about the challenges that women face in negotiating professional and personal lives in academic medicine In 2003, we conducted a survey to follow-up on these now mid-career women to determine how they crafted their careers and viewed personal and professional success.

METHODS:  We conducted a semi-structured survey of the 21 original respondents; obtained their Curricula Vitae (14) to quantify traditional academic productivity; and analyzed survey responses qualitatively. Specifically, we reviewed responses independently to identify themes then worked iteratively and collaboratively to refine the themes and compose a conceptual model which described the data. Inter-rater reliability for theme identification for a randomly selected 8 of the surveys, for rater pairs was 60%–100% (Kappa .35–1.0).

RESULTS:  16/21 (76%) responded. Average age was 48 (range: 45–56). Three were full professors, 10 associate professors, 2 left academic medicine and all of those who remained in academia were promoted. 11/16 women had children (average 2.4 children; range: 1–3), 3 chose to work part-time. Total grant funding obtained ranged from $120,000–2.2 million, number of first author papers 4–42. The conceptual model expands on the following 3 key themes: Defining, Navigating and Negotiating Success; Making Life Work; and Making Work Work. As a group the women who were satisfied with their careers since leaving fellowship, had clarity of values and goals, and a sense of control over their time. Those who were less satisfied with their careers described the personal cost of the struggle to balance their personal and professional lives with a few describing explicit institutional barriers to fulfillment of their potential and desires. Specific strategies employed and advice given were establishing oneself early, buying and creating flexibility with grants, and having help at home and work.

CONCLUSION:  These highly motivated women, selected and trained to achieve academic success, have done so by explicitly creating a personal vision of work and home balance, making values-driven choices, and learning to recognize institutional values and culture as potentially unfriendly.

DEPARTMENT CHAIRS’ PARTICIPATION IN THE INTERNAL MEDICINE CLERKSHIP—RESULTS OF A NATIONAL SURVEY.P.A. Hemmer1; E.J. Alper2; R.Y. Wong3. 1Uniformed Services University of the Health Sciences, Bethesda, MD; 2University of Massachusetts Medical School (Worcester), Worcester, MA; 3None Given, Loma Linda, CA. (Tracking ID #117016)

BACKGROUND:  Medical students value the participation of Department Chairs in their clincal education. We sought to characterize the roles and extent to which U.S. Internal Medicine Department Chairs are involved in the education of medical students.

METHODS:  In 2003, the Clerkship Directors in Internal Medicine (CDIM), the national organization of individuals responsible for teaching internal medicine to medical students, conducted a voluntary, confidential survey of its membership. Along with demographics, clerkship directors (CDs) were asked: if the Chair participated in the clerkship, and if so, in what capacities; how many hours per month the CD met with the Chair to discuss issues related to the clerkship; and if Expectations were discussed.

RESULTS:  Overall survey response rate was 71% (87/123); 61% (75/123) responded to questions concerning the Chair. 81% (61/75) responded the Chair taught in the IM clerkship: 63% Ward Attending; 13% Ambulatory Attending; 57% Teaching Attending; 11% Other (Lectures, student rounds, Professor rounds, morning report). 43% of Chairs performed at least two activities; 11% three activities; 3% four activities. Only 36% of CDs explicitly discussed published Expectations with their Chair. CDs spent 1.6 hours per month with the Chair discussing issues related to the clerkship, but 21% reported spending zero hours per month. Chairs who teach in the clerkship spend more time with the CDs each month compared to Chairs who do not teach (2.1 vs. .93 hours, P = .01, Mann-Whitney). There was no association between the likelihood of the Chair teaching and the CD demographics of age, sex, academic rank, years as CD, percent of time CD devotes to the clerkship, or discussing Expectations.

CONCLUSION:  In contrast to the assertion from the Association of American Medical Colleges that Department Chairs are “uninvolved in the clinical education of medical students”, our findings suggest IM Chairs are deeply involved in student education. Chairs who teach are more likely to spend time with the CD each month, a reflection of their interest in students and education. Nevertheless, Chairs should discuss Expectations with CDs and spend time each month with the CD.

DEPRESSION AND RELAPSE TO SMOKING IN PATIENTS HOSPITALIZED WITH ACUTE CARDIOVASCULAR DISEASE.A.N. Thorndike1; N.A. Rigotti1; S. Regan1; S. Swartz2; R. Pasternak1; Y. Chang1; K.M. Emmons3; D.E. Singer1. 1Massachusetts General Hospital, Boston, MA; 2Maine Medical Center, Portland, ME; 3Dana-Farber Cancer Institute, Boston, MA. (Tracking ID #116668)

BACKGROUND:  Depressed smokers are less likely to quit smoking than non-depressed smokers in the outpatient setting. Depression is common among cardiac patients and increases their morbidity and mortality. We hypothesized that depression would increase relapse to smoking immediately after hospitalization for acute cardiovascular disease (CVD).

METHODS:  We analyzed data from a randomized, double-blind, placebo-controlled trial of bupropion SR in adult smokers admitted to the hospital with CVD. We enrolled 248 smokers with acute myocardial infarction, unstable angina, or peripheral vascular disease. Patients started drug during hospitalization and continued for 12 weeks post-discharge. All smokers received intensive smoking counseling in the hospital and 5 follow up phone calls post-discharge. Patients were followed for one year. We used the Beck Depression Inventory (BDI) to assess depressive symptoms at baseline and considered a BDI score ≥16 to indicate current depression. We assessed abstinence at 3 and 12 months and verified with saliva cotinine. We used a chi-square test and logistic regression model to compare differences in quit rates. Time to relapse was determined by patient self-report of relapse and analyzed with survival analysis.

RESULTS:  At baseline, 53 (21%) of smokers scored ≥16 on the BDI. Depressed smokers were less likely to quit at 3 months (17% vs. 35%, P = .01) and 12 months (8% vs. 27%, P = .004). These differences remained significant when controlling for study arm, race, sex, age, cigarettes/day, and nicotine dependence. Time to relapse by baseline BDI is shown in the figure. For both depressed and non-depressed smokers, there was no difference in quit rates or days to relapse for drug vs. placebo group.

CONCLUSION:  Baseline depressed mood is a strong predictor of relapse to smoking after hospitalization for acute CVD. The antidepressant bupropion does not appear to improve cessation rates among depressed smokers. Hospital clinicians should be alert to depressive symptoms among smokers admitted with CVD because relapse occurs early, often before follow-up as an outpatient.

inline image

DEPRESSION AND TREADMILL EXERCISE CAPACITY IN OLDER ADULTS: THE HEART AND SOUL STUDY.B. Ruo1; J. Rumsfeld2; S. Pipkin1; M.A. Whooley1. 1University of California, San Francisco, San Francisco, CA; 2University of Colorado at Denver, Denver, CO. (Tracking ID #117450)

BACKGROUND:  Depressive symptoms are strongly associated with self-reported functional status in older patients, but the relation between depressive symptoms and objective measures of functional status is not known. We sought to examine the association between depressive symptoms and treadmill exercise capacity, an objective measure of functional status, in older patients with coronary artery disease.

METHODS:  In a cross-sectional study of 944 older adults (mean age 66) with stable coronary disease, we measured depressive symptoms using the Patient Health Questionnaire (PHQ) and exercise capacity using total metabolic equivalent tasks (METS) achieved on treadmill testing. We evaluated the association between depressive symptoms and exercise capacity using logistic regression, adjusted for measures of cardiac disease severity (left ventricular ejection fraction, ischemia by stress echocardiography) and other patient characteristics.

RESULTS:  A total of 172 participants (18%) had depressive symptoms (PHQ score ≥10). The proportion with poor exercise capacity (<5 METS) ranged from 21% in those with no or minimal depressive symptoms (PHQ score 0–3) to 34% in those with a depression score ≥10 (P <.001). In multivariable analyses adjusted for measures of cardiac disease severity and other patient characteristics, the presence of depressive symptoms (PHQ ≥10) remained independently associated with poor exercise capacity (OR 1.8, 95% CI 1.1–2.7; P = .01).

CONCLUSION:  In older patients with coronary disease, depressive symptoms are independently associated with poor treadmill exercise capacity, an objective measure of functional status. Depression should be considered in the differential diagnosis of poor functional status.

DEPRESSION SCREENING AND SELF-RATED HEALTH AMONG POSTMENOPAUSAL WOMEN WITH HEART DISEASE: THE HERS TRIAL.B. Ruo1; D. Bertenthal2; V.A. Bittner3; M.A. Hlatky4. 1University of California, San Francisco, San Francisco, CA; 2VA San Francisco Healthcare System, San Francisco, CA; 3University of Alabama at Birmingham, Birmingham, AL; 4Stanford University, Stanford, CA. (Tracking ID #102034)

BACKGROUND:  Self-rated health is a strong determinant of mortality. Prior studies have shown an association between depression and self-rated health among patients with coronary disease. However, it is unclear whether depression at baseline or the development of depressive symptoms over time affects these outcomes. The goal of this study was to clarify the association between depression and self-rated health using longitudinal data.

METHODS:  We performed a prospective cohort study of 2194 postmenopausal women with stable coronary disease using data from the Heart & Estrogen-Progestin Replacement Study (HERS) trial. The primary predictor was presence of depression based on the Burnam depression screen at baseline and annually during 4.1 years of follow-up. The outcome was decline in overall health from baseline to year 4 of follow-up. Overall health was self-reported as “excellent”, “very good”, “good”, “fair”, or “poor” and worsening by one or more categories from baseline to follow-up was considered a decline. We used multivariable logistic regression to examine the relationship between depression and change in overall health.

RESULTS:  The average age of participants was 66 ± 7 years. 11% had a depression at baseline and 9% had depression at follow-up. Among women who did not have depression at baseline, 131 (6%) developed depression during follow-up. After adjustment for age, comorbidities, baseline overall health, and interim events, baseline depression was associated with a decline in overall health (OR 2.1, 95% CI 1.2–3.6). Developing depression was associated with an approximately 4-fold increased odds of decline in overall health (OR 3.8, 95% CI 2.4–5.9). Individuals with an improvement in depressive symptoms were more likely to report improved overall health at follow-up.

CONCLUSION:  Women with depression or who subsequently developed depression were more likely to report a decline in overall health. Because depression is a modifiable risk factor, efforts to improve overall health in postmenopausal women with heart disease should include interval screening for and treatment of depression.

inline image

DESCRIBING HIGH UTILIZING PATIENTS WITH MEDICALLY UNEXPLAINED SYMPTOMS.F.C. Dwamena1; R.C. Smith1; J.S. Lyles1. 1Michigan State University, East Lansing, MI. (Tracking ID #117268)

BACKGROUND:  Medically Unexplained Symptoms (MUS) are common in primary care and result in significant functional disability for many patients, excessive healthcare utilization, and physician frustration. We sought to describe the personal experiences, characteristics, and perceptions of a random sample of high utilizing patients with MUS.

METHODS:  We randomly selected 19 patients from a group of MUS patients with an average of 14 clinic visits per year. We identified MUS patients from a large Health Maintenance Organization using a validated chart abstraction method. One of the authors (FCD) conducted open-ended semi structured interviews lasting from 60 to 90 minutes with each patient. We audiotaped and transcribed the interviews and were guided by grounded theory to identify and test major categories that characterize the patients’ experiences and behaviors. All three authors iteratively identified relevant themes and then described each patient in terms of these themes. We then categorized patients based on these themes.

RESULTS:  We identified 5 main themes – 1) Behavior/Action (pleasure in life, relationships and coping skills), 2) Mechanisms of illness (sense of self, personality structure, family history and patients’ explanatory models), 3) Physical symptoms (fear of physical disease, and evidence of secondary gain), 4) Diagnosis (biomedical, psychological and MUS), and 5) Emotionality (expression of emotion and psychological insight). Three distinct groups of patients emerged from the analysis. Group 1 patients (3 out of 19) all had evidence of pleasure in life, were involved in positive relationships and were other-directed; they easily expressed emotion, exhibited self-awareness, and showed evidence of recent psychological growth. Group 2 (4 out of 19) had some evidence of pleasurable life activities and good relationships; they were able to express emotion but were less self-aware and had less psychological insight than Group 1. The most distinguishing feature of Group 2 was an unrecognized, pathological fear of physical disease. Group 3 (12 out of 19) had little pleasure in life, had poor relationships and coping skills, were self-centered, had little, if any, psychological insight, tended to focus on their symptoms, and frequently had evidence of secondary gain from their illness. Obesity and clinical characteristics of disordered personality were also common in this group. Group 3 did not express fear of physical disease, even though they focused on their symptoms.

CONCLUSION:  Our qualitative study identified at least 3 groups of high utilizing patients with MUS differentiated by their ability to recognize and express emotion, presence of a pathological fear of physical disease, and the tendency to focus on physical symptoms. This may have significant implications for treating these types of patients, especially if quantitative measures of physical and mental dysfunction, psychosomatic symptoms and psychological disease such as depression and anxiety support our categorization.

DETERMINING THE FACTORS THAT ATTRACT AND DISSUADE CANADIAN MEDICAL STUDENTS FROM A CAREER IN INTERNAL MEDICINE.K. Tzanetos1; S. Straus1; L. Hicks1; K. Yeates2; N. Chahine-Malus1. 1University of Toronto, Toronto, Ontario; 2Queen's University, Kingston, Ontario. (Tracking ID #117332)

BACKGROUND:  If medical educators are to ensure an adequate number of future internal medicine specialists to care for the aging Canadian population they need to influence medical students to pursue internal medicine residency training. Towards this end, this study explored the factors that attract and dissuade Canadian medical students from choosing a career in internal medicine.

METHODS:  Senior medical choosing and rejecting internal medicine training were compared with respect to demographic characteristics and in terms of how much importance they placed on 34 non-demographic survey items (1 = not imoprtant, 5 = extremely important). The non-demographic items represented aspects of internal medicine training and a career in this field that are potential considerations to medical students making thier specialty decisions.The non-demographic survey items were grouped according to whether they tapped a similar underlying construct. A multivariable regression analysis was then carried out to model whether internal medicine is chosen, with the factors and demographics serving as explanatory variables.

RESULTS:  154 students from two (out of 16) Canadian universities responded to the survey. Students choosing internal medicine did not differ from those rejecting the specialty in terms of age (p-value 0.283), gender (p-value 0.877), total financial debt (p-value 0.577), involvement in a current relationship (p-value 0.205), or number of dependents (p-value 0.355). Having a career prior to medical school was negatively associated with choosing internal medicine (p-value 0.041). Those who chose internal medicine identified the following 5 factors as being most important to their decision: intellectual stimulation (mean 4.43), experiences with role models or mentors (mean 4.20), experiences with staff (mean 4.09), experiences with residents (mean 4.05), and experiences with patients (mean 3.82). Students who did not choose internal medicine did not place high importance on any of the items. In a regression analysis, the following factors were identified as being the most significant predictors of choosing internal medicine (descending magnitude of odds ratio): nature of the practice of internal medicine, geographic location of training program, past experiences in the area, and challenging academic work.

CONCLUSION:  This study suggests that in order to attract students to internal medicine training medical educators should emphasize positive medical school experiences, the intellectual challenge, and the specialty's ability to foster close patient-doctor relationships. Further research should focus on collecting qualitative data to explore and confirm these findings and to more clearly delineate the aspects of the specialty that act as deterents.

DEVELOPING A PATIENT-DERIVED INTERVENTION TO INCREASE ADHERENCE TO EXERCISE: A CONCEPTUAL MODEL.T.C. Collins1; P. Krueger1; T. Kroll2; B. Sharf2. 1Baylor College of Medicine, Houston, TX; 2Texas A&M University System, College Station, TX. (Tracking ID #115700)

BACKGROUND:  We explored patients’ perceptions of the role of exercise for the treatment of peripheral arterial disease (PAD) in order to design a clinical intervention that emphasizes patient-physician communication.

METHODS:  We conducted qualitative interviews (five focus groups and 13 one-on-one interviews) involving 36 patients with a documented ankle-brachial index of <.955 in at least one leg. Of the 36 patients interviewed, 20 were men and 16 were women. By race, we recruited 14 whites, 12 African Americans, and 10 Hispanics. Using Atlas TI software, we analyzed verbatim transcripts to identify significant dimensions of patients’ perceptions of doctor-patient communication that would increase their use of exercise (i.e., walking).

RESULTS:  Based on 30 emerging codes, we identified the main factors that summarized patients’ perceptions of the causes and outcomes of PAD, the importance of doctor-patient communication, the impact of PAD on their lives, and the factors that would lead to their use of exercise to treat PAD. From this information, our conceptual model focuses on an intervention plan that includes addressing the known causes of PAD, a patient's behavior prior to the diagnosis of PAD, patient's perceptions of PAD, and those factors that shape a patient's overall assessment of the role of exercise for PAD. We also include the impact of the physician's perception of PAD coupled with the patient's perception to shape the intervention—a partnership between the clinician and patient—that we posit is likely to lead to behavior change. The impact of the intervention plan is expected to vary with time as a function of how each patient's life changes daily due to multiple internal (e.g., perceived disease severity, co-existing illnesses) and external influences (e.g., exercise barriers, family system, detailed advice).

CONCLUSION:  Based on qualitative data analysis, we developed a patient derived model to influence the use of exercise for PAD. This model will be used to develop a communication primer for use within a randomized trial of unsupervised exercise therapy for patients with PAD. Ultimately, we anticipate the use of this primer within primer care clinics to increase the use of walking among patients with PAD.

DEVELOPING A STRUCTURED CLINICAL QUESTION RUBRIC (SCQR).M.A. Fischer1; M. Keough1; C. Stille1; J. Moldoff1; J.H. Gurwitz1. 1University of Massachusetts Medical School (Worcester), Worcester, MA. (Tracking ID #116245)

BACKGROUND:  Evidence Based Medicine (EBM) has been proposed as a means to improve quality of patient care and support lifelong learning by physicians. Its importance is reflected in several ACGME-required physician-training competencies. No existing research on teaching the practice of EBM has focused on the important first step of designing structured clinical questions (SCQ). Our objective was to develop, validate and evaluate a rubric (SCQR) for use in training residents to write SCQs.

METHODS:  We developed the SCQR after identifying characteristics from the published EBM literature that improve the quality of SCQs. We used two techniques to assess the validity of the SCQR. First, two trained faculty independently evaluated 91 questions generated by housestaff. Reliability was measured by coefficient alpha scores. Next, three faculty rated the SCQs for clinical usefulness, and we evaluated the correlation between SCQR and effectiveness scores. To evaluate the SCQR, we compared 168 SCQR scores for 84 questions relating to treatment and diagnosis generated by 26 housestaff before and after an EBM training seminar.

RESULTS:  We based the SCQR on four key categories: patient features, disease features, diagnostic test features and test characteristics or treatment and outcome features. Question evaluation resulted in a “total element” score of all pertinent elements included (1–29) and a “category” score (1–4). The SCQR's reliability values were 0.96 for total element score, and 0.90 for category score. Correlations between SCQR and effectiveness scores support scale validity. For instance, learners who included more key elements in their questions were more likely to receive a higher expert effectiveness rating. Preliminary implementation of the SCQR showed that after a three-hour training seminar, learners increased their inclusion of key elements in SCQs (mean elements included 6.01 vs. 4.58, P < .005), and included elements from more key categories in SCQs (mean categories included 3.2 vs. 2.7, P < .005).

CONCLUSION:  The SCQR may serve as a viable instrument to rate SCQs. It can be used as a tool for evaluation of resident learning as well as to help self-directed learning. Ongoing projects use the SCQR in interactive online modules to help learners assess and improve their own development of SCQs, and formulate and answer SCQs during clinical encounters. This work can help to support residencies as they implement changes to address required ACGME competencies of patient care, medical knowledge and practice-based learning and improvement. Long-term goals include evaluating if better SCQs result in better patient care and deeper learning.

DEVELOPING A WEB-BASED “OSTEOPOROSIS CARE TOOLBOX”: ELICITING INPUT FROM TARGET USERS.D.A. Levine1; K.G. Saag1; L.L. Casebeer1; C. Colon-Emeric2; K. Lyles2; R.M. Shewchuk1. 1University of Alabama at Birmingham, Birmingham, AL; 2Duke University, Durham, NC. (Tracking ID #115679)

BACKGROUND:  Barriers prevent osteoporosis diagnosis and treatment in nursing home residents. Successful interventions benefit from target recipient input during development. We sought to determine whether a highly structured process based on the Nominal Group Technique (NGT) would elicit useful and practical suggestions from nursing home directors of nursing (DONs, n = 9) concerning the content of a “toolbox” being developed as a component of a multi-modal intervention to improve the care of nursing home residents with osteoporosis.

METHODS:  We identified Alabama directors of nursing using data from the Alabama Quality Assurance Foundation. The first 15 respondents were recruited. 60% of these respondents participated. Information was elicited from participants in two phases. In the first phase, a conference call generated 42 possible elements for inclusion in the toolbox. Experienced moderators facilitated the discussion and utilized pre-formulated probes based on previous research identifying perceived barriers to osteoporosis care in nursing homes. These perceived barriers included reimbursement for bone mineral density testing and anti-osteoporotic therapies, facility staff osteoporosis knowledge and facility staffing availability for implementation of interventions. Then, a systematic distillation procedure combined substantive similar suggestions and discarded clearly idiosyncratic items resulting in the retention of 20 potential toolbox elements. In the second phase conducted by mail, the same DONs rated the 20 elements with respect to perceived practicality and feasibility for inclusion. Elements were grouped into tertiles based on the mean ratings of these attributes and cross-tabulated to obtain a feasibility versus practicality grid.

RESULTS:  All DONs (n = 9) participants completed both phases. The most feasible, most practical elements included: 1) Assessment tools for osteoporosis, 2) Fall and fracture prevention information, 3) Osteoporosis treatment protocols, 4) Pain management protocols, 5) Reimbursement information for treatments, and 6) Osteoporosis educational programs for nurses and clinical nurse assistants. These elements were incorporated into the osteoporosis intervention toolbox in participant-recommended formats. The least feasible, least practical elements included: 1) Sources for authoritative osteoporosis treatment information, 2) Knowledge of fall prevention program implementation, 3) Information on better assistance devices for osteoporosis patients and 4) Osteoporosis educational conference call for directors of nursing. The described process achieved successful consensus among the participants.

CONCLUSION:  A modified nominal group technique provided useful information from health care professionals for the purpose of barrier identification and intervention development. The technique proved efficient, facile and well-received by participants.

DEVELOPMENT AND INITIAL TESTING OF A COMPUTER-BASED PATIENT DECISION AID TO PROMOTE COLORECTAL CANCER SCREENING FOR PRIMARY CARE PRACTICE.J.A. Kim1; M. Pignone1; A. Whitney1; C. Lewis1. 1University of North Carolina Division of General Medicine, Chapel Hill, NC. (Tracking ID #115174)

BACKGROUND:  We have previously developed an effective videotape decision aid for CRC screening, but it did not include colonoscopy screening and could not be tailored to patients’ information needs. To address these limitations, we developed and tested a novel computer-based patient decision aid for use in primary care practice.

METHODS:  We recruited a convenience sample of adults 50–75 years of age who were attending a scheduled appointment at our university internal medicine clinic. Eligible patients had no personal or family history of colon cancer. Participants viewed a computer-based decision aid based on “stages of change” theory that included an introduction to colorectal cancer screening, video clips explaining different screening modalities (FOBT, sigmoidoscopy, colonoscopy, or barium enema), and comparative information about the tests. A physician and several patients provided narration. Subjects completed surveys before and after watching the decision aid that included intent to ask one's provider for screening, measured on a 4-point Likert scale (4 = very likely to ask, 1 = very unlikely to ask); and stage of readiness to be screened (green = ready to be tested, yellow = need more information, and red = did not want to be tested). We performed chart audits to determine the proportion of patients with tests ordered and completed.

RESULTS:  To date, 73 patients have participated. Mean age is 60 years; 69% are White, 21% African-American, and 64% have greater than a high school education. 45% were screened for colon cancer in the past and 62% had previously discussed screening with their provider; 90% preferred to play a major role in deciding how to be screened. At baseline, the mean intent to ask for screening was 2.8. After the decision aid, the mean intent increased to 3.2 (P < .0001, paired t test). 95% said the information increased their knowledge about colon cancer and 84% thought the information helped them decide whether to be screened. After viewing the decision aid, 62% were ready to be screened, 17% wanted more information but were considering screening, and 16% did not want to be screened. 38% of participants had a CRC screening test ordered and 29% of patients have completed their tests to date, with an additional 7% who are awaiting endoscopy.

CONCLUSION:  A computer-based colorectal cancer decision aid improved patients’ intent to ask for screening and subjectively improved patient knowledge about screening. Almost 2/3 of patients want to be screened, but actual screening rates are lower, suggesting the need for additional office systems support, such as standing orders.

DEVELOPMENT AND VALIDATION OF AN INDEX TO PREDICT ADL DEPENDENCE IN COMMUNITY LIVING ELDERS.K.E. Covinsky1; J.F. Hilton1; L. Karla1; R.A. Dudley1. 1University of California, San Francisco, San Francisco, CA. (Tracking ID #117155)

BACKGROUND:  Maintaining independence in daily functioning is an important health outcome in older adults. A key measure of functional independence in elders is the ability to do activities of daily living (ADL) without the assistance of another person. However, few prognostic indices have been developed that stratify elders into groups at variable risk for developing ADL dependence.

METHODS:  We studied subjects enrolled in AHEAD, a nationally representative cohort of elders over age 70. We included 5239 subjects (mean age 77) reporting that they could do each of 5 ADL (bathing, dressing, toileting, transferring, eating), without the assistance of another person at baseline. Our outcome was the need for help (dependence) with at least one ADL at 2 years. Predictor variables encompassed several domains: demographic characteristics, comorbid conditions, functional status, cognitive status, and general health indicators. After dividing subjects into development (n = 3,245) and validation cohorts (n = 1,994), we used logistic regression to determine independent predictors of 2 year ADL dependence in the development cohort. We then created a risk index, which we tested in the validation sample.

RESULTS:  The 8 independent predictors of 2-year ADL dependence were age over 80, diabetes, difficulty walking 3 blocks, difficulty bathing or dressing, need for help with personal finances, difficulty lifting 10 pounds, inability to name the Vice-president, history of falling, and low BMI. We created a risk index by assigning 1 point to each risk factor. In the development sample, rates of 2-year ADL dependence in subjects with 0, 1, 2, 3, 4, and 5 or more risk factors were 1.3%, 2.8%, 3.8%, 10.1%, 21.6%, and 33.3% respectively (P < .001, roc area = .79). In the validation sample, the rates were 0.7%, 4.3%, 8.7%, 11.3%, 17.8%, and 39.5% (P < .001, roc area = .77).

CONCLUSION:  Using data available from patient reports, we validated a risk index that accurately distinguished between subjects at variable risk of ADL dependence.

DEVELOPMENT AND VALIDATION OF AN INDEX TO PREDICT MORTALITY IN FRAIL ELDERS.E. Carey1; L. Lui1; L. Walter1; C. Eng1; K.E. Covinsky1. 1UCSF, San Francisco, CA. (Tracking ID #117365)

BACKGROUND:  Most prognostic indices for older people focus on specific diseases or hospitalized elders. However, the majority of elder deaths are associated with multiple chronic illnesses and frailty, the prognosis for which is notoriously difficult to determine. Our goal was to create a predictive index for mortality in frail older adults.

METHODS:  We developed and validated a predictive index for mortality using a cohort of frail elders who receive care from the Program of All-Inclusive Care of the Elderly (PACE), a program which enables frail, nursing-home eligible elders to continue living at home. Participants were enrolled in PACE between 1988 and 1996. We developed the index in 2232 participants from 6 sites in the western U.S. (mean age 79, 68% female, 40% white, overall 1- and 3-year mortality 13% and 55% respectively), and validated it in 1667 participants from 5 sites in the eastern and midwestern U.S. (mean age 79, 76% female, 65% white, overall 1- and 3-year mortality 13% and 50% respectively). Participants were evaluated by the PACE multidisciplinary team at the time of PACE enrollment, and a full geriatric assessment was completed, including assessment of comorbid illnesses, personal care needs, caregiver support, and functional status.

RESULTS:  In the development cohort, 10 independent predictors of mortality were identified and weighted, using Cox regression, to create a risk score: male sex, 2 points (pts); age [75–84, 2 pts; >84, 3 pts]; dependence (dep) in toileting, 1 pt; dep in dressing [partial dep, 1 pt; complete dep, 2 pts]; bowel incontinence, 1 pt; malignant neoplasm, 3 pts; diabetes mellitus, 1 pt; CHF, 3 pts; COPD, 1 pt; and renal failure or insufficiency, 3 pts. We calculated each subject's risk score by adding the points of each independent risk factor present. The 1- and 3-year mortality rates by risk score and C-statistic (ROC area) for both the development and validation cohorts are presented in the table.

CONCLUSION:  Our prognostic index provides a simple and accurate method of stratifying frail elders into groups at varying risk of mortality.

Table 8. 1- and 3-Year Mortality Rates by Risk Score in the Development and Validation Cohorts
Risk Score1-Yr Mortality3-Yr Mortality
Dev. CohortVal. CohortDev. CohortVal. Cohort
0–3 points 36/650 (6%) 30/477 (6%) 95/650 (18%) 82/477 (21%)
4–6 points125/976 (13%) 73/707 (10%)296/976 (37%)198/707 (34%)
>6 points129/606 (21%)119/483 (25%)286/606 (57%)238/483 (58%)

DEVELOPMENT OF A MEASURE OF A PRACTITIONER's ORIENTATION TOWARD INTEGRATIVE MEDICINE.A.F. Hsiao1; R.D. Hays1; G.W. Ryan2; R.M. Andersen1; M.L. Hardy1; D.L. Diehl3; N.S. Wenger1. 1UCLA, Los Angeles, CA; 2RAND, Santa Monica, CA; 3NYU, New York, NY. (Tracking ID #115677)

BACKGROUND:  Patients often turn to complementary and alternative medicine (CAM) and use it concurrently with conventional medicine to treat illnesses and promote wellness. Yet, some clinicians are not open to combinations of treatment paradigms. Because integration of CAM with conventional medicine may have outcome and economic implications for care, we developed a survey instrument to assess a practitioner's orientation toward integrative medicine.

METHODS:  We used semi-structured interviews, cognitive interviews, and multitrait scaling analysis to develop and evaluate a 32-item survey to measure a practitioner's orientation toward integrative medicine (IM-32). The IM-32 assesses 6 domains of a practitioner's orientation toward integrative medicine: openness, readiness to refer, learning from alternate paradigms, spiritual orientation, patient-centered care, and safety of integration. We mailed the survey to a convenience sample of 294 acupuncturists, chiropractors, internists/family practitioners, and physician acupuncturists in academic and community settings in California.

RESULTS:  Two hundred and two participants (75%) completed the survey. The mean age was 47 years; 68% were men. Seventy percent were white and 23% were Asian Americans. The 32-item scale had a mean score of 64.7 (0–100 possible range) and alpha internal consistency reliability = .90. All sub-scales except one had reliability of 0.70 or above (spiritual orientation, alpha = .66) and item discrimination successes of 80% or above (readiness to refer, item discrimination = 74%). Construct validity was supported by the association of the IM-32 total scale score with hypothesized constructs: physician acupuncturists vs. physicians (71.2 vs. 49.7, P = .001), dual-trained practitioners vs. single-trained practitioners (71.3 vs. 61.8, P = .001), and upper versus lower tertile of practitioners’ self-perceived “integrativeness” (73.1 vs. 67.4, P = .01).

CONCLUSION:  This study of conventional and CAM practitioners provides initial support for the reliability and validity of the IM-32. The IM-32 could be employed in measuring the relationship of integration with patient satisfaction, health outcomes, and cost-effectiveness of care.

DIABETES CASE MANAGEMENT: EFFECTS ON GLYCEMIC CONTROL.W.T. Shimeall1; E.S. Holmboe2; R. Escobedo3; B.M. Sylvia1; C. Clark4; P. Kelley1. 1Uniformed Services University of the Health Sciences, Bethesda, MD; 2Yale University, New Haven, CT; 3Naval Hospital Jacksonville, Jacksonville, FL; 4National Naval Medical Center, Bethesda, MD. (Tracking ID #117546)

BACKGROUND:  Background: Diabetes is one of the most prevalent and costly chronic diseases in the U.S. The average HbA1c among diabetics is 9.2%, well above American Diabetes Association (ADA) goal of <7.0%. Eliminating barriers to glycemic control among individuals living with diabetes and their primary care providers remains a significant challenge. The purpose of this study is to investigate the effectiveness of a patient-centered hybrid case management intervention compared to usual diabetes care in a large military academic hospital.

METHODS:  Methods: This is a one year randomized, controlled, longitudinal study. The hybrid case management program consists of an initial separate diabetes clinical assessment using multiple providers with subsequent nurse case management. The initial multidisciplinary visit is followed by ongoing contact either in person or via telephone on a regular basis with the nurse case manager who coordinated care with the patient's usual primary care physician. The program targeted higher risk diabetic patients with HbA1c > 8%. The primary outcomes for the study were change in HbA1c at 12 months and compliance with ADA recommended diabetes care process measures during this time, including quarterly HbA1c measurement when above target value, pneumococcal and influenza vaccination, microalbuminuria screening, and monofilament foot exams. Qualitative interviews with 20 participants were conducted after 12 months to assess perceptions of care. The constant comparative method of analysis was used to review interviews.

RESULTS:  Results: 81 participants were enrolled, with 43 randomized to case management and 38 to usual care. The groups are statistically similar by age, gender, co-morbidities and initial HbA1c. Four participants in the case management group and three in usual care withdrew early in the course of the study. The study is ongoing and 35/39 case managed and 35/35 usual care participants have completed 12 months since randomization. HbA1c screening compliance (>3 tests completed in study year to approximate quarterly screening) was significantly greater in the case management group (97.4% vs. 37.1%, P < .001). The case management group's mean HbA1c improved from 10.27 to 7.95 (P < .001), but the HbA1c screening noncompliance in the usual care group prevented a statistical comparison of final HbA1c changes between the groups. There were also clinically and statistically improved compliance measures in the case management group for pneumococcal vaccination (74.4% vs. 44.1%, P < .008), influenza vaccination (69.2% vs. 47.1%, P < .05), monofilament foot exams (71.8% vs. 45.5%, P = .034), and microalbuminuria screening (89.5% vs. 71.4%, P = .05). Ten participants from each group underwent the qualitative interview. The most common themes regarding the positive impact of the case management program from the patient perspective that emerged were 1) a greater sense of mastery of self-care, 2) the importance of having a nurse function like a coach, 3) a personal sense of feeling better, and 4) the importance of a partnership with providers to facilitate self-care.

CONCLUSION:  Conclusion: In a large academic military hospital, a primary care-based nurse case management program following multidisciplinary assessment is effective in improving glycemic control and diabetes care process outcomes. This study extends earlier work by demonstrating the positive impact on participants’ perceptions of care, well-being and engagement in the self-management of their chronic disease.

DIABETIC FOOT CARE IMPROVES BY PARTICIPATION IN ACGME RECOMMENDED PERFORMANCE IMPROVEMENT PROCESS.S. Kraemer1; L. Staton1; S.R. Patel1; C. Estrada1. 1East Carolina University, Greenville, NC. (Tracking ID #116623)

BACKGROUND:  A Performance Improvement (PI) Process is recommended by the AGGME. The PI activities must relate to the core competencies, involve residents and faculty, and produce measurable improvements in patient care or residency education. Although PI projects are utilized in health care, no data is available documenting the effectiveness of this ACGME requirement. We developed a PI project examining the impact of 2 sequential retrospective electronic medical record chart reviews using a Diabetes Quality Improvement Project (DQIP) guidelines-based questionnaire on diabetic foot care.

METHODS:  Medical residents reviewed the electronic medical record twice to identify Internal Medicine continuity clinic visits of patients with diabetes mellitus (DM). Audit #1 reviewed patient encounters for the academic year 2002–2003 and Audit #2 July–Sept 2003. Residents reviewed 2–5 charts during each audit. The questionnaire required yes or no answers of diabetic foot history and review of systems. We also utilized DQIP guideline-based questions to review the foot examination (skin, vascular, neurologic) and if any interventions were performed.

RESULTS:  Patients had an average of 3.9 visits during audit 1 (n = 105) and 3.4 visits for Audit #2 (n = 142). Any mention of feet in the review of systems was similar in both audits (P < .20). Audit #2 showed improved documentation of monofilament or fork test (23% vs. 13%; P < .08), pedal pulses (73% vs. 45%; P < .01), and skin exam (72% vs. 51%; P < .001). Audit #2 also showed improved documentation of a complete foot exam (18% vs. 6%; P < .004). No differences were found in the prevalence of foot abnormalities (P = .18) or frequency of interventions (P = .76)

CONCLUSION:  The PI process improved documentation and performance of the DM foot exam by Internal Medicine residents. This study supports the ACGME recommendations for PI activities by internal medicine residency programs. Additionally the educational impact of this chart audit is highlighted by the absence of a coinciding didactic or formal educational instruction about diabetic foot care. The PI process we used incorporated four ACGME core competencies: effective patient care, application of medical knowledge to patient care, practice based learning and system-based practice. It is unknown if ACGME requirements for the PI process will lead to improved clinical outcomes.

DIAGNOSTIC TESTS FOR UROGENTIAL CHLAMYDIAL AND GONOCOCCAL INFECTIONS USING NON-INVASIVE SAMPLES: A META-ANALYSIS.R.L. Cook1; S.L. Hutchison1; L. Ostergaard2; R.S. Braithwaite1; R.B. Ness1. 1University of Pittsburgh, Pittsburgh, PA; 2Aarhus University Hospital, Aarhus. (Tracking ID #117513)

BACKGROUND:  Nucleic acid amplification tests (NAATs) represent a new generation of test assays for chlamydial and gonococcal infections. They can be used on non-invasive samples such as urine and self-collected vaginal swabs. Yet, the majority of clinicians still obtain test samples for genital chlamydial and gonococcal infections using invasive methods such as swabs or speculums. The specific objectives of this study were to determine overall estimates of diagnostic accuracy for non-invasive tests for chlamydial and gonococcal infections, and to determine whether diagnostic accuracy varies according to test characteristics, population characteristics, or study quality.

METHODS:  Articles were identified through a MEDLINE search, references of identified articles, and hand-searching of relevant journals. Articles were included if they presented sensitivity and specificity results for a non-invasive, commercially available nucleic-acid amplification test (NAAT); used an appropriate reference standard for comparison; and presented data separately for men and women. From 46 eligible studies, two authors independently abstracted data on type of test assay, test sensitivity and specificity, reference standard, population characteristics, funding source, and quality score (using a 7-item checklist). Summary estimates for males and females were calculated separately for chlamydial and gonococcal infections and were stratified for various patient and test-related subgroups.

RESULTS:  For Chlamydia trachomatis, the overall sensitivity and specificity for women was 88.1 (95% CI 85.7–90.5) and 99.8 (95% CI 99.7–100) and for men was 89.3 (95% CI 86.9–91.8) and 99.4 (95% CI 99.1–99.7). For Neisseria gonorrhoeae, the overall sensitivity and specificity for women was 85.5 (95% CI 76.3–94.8) and 99.7 (95% CI 99.2–1.0), and for men was 91.4 (95% CI 83.4–99.5) and 99.4 (95% CI 98.8–1.0). There were minimal differences in sensitivity or specificity according to differences in population characteristics (e.g. presence of symptoms) or test characteristics (e.g. method of sample collection). The sensitivity of chlamydial assays tended to be higher in studies with better quality scores.

CONCLUSION:  NAAT tests on non-invasive samples have excellent sensitivity and specificity, and the results are nearly identical to those obtained through more “invasive” sample collection directly from the cervix or urethra. Non-invasive test collection methods are preferred by patients, use fewer resources, and may increase acceptability of screening. Non-invasive testing should become the standard of care.

DIFFERENCES IN MEASURED QUALITY PERFORMANCE UTILIZING PATIENT PANEL COMPARED TO VISIT DATA.G.P. Barnas1; J. Kulp2; L. Voigt2; G. Schectman1. 1Medical College of Wisconsin, Milwaukee, WI; 2Milwaukee VA Medical Center, Milwaukee, WI. (Tracking ID #115801)

BACKGROUND:  Physician profiling is increasingly being utilized by insurers, managed care organizations, corporations, and patients to help make decisions regarding provider selection or reimbursement. The use of visit-based (encounter or claims) data for performance monitoring has been the historical method used to assess clinical outcome measures. However, compared to sampling a provider's entire panel of patients independent of the patients’ most recent clinic visit, the visit-based method is less likely to include patients of a provider, who for various reasons, are seen less frequently. In this study we assessed the impact on measured clinical outcomes utilizing visit-based data compared to enrollment (patient panel) data.

METHODS:  All patients with outpatient clinic visits to a Primary Care provider (PCP) at the Milwaukee VAMC Apr–Jun 2003 were electronically sampled and compared to all patients that were assigned to a PCP during the same interval. Clinical outcome data for seven performance measures were assessed utilizing computerized clinical reminder data extracts.

RESULTS:  25,910 patients were assigned to a PCP and 14,401 patients had visits to a PCP within the study period. Measures of blood pressure (BP) control (<140/90) in hypertensive patients and diabetic (DM) control (HgbA1c < 9) were similar in both samples but the other measures such as diabetic BP control (DM-BP), diabetic foot exams (DM-Foot), colon cancer screening, and pneumococcal and influenza immunizations were all reported to have better outcomes in the visit group with the difference ranging from 1.6% to 19%.

CONCLUSION:  Use of visit-based denominator data for quality performance measures may overestimate actual compliance compared with measures across all enrolled patients. Patient sampling methods selected for quality monitoring may affect the interpretation of performance data in the ambulatory clinic setting.

PCP__BPHgbA1cDM-BPDM-FootColon CaPneumovaxFluvax
DIFF: 0.4% 0.1% 1.9% 1.6% 6.6% 5.8%19.1%

DIFFERENTIATING WEAK AND STRONG EVIDENCE IN DECISION MAKING.C.S. Seibert1; L.J. Zakowski1; S. Vaneyck1; C.L. Gjerde1. 1University of Wisconsin-Madison, Madison, WI. (Tracking ID #115409)

BACKGROUND:  Medical students need to learn to assess the strength of medical evidence before they apply it to make clinical decisions. The purpose of this study was to assess the ability of second-year medical students to differentiate between weak and strong evidence before and after instruction in EBM.

METHODS:  Students were instructed to read a clinical scenario and make a clinical decision about initiating a treatment or ordering a test. Then, they were given either strong or weak evidence related to the clinical scenario and were asked to reconsider their clinical decision based upon the evidence provided. Students were blinded as to whether they receive weak or strong evidence. Each student completed two clinical scenarios (one with weak, one with strong evidence) before and after EBM instruction. Comparisons were made between the percentage of students who made correct clinical decisions before and after EBM instruction.

RESULTS:  Before EBM instruction, decision-making data was obtained from 87 of 143 students and after EBM instruction, data was obtained from 121 of 143 students. Before EBM instruction, initial responses to clinical scenarios were 47% correct. After reading strong evidence, there was a significant increase in correct responses to 85% correct (P = .000). After reading weak evidence, there was a significant increase in correct responses to 81% correct (P = .000). There was no difference in the correctness of final answers achieved after considering weak versus strong evidence (chi-square = 0.3; P = .56.) After EBM instruction, initial responses to clinical scenarios were 50% correct. After reading strong evidence, there was a significant increase in correct responses to 88% correct (P = .000). After reading weak evidence, there was a significant increase in correct responses to 75% correct (P = .000). A significant difference was achieved in the correctness of final answers after considering weak versus strong evidence (chi-square = 7.0; P = .008).

CONCLUSION:  While initial decisions were at the guess level (~50%), most students made the correct clinical decision after considering strong or weak evidence. After EBM instruction, students were better able to discriminate between strong and weak evidence by making significantly more correct decisions with strong evidence.

DIFFERING INTERPRETATIONS OF PRIMARY CARE RESIDENTS’ PERSONAL STATEMENTS.D. Zipkin1; R.B. Baron2; T.E. Baudendistel1; R. Gonzales2. 1California Pacific Medical Center, San Francisco, CA; 2University of California, San Francisco, San Francisco, CA. (Tracking ID #115779)

BACKGROUND:  Primary care internal medicine residency training programs aim to select applicants committed to careers in primary care. Indeed, to maintain Health Resources and Services Administration support, a program must demonstrate that a high proportion of graduates pursue careers in primary care. To gauge their applicants’ commitment to primary care, selection committees rely on information from multiple sources, particularly applicants’ personal statements. We examined the level of agreement between experienced selection committee members in their interpretations of the personal statements of residents accepted into a well-established primary care internal medicine residency.

METHODS:  A convenience sample of forty primary care residency application personal statements submitted between the years 1992 and 1999 were reviewed by two raters who were blinded to the other's assessments. The raters scored the statements for the presence of an interest in or acknowledgement of the following core primary care concepts, which had been established a priori by a consensus panel of five faculty members: continuity of care, primary care experience, access to care, preventative care, psychosocial medicine, international health, and clinical research. In addition, the raters graded the presence or absence of references to sub-specialty interests as potential negative predictors of a career in primary care. Interrater reliability for identifying these concepts was measured using the Kappa statistic.

RESULTS:  Core concepts least frequently cited included sub-specialty interest (13%) and international health (21%), and concepts most frequently cited included access to care (76%) and psychosocial medicine (76%). The average number of concepts per statement was 3.7 (SD = 1.4). Reviewers agreed modestly in their ratings of access to care (kappa coefficient 0.52 [95% CI 0.22–0.83]), prevention (kappa 0.59 [0.34–0.84]), and continuity of care (kappa 0.60 [0.35–0.85]). Better correlation occurred with applicant's self-reported primary care experience (kappa 0.74 [0.52–0.95]), and the strongest correlation was seen with specialty interest (kappa 0.77 [0.47–1.07]). The weakest correlation was seen with interest in psychosocial medicine (kappa 0.29 [0.01–0.57]).

CONCLUSION:  Experienced selection committee physicians reading primary care applicants’ personal statements display a high degree of variability in identifying the presence of core primary care topics. How personal statements are evaluated in the resident selection process, and the role of the statement as a tool to predict an applicants’ commitment to primary care should be further defined.

DISAGREEMENT IN PAIN ASESSMENT BETWEEN PATIENTS AND PHYSICIANS. DOES RACE PLAY A ROLE?L. Staton1; M. Panda2; I.A. Chen3; C. Sam4. 1East Carolina University, Greenville, NC; 2University of Tennessee at Chattanooga, Chattanooga, TN; 3Eastern Virginia Medical School, Norfolk, VA; 4University of North Carolina, Chapel Hill, NC. (Tracking ID #116629)

BACKGROUND:  Physicians often understimate patients’ pain. Agreement in pain assessment has been shown to be associated with improved quality of life and patient satisfaction. We explored patient factors associated with disagreement in pain assessment between physicians and their patients.

METHODS:  The Patient and Physician Perception of Pain Study (4P) is a 12 center, cross sectional evaluation of chronic non-malignant pain patients and their physicians in a primary care setting. We recruited patients with greater than 3 months of chronic non-malignant pain. Patients completed a 160 item survey which included demographic factors, and attitudes regarding quality of life, self-efficacy, disability and satisfaction. To measure pain intensity, patients completed a 11-point numeric rating scale for which pain scores range from 0 (no pain) to 10 (unbearable pain). Physicians were given access to patients’ self rating of pain and was asked to rate patients’ pain on the same scale after the visit. Discordance was defined as underestimation of pain by 2 or more points, which is clinically relevant. Race was defined as black vs. other. We compiled descriptive data and performed bivariate analyses examining factors associated with discordance. We then performed logistic regression analysis to control for possible confounders (age, gender, education level, insurance status, degree of pain, marital status, depression, narcotic use and comorbid illnesses).

RESULTS:  Of 601 patients evaluated, 463 (77%) patients completed the survey. Forty-eight percent were white, 39% were black and 67% were female. Mean age was 53 years, 34% were married, 35% admitted to being depressed and 36% reported using narcotics. Underestimation of pain level occurred 39% of the time and was significantly associated with black race (P = .005). After combining race in a regression model with demographics and other possible confounding variables, black patients remained signficantly more likely to have their pain underestimated by physicians, OR 1.9 (95% CI 1.2, 3.0).

CONCLUSION:  We found that race is significantly related to discordance in pain assessment. This association may suggest other biases and inadequacies in pain treatment in African Americans. Further studies could address if efforts to improve awareness of disparities in health care and to enhance communication are possible ways to improve congruence in pain assessment.

DISCIPLINARY ACTION AGAINST PHYSICIANS: CHARACTERISTICS AND PREDICTORS.A.A. Khaliq1; L. Narine2; R.A. Smego3. 1University of Oklahoma, Oklahoma City, OK; 2University of North Carolina at Charlotte, Charlotte, NC; 3University of North Dakota School of Medicine & Health Sciences, Fargo, ND. (Tracking ID #116803)

BACKGROUND:  We sought to report the type, frequency, and severity of disciplinary actions against physicians (MDs), and to determine physician characteristics and disciplinary predictors.

METHODS:  Descriptive and predictive analysis of publicly available data maintained by the Oklahoma State Board of Medical Licensure & Supervision (OSBMLS).

RESULTS:  Among 14,314 currently or previously licensed MDs, 396 (2.8%) had been disciplined (515 disciplinary actions). Of these, 252 had one action, 95 had two actions, and 16 had three or more actions taken against them over time. For the year 2001, 543 complaints lodged by 371 complainants were registered by OSBMLS. The most frequent complaints against MDs involved quality of care issues (24.8%), incompetence (18.3%), negligence (16.7%), medication/prescription violations (19.6%), billing issues (9.4%), inadequate records (8.4%), fraud (6.7%), sexual misconduct (4.6%), substance abuse (3.8%), criminal activity (3.0%), application fraud (0.3%), and other (31.8%). Complaints against MDs originated from the public (66%), other MDs (5.3%), self-reporting (4.8%), and staff (4.0%). In contrast to 1.8% of MDs in all other primary specialties combined (n= 8,705), 4.6% of primary care MDs (includes Internal Medicine, Family Medicine, and General Practice) (n= 4,163), 4.0% of obstetricians/gynecologists (n= 602), and 3.5% of psychiatrists (n= 687) had been disciplined (P < .001). Logistic regression modeling revealed that older age (for those > 40 but < 55 yrs, OR = 6.03, 95% CI = 3.07–11.83; P < .001; for those > 55 yrs, OR = 9.37, 95% CI = 4.80–18.28; P < .001), lack of board certification (OR = 2.82, 95% CI 2.06–3.88; P < .001), and male gender (OR = 1.76, 95% CI 1.05–2.94; P < .001) were strong predictors of being disciplined. Race and foreign medical school training were not risk factors. Disciplinary actions against 189 actively licensed MDs included probation (52%), suspension (24%), reprimand (12%), and revocation (5%).

CONCLUSION:  In order to optimize physician behavior, medical schools and residency training programs must continue to emphasize both patient care and medical professionalism as critical core competencies.


BACKGROUND:  In response to public scrutiny of its conflict-of-interest procedures, the FDA drafted a guidance document, effective January 2002, requiring more explicit disclosure for Advisory Committee (AC) members and invited consultants for meetings in the FDA's Center for Drug Evaluation and Research (CDER) that considered specific products. The objectives of this study were to characterize disclosed conflicts, examine the extent of compliance with the guidance, and examine the guidance's impact.

METHODS:  Data were extracted from the committee rosters, meeting agendas, and transcripts available on the FDA website for product-specific CDER meetings from January 2001 to June 2003. Details of the reported financial interests for all AC members and CDER consultants covered by the guidance were recorded and analyzed using Access.

RESULTS:  A total of 81 meetings involving 1220 persons were covered by the guidance and hence included in the study. In 69 meetings (81%), at least one attendee disclosed a conflict. At the attendee level, 269 (22%) disclosed a conflict. The most common specified conflicts were consulting arrangements (41% of conflicts), investments (28%), and lecture honoraria and grants (13% each). Eighteen percent of these consulting arrangements involved over $10,000, 28% of individuals collected over $10,000 in lecture honoraria and 11% of grants exceeded $300,000, all generally within the last 12 months. Disclosure of the specific monetary value of a conflict increased from 15% before the guidance went into effect to 96% after, while disclosure of whether the conflict was with the sponsoring company or a competitor went from 27% to 96%.

CONCLUSION:  Conflicts of interest at CDER AC meetings are common and often of considerable monetary value. More uniform disclosure of the conflict value and competitor status has occurred as a result of the FDA guidance. The FDA should ensure 100% compliance with the guidance as it is written and also expand the guidance to include non-product meetings and longer time periods.

DISCONNECT BETWEEN PERCEIVED AND TRUE RISK FOR CARDIAC DISEASE IN URBAN MINORITIES.K.B. Desalvo1; B. Pedersen2; M. Kleinpeter2; J.W. Peabody3. 1Tulane University, New Orleans, LA; 2Tulane School of Medicine, New Orleans, LA; 3University of California, San Francisco, San Francisco, CA. (Tracking ID #116901)

BACKGROUND:  Socially and economically disadvantaged populations are disproportionately affected by cardiac risk factors (CRF) and heart disease. Because many of these risk factors can be modified, we need to understand patients perceptions of CRF. We studied the actual prevalence of CRF and perceptions of heart disease risk in an urban, disadvantaged clinic population.

METHODS:  Trained interviewers surveyed 199 patients selected to participate in a future randomized, controlled educational trial. The survey instrument included questions about actual and self-perceived risk of heart disease, as well as novel CRF such as stress and socioeconomic status.

RESULTS:  Most participants were female (92%) and black (89%). The mean age was 56 years [SD  9.9]. 39% of participants lacked a high school diploma, 64% earned less than $750/month, and 57% were unemployed. Traditional CRF were commonly found in this population: 81% had hypertension; 77% were obese; 55% had hyperlipidemia; 54% reported a family history of heart disease; 37% had diabetes; and 22% smoked. 93% of respondents had 2 or more CRF and 81% had 3 or more. The most common combination of risk factors was hypertension and obesity. Although 93% has 2 or more CRF, only 1/3 perceived themselves at high risk. When we compared perceived risk with actual risk, the patients’ perceptions did not correlate with their actual risk factors (Pearson's r = .09, P > .05). Among those at high risk, 63% inappropriately perceived themselves to be at low risk for heart disease. The inability to recognize cardiac risk was predicted only by perceived stress (OR = 4.23, [95% CI 1.84, 9.68]) and the diagnosis of hyperlipidemia (OR = 2.35, [95% CI 1.05, 5.26]).

CONCLUSION:  This urban, disadvantaged population carries a high burden of cardiac risk factors, yet, routinely underestimate their true risk of heart disease largely due to more immediate preoccupation with life stressors. Future heart disease risk reduction and educational programs in these populations need to consider incorporating stress coping modules.

DISEASE AND MEDICATION BELIEFS OF LATINOS AND AFRICAN AMERICANS WITH MODERATE AND SEVERE ASTHMA.A. Oster1; E.A. Halm2. 1Mount Sinai School of Medicine, New York, NY; 2Mount Sinai School of Medicine, Mamaroneck, NY. (Tracking ID #116252)

BACKGROUND:  Latino and African American inner city residents have higher rates of asthma, higher rates of emergency department (ED) visits, hospitalizations, and asthma-related deaths. Although Latinos and African Americans differ in cultural and historical backgrounds, it is unknown whether these groups differ in adherence rates, asthma health beliefs, medication beliefs, or disease self-management practices.

METHODS:  All adults hospitalized with a diagnosis of asthma at a large inner city teaching hospital during 12 consecutive months were interviewed to assess demographic data, access to primary care, asthma severity, adherence to treatment, asthma knowlege, disease beliefs, medication beliefs, and self-management practices. Chi-squared tests, t tests, or Mann-Whitney rank-sum tests were used as appropriate.

RESULTS:  There were 335 hospitalizations by 218 individual patients. Of these, 49% self-identified as Latino, 21% as African-American, 14% as white, and 16% as other. There were no differences between Latinos and African Americans in age (mean 50.3 years, P = .72), or gender (79% female, P = .94). However, Latinos were more likely to have lower income (P < .00), live in public housing (60% vs. 38%P = .02) and rely on Medicaid (60% vs. 42%, P = .03). The two groups reported similar asthma histories: age of initial diagnosis, number of ED visits, number of hospitalizations, oral corticosteroid use (87% vs. 89% African-Americans), and history of intubations, did not differ. Similar percentages reported regular source care and ease in obtaining appointments. However, Latinos were twice as likely to report difficulty receiving telephone advice regarding their asthma (49% vs. 28% of African Americans, P = .04). Twice as many Latinos believed they could abort attacks with inhaled beta-agonists (30% vs. 16% African-Americans, P = .03). Latinos were more likely to be at moderately or very worried about side effects of both beta-agonists (39% vs. 13%, P < .00) and inhaled corticosteroids (ICS) (53% vs.25%, P = .03). However, the groups did not differ in actual ICS use: although 64% of both recalled having them prescribed, most used them only a few times a week, used them less often when without symptoms, and most thought using them when without symptoms, was only “a little important”. More than half of both groups (55% of Latinos and 53% of African-Americans) believed their asthma was present only when they experienced symptoms. There were no differences between Latinos and African Americans in measures of participatory decision-making, satisfaction with health care, self-efficacy, disease self-managment, or awarenes of disease symptoms.

CONCLUSION:  Latinos and African Americans with moderate and severe asthma have similar, though subopitmal health beliefs about their illness. Both have significant knowledge deficits regarding their illness. Latinos express greater concern regarding the negative effects of continuous medication use. Further research regarding the effects of this concern on adherence to treatment is warranted.Educational efforts to improve self-management skills need to be sensitive to cultural differences in perception of risks of medications if they are to be successful in improving adherence rates and imporving asthma control.


BACKGROUND:  Prior reports of disease management (DM) for congestive heart failure (CHF) have shown improved outcomes, yet there are no reports examining health insurance claims in an HMO sponsored DM program. This study compared CHF outcomes and health care costs in an HMO setting for patients in DM versus those not in DM who were included in a Center for Medicare and Medicaid Services’ (CMS’) Quality Assessment and Performance Improvement (QAPI) program.

METHODS:  We retrospectively examined QAPI-based outcomes and paid health care claims over 12 months among 272 continuously enrolled Geisinger Health Plan (GHP) Medicare+Choice (M+C) participants who were in a QAPI study of CHF. Two groups were compared: those who were enrolled in an “opt-in” CHF DM program vs. those who were not enrolled. We compared QAPI® data on cardiac imaging (CI), angiotensin converting enzyme inhibitor (ACEI) and beta-blocker (BB) use, as well as per member per month (PMPM) paid charges for total care, for CHF care, and for inpatient, outpatient and emergency room (ER) care. All QAPI® measures were independently obtained outside any DM data sources, using a “hybrid” method of claims and chart audits. The mean PMPMs were tested for statistical significance by excluding patients with outlier claims that were >3 standard deviations from the calculated mean.

RESULTS:  Of 415 patients with CHF randomly selected for QAPI® measurement, 44 (10.6 %) were enrolled in DM (“program”) vs. 228 (54.9%) who were not enrolled (“non-program”). 143 (34.5%) were excluded because of lack of continuous enrollment during the study period. Both program and non-program patients were similar in M/F ratio (52%/48% vs. 55%/45%, P = NS) and mean age (78.5 vs. 78.8 years, P = NS). There was no statistically significant difference between program vs. non-program patients in the number with CI (N = 42 or 95.5% vs. 202 or 88.5%, P = NS), or in the use of ACEI in the setting of reduced systolic function (N = 5 of 13 or 38.5% vs. 41 of 69 or 59.4%, P = NS). Use of BB was statistically significantly higher among program vs. non-program patients (N = 25 or 56.8% vs. 110 or 48.2%, P < .05). The PMPM for program patients was $1635 vs. $2415 in non-program patients (P < .05). This difference was accompanied by a statistically significant lower PMPM for CHF care ($1333 program vs. $2017 non-program patients, P < .05). Differences observed in total inpatient ($876 vs. $1612), outpatient ($309 vs. $385) and ER care ($81 vs. $54) were not statistically significant.

CONCLUSION:  In this M+C HMO, opt-in CHF DM appeared to be associated with a significant reduction in health care costs and an improvement in the use of beta-blockers, supprting the contention that DM may be useful option in the care of patients with CHF.

DISPARITIES IN DIABETES OUTCOMES: IMPACT OF MENTAL ILLNESS.S.M. Frayne1; H. Lin2; J.H. Halanych3; F. Wang2; D.R. Berlowitz2; K.M. Skinner2; E.J. Sharkansky4; T. Keane4; L. Pogach5; D.R. Miller2. 1VAMC, Palo Alto, CA; 2CHQOER, VA Hospital, Bedford, MA; 3U. Alabama, Birmingham, AL; 4VAMC, Boston, MA; 5VAMC, East Orange, NJ. (Tracking ID #116994)

BACKGROUND:  There is emerging evidence that patients with mental health conditions (MHC) may receive less intensive care for medical conditions. We examined disparities in glycemic/lipemic control for diabetics with vs. without MHCs.

METHODS:  In a national sample of all 339,559 Veterans Health Administration (VA) patients with diabetes (based on ICD9 codes and antiglycemic prescriptions) whose facility submitted lab data to a central registry, we identified MHC from ICD9 codes in VA administrative data. We compared diabetics with vs. without MHC on FY1999 Diabetes Quality Improvement Project measures (poor glycemic control i.e. HbA1c ≥9.5 or test not done; poor lipemic control i.e. LDL ≥ 130 or test not done) in univariate and multivariable analyses (adjusting for age, sex, race and physical comorbidities, in aggregate and in several major MHCs). Finally, we repeated analyses for those self-reporting VA as their exclusive care provider in the 1999 Large Survey of Veterans.

RESULTS:  Comparing the 83,473 with MHC to the 256,086 without MHC, 29% vs. 24% had poor glycemic control (P < .0001), and 54% vs. 50% poor lipemic control (P < .0001). Increasing number of MHC diagnoses was associated with monotonic increases in poor glycemic and lipemic control. Disparities persisted after case mix adjustment, and were greater for certain MHCs (Table). In the 55,694 with care only in VA (i.e. with all care captured), magnitude and direction of disparities were confirmed.

CONCLUSION:  Diabetics with MHC are more likely than those without MHC to have poor glycemic and lipemic control. National diabetes quality improvement efforts should pay special attention to the mentally ill, especially those with substance abuse, psychosis or personality disorders and those with more MHCs; potential patient-level, provider-level or system-level barriers should be explored. Optimizing care for patients with comorbid medical and mental illness may require paradigm shifts for general internists, such as a move toward integrated medical-mental health systems of care.

Table TABLE. Adjusted OR (95% CI) for diabetics with (vs. without) MHCsThumbnail image of

DIURETIC USE AND BONE MINERAL DENSITY IN OLDER MEN: THE OSTEOPOROTIC FRACTURES IN MEN STUDY.L.S. Lim1; H.A. Fink2; M.A. Kuskowski3; J.A. Cauley4; K.E. Ensrud2. 1Mayo Clinic, Rochester, MN; 2University of Minnesota, Minneapolis, MN; 3GRECC, VA Medical Center, Minneapolis, MN; 4University of Pittsburgh, Pittsburgh, PA. (Tracking ID #115089)

BACKGROUND:  Osteoporosis in men is a major health problem that has gained increasing recognition. As the size of the aging population increases, it is expected that the number of men developing osteoporosis and osteoporosis-related fractures will increase. Diuretics, including loop, potassium-sparing, and thiazide diuretics, are commonly used by older adults. They have the potential to affect renal calcium balance. Although much is known about thiazide effects on bone mineral density (BMD), data is lacking on the effects of nonthiazide (loop or potassium-sparing) diuretics on BMD. Previous epidemiologic studies have reported mixed results in the association between loop diuretics and BMD. Our aim is to investigate the association between thiazide and nonthiazide diuretics and BMD in a cohort of older, community-dwelling men.

METHODS:  The study involved participants from the Osteoporotic Fractures in Men (MrOS) study. Cross sectional analysis was performed on 5,835 men aged 65 years and older (mean age 74 years) from 6 metropolitan sites in the United States. Participants were grouped according to diuretic usage into 4 groups: loop, potassium-sparing, thiazide diuretic users, and nonusers. BMD was assessed by dual X-ray absorptiometry at the total hip, femoral neck, trochanter, lumbar spine, and total body. ANCOVA was used to assess the age- and multivariate-adjusted associations between each group of diuretic use and BMD.

RESULTS:  Loop diuretic users had a higher mean age-adjusted BMD that was 2.8 to 4.4 percent greater at the total hip, femoral neck, lumbar spine and total body compared to diuretic nonusers (P < .01). This difference was smaller in magnitude and not statistically significant after multivariate-adjustment. There were no differences in mean age- or multivariate-adjusted BMD between users of potassium-sparing diuretics and diuretic nonusers. Thiazide users had a 2 to 3.4 percent greater mean age-adjusted BMD at the total hip, femoral neck, trochanter, lumbar spine and total body compared to diuretic nonusers (P < .01). After multivariate-adjustment, this difference remained statistically significant only for the total body site (difference 1.3%, P < .01). There were no significant differences in mean age- or multivariate-adjusted BMD between categories of diuretic users.

CONCLUSION:  Our results demonstrate that loop and potassium-sparing diuretic use in older men do not adversely influence BMD. Overall, the results suggest that diuretic use is not a major correlate of BMD in older, community-dwelling men.

DO BRIEF MEASURES OF READINESS TO CHANGE PREDICT ALCOHOL CONSUMPTION AND CONSEQUENCES IN PRIMARY CARE PATIENTS?R. Saitz1; N.J. Horton2; J.H. Samet1. 1Boston University, Boston, MA; 2Smith College, Northampton, MA. (Tracking ID #115939)

BACKGROUND:  Measures of readiness to change drinking are recommended in clinical practice to help tailor interventions and predict outcomes, but their validity has not been well established. We hypothesized that readiness measures would predict alcohol consumption and consequences among hazardous drinkers in primary care.

METHODS:  We studied a prospective cohort of adults in a randomized trial of providing alcohol screening results to physicians in an urban primary care practice. Subjects were current drinkers, of hazardous amounts or with a + CAGE alcohol screening question, who were followed-up 6 months later. Assessments of 6 predictors at study entry included: 1) readiness to change drinking; 2) importance of changing; 3) confidence in ability to change; 4) intention to cut down; 5) intention to abstain completely (single item measures); 6) stage of change using the validated multi-item Readiness to Change Questionnaire (RCQ). We fit multivariable regression models for all predictors and 6 outcomes: 1) abstinence; 2) binge drinking (>3 drinks for women or elderly, 4 for men); 3) hazardous drinking (binge or >7 drinks per week for women or elderly, 14 for men); 4) drinks/day; 5) drinks/drinking day; 6) consequences. Analyses adjusted for clustering within physician, physician training, randomization, sex, race, medical comorbidity, and education. Models predicting consequences or consumption also controlled for consequences or drinks/drinking day at study entry, respectively.

RESULTS:  Of 312 hazardous drinkers enrolled, 235 completed follow-up and 228 had complete data; 33% were in Action and 31% the Contemplation stage of change (RCQ). Both confidence (scale from 1–10) and intention to abstain (5 point Likert scale) were associated with reduced consumption: less binge drinking (Adjusted Odds Ratio [AOR] for an additional point 0.88, 95% CI 0.80–0.98 for confidence, AOR 0.79, 95% CI 0.64–0.98 for intention); and hazardous drinking (AOR 0.89, 95% CI 0.79–1.00 for confidence; AOR 0.78, 95% CI 0.62–0.98 for intention). Intention to abstain was also associated with more abstinence (AOR 1.43, 95% CI 1.09–1.88). No other predictors were associated alcohol consumption. All predictors except intention to abstain were significantly associated with later alcohol consequences (more readiness, more consequences).

CONCLUSION:  Only confidence to change drinking and intention to abstain are associated with later reductions in drinking. The association between readiness and consequences may seem counterintuitive but is likely explained by consequences serving as powerful motivators. Among hazardous drinkers in primary care, tools to assess readiness to change drinking can be valid but must be selected carefully because measures do not consistently predict changes.

DO BUSY AND TIRED RESIDENTS TEACH LESS?C.A. Feddock1; A.R. Hoellein1; J.P. Shah1; B.L. Houk1; J.F. Wilson1; T.S. Caudill1; C.H. Griffith1. 1University of Kentucky, Lexington, KY. (Tracking ID #116790)

BACKGROUND:  When third year medical students rotate on inpatient internal medicine services, resident physicians provide a substantial amount of clinical teaching. To our knowledge, no studies have examined the effect of patient care duties on the quality or quantity of resident teaching. The purpose of our study was to identify the physician factors which influence the amount of medical student teaching.

METHODS:  Over a three month period, residents on our inpatient internal medicine ward services were approached daily by a research assistant and asked to complete a brief questionnaire. Residents were asked to subjectively rate their current workload (light, medium, heavy or extremely heavy) and to rate their last night's sleep (sufficient versus insufficient). Residents also completed a 5-item satisfaction questionnaire, which included rating the statement, “I spent enough time teaching students,” on a 5-point Likert scale (1 = strongly disagree, 5 = strongly agree). For the purposes of analysis, workload was subdivided into low (light or medium) and high (heavy or extremely heavy). Analysis included correlation and multiple regression approaches to assess the effect of workload and fatigue on adequacy of student teaching.

RESULTS:  Forty-three different residents completed a total of 200 surveys. Nineteen percent of residents reported high workloads, 27% reported insufficient sleep, and 13% reported both a high workload and insufficient sleep. Overall, residents gave themselves low ratings for the adequacy of their time spent teaching (2.65  1.09). Lower ratings correlated with higher resident workload (r =–.20, P = .005), with insufficient sleep (r =–.16, P = .019), and marginally with the resident being post-call (r =–.14, P = .056). In the regression analysis, independent predictors of inadequate student teaching were high workload, inadequate sleep and the interaction of workload and sleep. Student teaching was most significantly impacted by residents who had both high workload and insufficient sleep. For example, if residents had high workload and insufficient sleep, 45% rated their teaching as completely inadequate (1 on the 5-point scale) versus 15% if residents had either light workload or heavy workload with sufficient sleep (P = .0006).

CONCLUSION:  High workloads and insufficient sleep dramatically decrease the adequacy of time residents spend teaching medical students. Further studies are needed to identify other factors which impact resident teaching of medical students. Academic medical centers should consider the effects that resident schedules and patient care duties may have on medical student education.

DO COLLABORATIVE QUALITY IMPROVEMENT PROGRAMS REDUCE CARDIOVASCULAR RISK FOR PERSONS WITH DIABETES?R.B. Vargas1; C.M. Mangione1; J. Keesey2; S. Asch3; M. Schonlau2; E. Keeler2. 1University of California, Los Angeles, Los Angeles, CA; 2RAND, Santa Monica, CA; 3West Los Angeles Veteran's Administration, Los Angeles, CA. (Tracking ID #117153)

BACKGROUND:  In patients with diabetes, cardiovascular morbidity and mortality is a significant and potentially preventable complication. There is a need to identify successful interventions to improve clinical outcomes for these patients that can be reproduced in real world settings. We aim to examine the impact of implementing the Chronic Care Model (CCM) through Breakthrough Series collaboratives on the reduction of cardiovascular disease risk in patients with diabetes.

METHODS:  Representatives from 40 healthcare organizations participated in a collaborative instructing them on how to implement organizational changes to improve quality of care for diabetes in an intervention group of their patients. Of these, 17 organizations were large enough to provide matched diabetic patients from non-intervention clinical sites as controls, and 13 agreed to participate. We identified patients with diabetes by chart review from both intervention and within organization control sites and calculated the risk for fatal or non-fatal myocardial infarction or sudden death using a modified United Kingdom Prospective Diabetes Study (UKPDS) risk engine based on glycemia, blood pressure, and cholesterol level. The pre and post periods were one year before and after the intervention. Patients were eligible for inclusion if they were over 25 years old, had 5 out of the 6 variables needed to calculate UKPDS cardiovascular risk score (78 % of sample). We used hierarchical regression to compare the cardiovascular risk of patients from intervention and control sites pre and post participation in the collaborative. This model adjusted for patient level variation in age, medical co-morbidities, and severity of diabetes as well as clustering by health care organization site.

RESULTS:  564 patients from intervention sites and 606 patients from control sites met the inclusion criteria. The mean age was 64 years, 39% were female, and the mean 1 year risk of myocardial infarction or cardiovascular death as estimated by the UKPDS score pre-intervention was 3.1% for the intervention group and 3.0% for the control group. Although the UKPDS risk score improved in both groups, the intervention group had significantly greater risk reduction when compared to the control group (P < .0001).

CONCLUSION:  Collaborative interventions to improve quality of care can be implemented in practice based settings. Over a one year interval, implementation of the CCM through the Breakthrough Series significantly lowered the cardiovascular risk of patients with diabetes who were cared for in the participating organization's settings.

DO DRUG PRICES REFLECT DEVELOPMENT TIME AND GOVERNMENT INVESTMENT?S. Keyhani1; M. Diener-West1; N.R. Powe1. 1Johns Hopkins University, Baltimore, MD. (Tracking ID #117156)

BACKGROUND:  Antiretroviral and cancer drugs are among the most highly priced medications. The pharmaceutical industry has cited lengthy development times as one reason for high drug prices. We compared the price of antiretroviral and cancer drugs to other drugs after adjusting for development time, government support and other characteristics.

METHODS:  We conducted a retrospective study of 180 drugs. We assembled data on drug development times, drug characteristics, government support and prices. There were 168 drugs with development data available in the Government Printing Office (GPO) database approved between 1994–2002 by the Food and Drug Administration (FDA). We obtained data on 16 antiretrovirals developed between 1987 and 2002. Development data for antiretrovirals was obtained from multiple sources including the GPO database, new drug applications (NDA) filed with the FDA and newspaper reports. We obtained data on government support from the FDA, the NDA, and the US Patent and Trademark Office. Prices were obtained from the 2002 Redbook. Drugs were grouped into 7 classes. First, we compared the median development period and government support of drugs by class. Second, for oral drugs, we assessed the independent effect of drug class on price per day in a multivariable analysis, controlling for drug characteristics, government support and development time.

RESULTS:  Thirty percent of antiretrovirals had government patents compared to 10% of antimicrobials, 3% of cardiovascular drugs and no cancer, neuropsychiatric, transplant and gastrointestinal drugs (P < .002). Fifty percent of all antiretrovirals had NIH trials listed in the NDA compared to 7% of cancer drugs, 7% of antimicrobials, 3% of cardiovascular and no other drug classes (P < .001). Thirty percent of antiretroviral drugs had regulatory support compared to 21% of cancer drugs, 3% of antimicrobials and no other drug classes (P < .001). The median development period of antiretrovirals (4.2 yrs) was lower than cancer drugs (5.6 yrs), antimicrobials (5.7 yrs), cardiovascular drugs (7.8 yrs), neuropsychiatric drugs (7.1 yrs), transplant drugs (6.7 yrs) and gastrointestinal drugs (7.8 yrs) (overall P < .004). By multivariable analysis antiretrovirals cost $8 per day more than the reference group (cardiovascular drugs) while cancer drugs cost $11 per day more after adjusting for development time, government support and drug characteristics (P < .001). The remaining drug classes did not differ in price from the reference category. Development time was not associated with drug price (P = .18).

CONCLUSION:  Antiretroviral and cancer drugs even after accounting for development time and government support are among the most highly priced medications. Development time is not associated with drug price.


BACKGROUND:  The National Board of Medical Examiners (NBME) will implement USMLE Step 2 Clinical Skills (CS) as a licensure requirement to practice medicine beginning in 2004. The exam will assess students’ history, physical exam and communication skills using a series of standardized patient (SP) encounters. SP's will be primarily responsible for scoring student performance during the encounter. Physician evaluators will score post-encounter clinical notes and have a limited role in evaluating the clinical skills of the students. This format assumes that SP's and faculty scoring of an encounter during the exam would be similar. The purpose of our study was to determine the correlation between faculty and SP scores during the clinical competency exam (CCX) conducted at the University of Cincinnati College of Medcine (UCCOM) which is similar to the planned USMLE Step 2(CS).

METHODS:  During academic years 2002 and 2003, the UCCOM conducted a CCX for all 4th year students. The CCX consisted of 7 SP-encounters. Faculty and SP's rated history, physical exam and communication skills of each student using checklist evaluations. Faculty and SP's participated in workshops to enhance their evaluation skills prior to the CCX. Mean scores and standard deviations were calculated for each skill for the two years the CCX has been administered. Pearson coefficients were used to determine the correlation between faculty and SP scores.

RESULTS:  To date, 313 students have participated in the CCX. The table below shows strong agreement between faculty and SP scores of students’ history and physical exam skills. A significantly weaker correlation existed between faculty and SP scores of students’ communication skills.

CONCLUSION:  Our results suggest that significant variation exists between faculty and SP's when evaluating communication skills of students. We recommend the NBME carefully monitor SP's evaluations of students’ communication skills as USMLE Step-2(CS) begins this year.

Clinical SkillFaculty Mean Score (sd)SP Mean Score (sd)r (p)
History144.4 (15.7)146.6 (17.1)0.82 (P < .0001)
Physical Exam 53.9 (10.3) 53.9 (10.3)0.90 (P < .0001)
Communication100.3 (9.9) 95.3 (9.2)0.41 (P = .001)

DO FACULTY-MEDIATED INTERN SUPPORT GROUPS REDUCE BURNOUT? R. Dillingham1; M.J. Goodman1; A.M. Wolf1; M.L. Plews-Ogan1; J.B. Schorling1. 1University of Virginia, Charlottesville, VA. (Tracking ID #116859)

BACKGROUND:  Burnout is common among internal medicine residents. In an attempt to reduce burnout among interns, we conducted support groups to provide a protected opportunity for housestaff to discuss their emotions relative to their professional role. The purpose of this study was to evaluate whether these faculty-mediated intern support groups could reduce the rate of burnout.

METHODS:  All 32 PGY-1 internal medicine residents were assigned to support groups in June 2002. Three groups of 10–11 interns met for one hour every other week in the hospital starting in July, 2002 at times convenient to the group members. Each group was facilitated by both a general medicine faculty member and a clinical psychology graduate student. The groups were designed so that no faciltator had any formal supervisory or evaluative role realtive to the residents in his or her group. Residents were asked to write for 5–10 minutes on specific topics at the beginning of the session, and discussions centered on the writing topics. The support group program was mandatory for the interns from July to December except if they were on-call or post-call and voluntary thereafter. The primary outcome measure was the Maslach Burnout Inventory (MBI) scores of the intern group for whom the support groups were required. The MBI was administered in February 2003. Residents participating in the support groups also completed an anonymous survey with thirteen questions measured on a Likert scale to determine their satisfaction with the groups and to evaluate the functioning of the groups.

RESULTS:  A total of 78% of all interns completed the two instruments. Of these, 84% met criteria for burnout. The criteria for identifying burnout were high scores on either the depersonalization or emotional exhaustion subscales of the MBI using the standards for health professionals. Eighty percent of the residents reported on the surveys that they valued the support groups, and 60% wished to continue meeting with their groups.

CONCLUSION:  The faculty-mediated intern support groups did not seem to reduce the prevalence of burnout in February of the intern year. Although the MBI was only administered at one point in time, the high rate of burnout identified makes any effect of the support groups highly unlikely, and the rate we observed was similar to that reported by others for medical residents who were not participating in such groups. However, the interns valued the opportunity to share their feelings about the residency and their new professional role with other interns and faculty. It is likely that successful efforts to reduce burnout among residents will need to address the stuctural issues of residency and not just provide outlets for them to address the emotional consequences of their work.

“DOC, THE METHADONE EATS MY BONES!” BONE HEALTH IN CHRONIC METHADONE TREATMENT.T.W. Kim1; D.P. Alford1; A.O. Malabanan1; B. Batch1; J.H. Samet1. 1Boston University, Boston, MA. (Tracking ID #116879)

BACKGROUND:  While methadone maintenance treatment (MMT) is effective for opioid dependence with an excellent safety record, bone pain complaints have raised concerns about these patients’ bone health. We assessed the prevalence of bone pain, vitamin D insufficiency and osteoporosis in persons receiving chronic methadone.

METHODS:  Subjects recruited from an urban MMT program from 8/03 to 12/03 underwent the following: administered questionnaire (demographics, perceptions of methadone and bone health, bone pain), serum 25, hydroxyvitamin D (Vit D) and dual energy x-ray absorptiometry (DXA) for bone mineral density (BMD). Vitamin D insufficiency was defined as Vit D < 20 ug/ml, osteopenia as T score ≤ −1 and > −2.5 and osteoporosis as T score ≤ −2.5.

RESULTS:  Study sample (n = 88) characteristics: median age 42 years (range 20–66); female 61%, of whom 34% post-menopausal; black 47%; HIV infection 27%; tobacco use 89%; median years illicit opioid use 12 (range 2–38); median years MMT 3 (range 0–25). Survey results: 43% (38/88) somewhat or very worried that methadone affects bone health; 20% (18/88) delayed initiation of MMT due to this worry; 33% (29/88) felt this worry would influence their MMT duration; 20% (18/88) had declined a recommended increase in methadone dose, with 41% (7/18) of those subjects citing concerns about bone health as reason for refusal; and 64% (56/88) reported bone pain. Serum analysis revealed 32% (19/59) had Vit D insufficiency. DXA results (n = 79) (see table below) revealed normal BMD in only 13% of men, 44% of postmenopausal women and 52% of premenopausal women.

CONCLUSION:  Almost half of the MMT subjects in our sample believed that methadone adversely affects bone health. The prevalence of vitamin D insufficiency was substantial and a high prevalence of osteoporosis was demonstrated, especially among our male subjects. Since osteoporosis and osteomalacia are treatable bone diseases, further study is warranted to address the abnormal bone health of patients in MMT.

male28% (9/32)59% (19/32)
female, premenopause32% (10/31)16% (5/31)
female, postmenopause31% (5/16)25% (4/16)

DOES HERBAL PRODUCT USE AFFECT THE USE OF APPROPRIATE MEDICATIONS FOR SECONDARY PREVENTION OF CORONARY DISEASE?S.W. Bent1; J.A. Simon1; M. Shlipak1. 1University of California, San Francisco, San Francisco, CA. (Tracking ID #116475)

BACKGROUND:  Herbal products are widely used in the United States, with recent estimates suggesting that approximately 25% of adults have used an herb to treat a medical illness within the past year. One great concern about the widespread use of herbs has been a possible “substitution effect,” whereby herb users may be less likely to take prescription medications with established benefits for their condition. In a randomized controlled trial of post-menopausal women with known coronary heart disease (The Heart Estrogen Replacement Study; HERS), we sought to determine whether patients using herbal products at baseline were less likely to use four medications with well-documented benefits: statins, beta-blockers, angiotensin-converting enzyme inhibitors (ACE-I), and aspirin.

METHODS:  The HERS study included 2763 postmenopausal women less than 80 years of age who had not had a hysterectomy and had known coronary artery disease. Medication and supplement use, including herbal products, was assessed at baseline (1993) and every four months until study completion (1998) by phone interview. We manually searched the entire medication database for names of herbal products using the Natural Medicines Comprehensive Database as the reference standard. Participants were classified as users of any herbal product or non-users. Using a students t test, we then determined whether herb users had a different likelihood of taking the four medications known to be beneficial in patients with coronary heart disease.

RESULTS:  The proportion of HERS participants using herbal products nearly doubled during the study period, from 6.1% of women at baseline (1993) to 11.0% of women at the final visit (1998). The most commonly used herbal products at baseline were Garlic (3.5%) and Ginkgo 0.3%. Herb users had similar likelihood of taking statins [OR = 1.26 ( 95% CI: 0.92 to 1.72 )], beta-blockers [OR = 0.84; (95% CI: 0.59 to 1.18), ACE-I [OR = 0.94 (95% CI: 0.62 to 1.42], and aspirin [OR = 1.31 (95% CI: 0.87 to 1.97)] compared to non-users.

CONCLUSION:  Although herb use increased steadily among women with CHD during the period of observation, herbal product use did not appear to interfere with the use of appropriate medications. While this finding is reassuring, we cannot determine whether it applies to other chronic disease populations among whom the prevalence of herb use is even higher, such as patients with cancer, HIV, depression, and chronic pain.

DOES INSURANCE STATUS AFFECT PHYSICIANS’ PRESCRIBING FOR SEVERE, UNCONTROLLED HYPERTENSION?S.M. Dy1; J.H. Young1; K. Weeks1; R. Wongus1; M.J. Klag1. 1Johns Hopkins University, Baltimore, MD. (Tracking ID #115685)

BACKGROUND:  Physicians often underestimate medication costs and the impact of levels of insurance coverage on out-of-pocket costs. Our objective was to determine whether insurance status and pharmacy benefits influence physicians’ prescribing of antihypertensive medications in persons admitted for severe, uncontrolled hypertension.

METHODS:  We conducted a cross-sectional survey. We prospectively identified all African-American urban residents admitted to the internal medicine service at Johns Hopkins from 1999–2003 whose admission blood pressure was greater than 180/110 mmHg. We reviewed patients’ medical records and interviewed patients about their history of hypertension, adherence, medications, and access to care. We excluded patients with a secondary cause of hypertension. We defined medication costs as the average cost to the patient for 30 days’ treatment with the lowest dose, based on survey data from retail pharmacies nationwide.

RESULTS:  We completed interviews with 159 patients; 87% were under 65 years of age. For the 77% of patients where we had information on prescribed medications before admission, the mean number of antihypertensive medications was 2.3 (range, 1–6), with an average cost of $53 (range, $6–$161). Patients were taking a mean of 2.3 non-antihypertensive prescription medications on admission. There were no significant differences in the number or cost of medications between the admission and discharge regimens. In regard to insurance, 33% had none, 14% had no medication coverage, and 29% had medication copays. The prevalence of reported difficulty affording medications in these insurance groups was 78%, 57%, and 38%, respectively. Of those who reported difficulty affording medications, 68% said that this sometimes resulted in nonadherence. Patients without insurance had lower discharge medication costs than those with insurance ($43 vs. $53, P < .05), but there were no significant differences in discharge costs among patients who reported difficulty affording medications, those without medication coverage, or those with medication copays. These same findings persisted in multivariate analyses adjusting for the cost of the medication regimen on admission, for those patients where we had information about prescribed medications on admission. In multivariate analysis, after adjusting for the number of discharge antihypertensives, the presence of an ACE inhibitor or calcium channel blocker in the discharge antihypertensive regimen significantly increased its cost (P < .001).

CONCLUSION:  Insurance status, but not other important indicators of being able to afford antihypertensive medications, influenced physicians’ prescribing patterns. When not medically contraindicated, physicians might decrease medication costs by substituting lower-cost medications for ACE inhibitors and calcium channel blockers.

DOES KNOWLEDGE ABOUT PROSTATE CANCER SCREENING INFLUENCE PHYSICIANS’ SCREENING PRACTICES?D.S. Bell1; J. Hoffman1; M. Wilkes2. 1University of California, Los Angeles, Los Angeles, CA; 2University of California, Davis, Davis, CA. (Tracking ID #116985)

BACKGROUND:  Professional education often has little effect on physicians’ practices; thus, the influence of knowledge on practice behavior has been questioned. Many physicians routinely order PSA testing for men of all ages, despite the conclusions of multiple expert groups that routine screening is unlikely to benefit men outside of the 50–69 year-old age group. We report an interim analysis from a study relating physicians’ knowledge about prostate cancer screening with their self-reported practices.

METHODS:  We developed a multiple-choice knowledge test on based on the U.S. Preventive Services Task Force guideline on prostate cancer screening. Each question was reviewed and approved by a urologist expert and by the Task Force chairman. We also developed a survey to assess prostate cancer screening attitudes and practices. Using an online survey system, we administered these questions to a sample of general internists and family physicians listed in the AMA Physician Masterfile as practicing in Los Angeles. To be eligible, physicians were required to spend at least 1 day/week delivering primary care to men over age 40. Participants were paid $75. A single knowledge scale was constructed by including questions having an item-total correlation of 0.10 or greater. A PSA overuse scale was constructed by assigning points for the self-reported routine use of PSA screening in age groups under 50 and over 70. The overuse scale ranged from 0 to 5, with 5 points representing routine use of PSA screening in all age groups, including those over 80.

RESULTS:  Of 364 physicians randomly sampled, 163 (45%) responded, either online or by telephone, to the eligibility question. Of these respondents, 40 (25%) were ineligible. Of those eligible, 52 (42%) have completed the study to date. The final knowledge scale included 22 questions, with Cronbach's α = 0.73. The mean knowledge score among participants was 8.3 out of a possible 22 (SD = 3.6, range 0–15). The mean PSA overuse score was 2.8 out of 5 (SD = 1.6), with 12% having a score of 0 and 16% having a score of 5. Higher scores on the knowledge scale were associated with lower scores on the PSA overuse scale (r = −0.53, P < .0001).

CONCLUSION:  Physicians with greater knowledge about prostate cancer screening reported performing significantly less PSA screening for men under age 50 and over age 70. Further studies are needed to determine whether education that changes knowledge about prostate cancer screening can reduce the use of PSA screening among men in these age groups, for whom the harms may outweigh the benefits.

inline image


BACKGROUND:  Pain is the most common reason for medical encounters in the US, yet remains under-diagnosed and under-treated. To comply with pain management standards, Bellevue Hospital implemented a mandatory computerized pain assessment tool in its electronic medical record system for every outpatient encounter. Beginning in August 2002, each healthcare provider was required to ask patients whether they were experiencing any pain and to document the magnitude, quality, and location of the pain as well as causative and alleviating factors. We sought to evaluate whether this intervention led to an increase in the diagnosis of pain-related conditions.

METHODS:  We conducted a chart review of all adult primary care patients seen on two randomly selected days six months prior to implementation of the computerized instrument (Pre-group) and six months after (Post-group). For all visits for which an electronic note was completed (68% before and 99% after August 2002) we determined if patients were asked or responded about pain as documented in the computerized note. All pain-related diagnoses in the notes and in the ICD-9 coding were assessed by two independent reviewers (kappa = 0.98). We also conducted an anonymous survey of the Medicine housestaff and faculty by mail to assess their usage and attitudes toward the computerized pain assessment tool.

RESULTS:  Of the 264 patients in the Pre-group, 41% had a pain-related diagnosis compared with 37% of the 392 patients in the Post-group (P > .05). A pain inquiry by the physician was noted for 51% of patients before the mandated assessment tool vs. 44% after (P > .05). Of 144 patients in the Post-group with pain-related diagnoses, 95 (66%) had the presence of pain documented in the free text portion of the progress note. However, for 64 (67%) of these 95 cases, the physicians recorded a contradictory “no pain” response in the assessment tool, therefore bypassing the mandatory screens that would have followed a “yes” response. In the survey (response rate 42%, 85/201) 49% felt the computerized assessment tool was “somewhat difficult” or “very difficult” and 32 % felt it was “somewhat easy” or “very easy”. While 30% never used the assessment tool, 47% used it sometimes, and 23% used it always. The majority of respondents felt that the tool did not change their pain assessment practice.

CONCLUSION:  The mandatory computerized pain assessment tool did not lead to increased documentation of pain-related diagnoses and may have actually hindered the documentation of pain assessment because of the perceived burden of using the application. Computerized documentation tools should be carefully assessed before they are integrated into an electronic medical record system.

DOES NIGHT FLOAT IMPACT THE QUALITY OF THE ADMIT HISTORY AND PHYSICAL?H. Kaushal1; J. Fischer1; H. Khurana2; A. Reddy2. 1University of Illinois, Peoria, IL; 2University of Illinois at Peoria, Peoria, IL. (Tracking ID #115869)

BACKGROUND:  The recent national mandate for an eighty-hour resident physician work week has required nearly all internal medicine training programs in this country to incorporate a “night float” system into their curriculum. Despite this widespread adoption of the “night float” system, little is known regarding its effects on important outcomes, most notably patient care. The objective of this study is to assess the difference, if any, in quality of the resident physician history and physical exam (H&P) write-up on night float (NF) versus non-night float (non-NF) patient service rotations.

METHODS:  The study was conducted at a university-affiliated teaching hospital. Admission H&Ps eligible for review included all those with the principal discharge diagnosis of pneumonia or COPD (chronic obstructive pulmonary disease) that had been completed by an internal medicine resident between July 2002 and June 2003. A criteria checklist for each diagnosis that assessed H&P quality was created utilizing consensus opinion of five attending physicians (face validity). These checklists allowed comparison of five quality variables (history length, medication list completeness, history documentation, exam documentation, ancillary data documentation) between the H&Ps completed during NF vs. non-NF rotations. Reviewers blinded to both patient and physician identifiers utilized these checklists to score the H&Ps. Interrater reliability was assessed. All five quality variables studied were analyzed using a Mann-Whitney U test.

RESULTS:  A total of 79 patient records (21 COPD, 58 pneumonia) were identified as eligible and all were reviewed. Fifty-two of these records reflected non-NF while 27 reflected NF duty. Kraskal-Wallis analysis showed excellent agreement amongst reviewers in the scoring of patient records (P = .87 to P = .97). No significant difference in H&P quality was found between NF and non-NF duty assignment for any of the five quality indicators studied. This held true when examining each diagnosis separately as well as when combined.

CONCLUSION:  With the increase in “night float” utilization comes an increase in the transfer of patient care information between physicians. As such, the quality of the written H&P gains even greater importance in assuring a smooth transition of care from one provider to the next. Our finding that the quality of resident H&Ps did not differ between NF and non-NF services is encouraging and serves as an initial step prompting further research into this timely and nationally pertinent issue.

DOES PARITY INFLUENCE BONE DENSITY? RESULTS FROM THE CARDIA STUDY.K.P. Palonen1; C.E. Lewis1; S. Sidney2; S.B. Hulley3; C.I. Kiefe1. 1University of Alabama at Birmingham, Birmingham, AL; 2University of California, Berkeley, Oakland, CA; 3University of California, San Francisco, San Francisco, CA. (Tracking ID #115268)

BACKGROUND:  Calcium metabolism changes during pregnancy and lactation. Bone density decreases transiently during lactation, before returning to baseline. The chronic effect of multiple pregnancies on bone density is less clear. The purpose of our study is to clarify this relationship further.

METHODS:  We studied 545 women who had whole bone density measured with dual-energy x-ray absorptiometry (DXA) at year 10 (1995–96) of CARDIA, a cohort of African-American and Caucasian women aged 18–30 years at enrollment in 1985–1986. Of them, 302 had total hip and femoral neck bone density measured. We studied bone density in relation to number of live births and used linear regression analysis to control for possible confounders.

RESULTS:  Women with whole body bone density measurement were on average 35.3 ± 3.6 years old, and 47% Caucasian; the table shows that unadjusted bone density increased with parity but this increase was not statistically significant after controlling for significant confounders (race, body weight, lean body mass, smoking, and level of education). Measurements of bone density of the total hip and femoral neck revealed similar findings.

CONCLUSION:  Bone density does not appear to decrease with multiple births.

 0 live births1–2 live births>2 live births
  • * P = .003

  • **

    P = .058

N (%)225 (47)254 (47)66 (12)
Bone density: mean (SD) g/cm2*1.099 (0.087)1.106 (0.100)1.145 (0.097)
Adjusted bone density (SD)**1.103 (0.085)1.106 (0.083)1.131 (0.084)

DOES PERSONALITY HARDINESS PROTECT AGAINST RESIDENT BURNOUT?R. Gopal1; J.J. Glasheen2; T.J. Miyoshi2; A. Prochazka3. 1Denver VAMC, Denver, CO; 2University of Colorado Health Sciences Center, Denver, CO; 3Denver VA, Denver, CO. (Tracking ID #116107)

BACKGROUND:  Burnout is very common in internal medicine residents. The most common features are a combination of emotional exhaustion, depersonalization, and reduced personal accomplishment. Personality hardiness has been thought to buffer one from burnout.

METHODS:  We administered a postal survey in May 2003 to first, second and third year internal medical residents, at the University of Colorado Health Science Center. The survey contained the Maslach Burnout Inventory, a 22-item questionnaire organized into three subscales: emotional exhaustion (EE, high score ≥ 27), depersonalization (DP, high score ≥ 10), and personal accomplishment (PA, high score ≤ 33), the thirty item Cognitive Hardiness Scale, the two question Primary Care Evaluation of Mental Disorders (PRIME-MD) Patient Health Questionnaire (PHQ). We used linear regression to examine predictors of Maslach subscale scores.

RESULTS:  The response rate was 87% (121/139) (33% PGY1, 31% PGY2, 35% PGY3). Two-thirds of the respondents were in the 26 to 30 year old age group with 31% in the 31–35 year old group. The mean scores were EE (24.7, sd 9.0), DP (11.6, sd 5.7), and PA (39.1, sd 5.2). Fifty-one (52%) of the residents had high EE, 74 (61%) residents had high DP, and 19 (15.7%) residents had low PA. Defining burnout as high EE or high DP, 81 residents (67%) met criteria for burnout. Sixty-two (52.5%) residents met criteria for depression using the two-question PRIME-MD screener. Depressed residents had significantly higher levels of DP and EE (P < .05) but there was no significant relationship with PA (P = .89). The mean hardiness t-score was 54 (sd 7.8, range 38 to 69) that is above average compared to the general population (mean 50, sd 10, P < .001). Residents with high levels of DP had lower levels of hardiness (57.7 vs. 52.5, P = .006). Similarly, those with high levels of EE had lower levels of hardiness (58.4 vs. 51.5, P < .001), while those with high PA had low hardiness (50.2 vs. 57.4, P < .001). Hardiness remained associated with EE (P = .001), DP (P = .001) and PA (P = .001) after adjustment for age, gender, PGY level, and work hours.

CONCLUSION:  Burnout continues to be a major problem in medical residency training programs and seems to be associated with depression. High levels of hardiness are associated with lower EE and DP and may protect the resident against the rigors of residency training and depression.

DOES POOR MEDICATION ADHERENCE PREDICT INCREASES IN ANTI-HYPERTENSIVE THERAPY?R.W. Grant1; N. Wong1; D.E. Singer1. 1Massachusetts General Hospital, Boston, MA. (Tracking ID #116633)

BACKGROUND:  Despite the availability of potent anti-hypertensive (AH) medicines, most hypertensive Americans do not achieve optimal blood pressure control. We analyzed pharmacy claims records to test the hypothesis that inadequate adherence to prescribed AH therapy predicts the need for increases in therapy.

METHODS:  We defined a cohort of patients enrolled in a single health insurance plan between 7/01/99 and 6/30/03 who were newly initiated on AH therapy. We used pharmacy claims data to calculate AH medication adherence. Overall adherence was determined by dividing the total number of pills dispensed by the total number of days from first to last prescription, and interval adherence was calculated by dividing the number of pills from a single prescription by number of days until the next prescription. We performed a cohort analysis of overall adherence to the initial AH medication by comparing patients remaining on a stable regimen to patients requiring an increased dose or addition of a second AH agent. We also performed a nested case-control analysis to determine whether AH adherence was lower in the interval preceding an increase in therapy among cases compared to a similar interval in randomly selected controls matched 1:1 by age, medication name, and elapsed time since first starting AH therapy.

RESULTS:  There were 5089 patients newly started on AH therapy in our cohort. Mean age was 47.8 (± 13) years and 50.1% were women. Over a mean of 35 (± 11.3) months of observation, 935 patients required an increase in regimen. The mean time to increased regimen in these patients was 220 (± 222) days. Patients with increased AH regimens were older (50.8 vs. 47.1 years, P < .001) and had higher overall adherence to their first AH (92.5% vs. 80.7%, P < .0001). Adherence in patients with increased AH regimens remained significantly higher after controlling for age, gender, and treatment for diabetes or ischemic heart disease (P < .0001). In our nested case-control analysis, 750 cases were successfully matched to 750 controls. Adherence among cases in the interval preceding an increase in AH therapy was high and not significantly different compared to a matched interval in controls (97% vs. 100.1%, P = .5, median adherence interval 30 days).

CONCLUSION:  Contrary to our expectations, we found that adherence to the first prescribed AH medicine was higher in patients requiring increases in AH therapy compared to patients remaining on the same dose of a single agent. Moreover, adherence in the period immediately preceding an increase in therapy was no lower than a matched adherence interval in similar control patients. These findings support the conclusion that elevated blood pressure in patients who require an increase in AH therapy is not primarily due to poor adherence to the initial AH regimen.

DOES PUBLICATON BIAS EXIST FOR CLINICAL TRIALS SUPPORTING NEW DRUG APPLICATIONS?S. Jeng1; J. Hamrick1; I. Sim1. 1University of California, San Francisco, San Francisco, CA. (Tracking ID #115104)

BACKGROUND:  The Food and Drug Administration (FDA) approves new molecular entities (NMEs) based on “pivotal” clinical trials submitted by pharmaceutical companies. “Pivotal” trials demonstrate the safety and efficacy of an NME for its indication. There is, however, no requirement that these trials be published. We determined the proportion of pivotal trials that are published and identified significant predictors of publication.

METHODS:  We reviewed the FDA's statistical and medical reviews for all NMEs approved between January, 1998 and December, 2000. Based on the reviews, we identified the pivotal trials for each NME and abstracted the following characteristics of each trial: intervention details, investigator, sample size, primary outcome results, statistical significance, randomization, double-blinding status, and use of intention-to-treat analysis. We then tried to match each pivotal trial to publications in PubMed, the Cochrane Library or CINAHL on the basis of at least two of the following characteristics: sample size, number and location of centers, dosing schedules and primary outcome results. If a pivotal trial was not matched in these databases, we reviewed the Medical Letter for potential references. We performed backward stepwise regression to determine which variables, if any, predicted publication status. Missing data was handled by regression imputation. For published trials, we recorded the journal's impact factor (a measure of a journal's citation frequency) and the length of time from FDA approval to publication.

RESULTS:  Over the 3 year study period, 324 pivotal trials were submitted in support of 90 NMEs. 208 (64%) of the trials were found in the literature, with 1 trial published in a journal that was not indexed in PubMed. For 12 NMEs (13%), none of the pivotal trials were published. Trials with statistically significant primary outcome results were more likely to be published (OR: 3.1, 95% CI: 1.8–5.4). This association remained significant across assumptions about the values of missing data. Sample size, randomization, double blinding, and use of an intention-to-treat analysis did not predict publication status. 70% of the published trials (146/208) were published after FDA approval. Of these 146 trials, the mean time to publication was 14 months (range <1 to 47 months). Trials were published in journals with a median impact factor of 2.9 (range 0.2 to 31.7). 41% of the NMEs (37/90) had at least one trial published in a relatively high-impact journal (impact factor ≥5).

CONCLUSION:  The publication of pivotal trials is incomplete. Over one third are unpublished within three years of FDA approval. Trials with statistically significant results are more likely to be published. Publication bias exists for clinical trials supporting the safety and efficacy of newly approved drugs in the United States.

DOES RACE, LANGUAGE, OR GENDER CONCORDANCE BETWEEN PROVIDER AND PATIENT AFFECT MISSED APPOINTMENT RATES IN PRIMARY CARE?.K.E. Lasser1; I.L. Mintzer1; A. Lambert1; H. Cabral2; D.H. Bor1. 1Cambridge Health Alliance, Cambridge, MA; 2Boston University School of Public Health, Boston, MA. (Tracking ID #101786)

BACKGROUND:  When patients miss primary care appointments, practices function less efficiently. This problem is of particular concern to publicly-funded neighborhood health centers, which are often under serious financial strain. No previous study has simultaneously examined whether race, language, or gender concordance between patient and primary care provider (PCP) has an impact upon missed appointment rates in primary care settings.

METHODS:  We surveyed 64 PCPs at 14 neighborhood health centers in Cambridge and Somerville, MA. We inquired about their self-identified race, language abilities, years in practice, and number of clinical sessions per week. Fifty out of 64 PCPs (78%) responded to the survey after receiving 2 e-mails. We then analyzed data on 56,870 adult primary care patient visits to these 50 PCPs during 2002. We defined a visit to be race concordant if the patient and PCP were of the same race, and to be gender concordant if the patient and PCP were of the same gender. If a PCP spoke a given language fluently, we defined visits with patients who spoke that language to be language concordant. We used chi-square tests to compare differences in groups in the proportion of missed appointments by patient, provider and visit characteristics. We then performed multiple logistic regression analyses of the dichotomous outcome (missed versus kept appointment) using generalized estimating equations.

RESULTS:  Sixteen percent of patients did not attend their first scheduled visit in 2002. Young, publicly insured, and Haitian Creole-speaking patients were significantly more likely to miss appointments than were older, privately insured English-speaking patients after adjustment for age, insurance, language, individual PCP open access, years PCP in practice, and health center. Odds ratios for missing an appointment for patients who had language, race or gender concordance with their PCP were .87 (95% confidence interval [CI], .79–.96), .89 (95% CI, .84–.96) and 1.02 (95% CI, .96–1.09) respectively. Patients whose PCP was at open access were slightly less likely to miss appointments than were other patients (OR .79; 95% CI, .73–.85) Odds ratios for missing an appointment for patients seen at particular health centers varied nearly 3-fold.

CONCLUSION:  Our study demonstrates the influence of the specific practice site on patients’ adherence to appointments; this finding warrants further scrutiny. We provide evidence to support the continued use of open access systems to decrease missed appointments. Finally, we encourage continued efforts to recruit minority PCPs and those with non-English language fluency to practice in community health centers.


BACKGROUND:  Little community-based data exists demonstrating that outpatient (OP) health services exposures can prevent stroke or death for persons with hypertension. This study seeks to identify health services exposures’ impact on stroke incidence and death for these patients.

METHODS:  This retrospective study followed a cohort of hypertensive enrollees in Tennessee's statewide Medicaid managed care system (TennCare) for 3–7 years to determine which health services exposures are associated with beneficial outcomes. OP health services exposures (incl. overall OP visits and visit subtypes) and demographic factors were evaluated using administrative data, linked to vital records, to assess stroke incidence and death. Key independent variables were measured during a 2-year run-in period. Associations with stroke incidence and death were assessed using Cox Proportional Hazards modeling.

RESULTS:  49,520 subjects were followed 4.7 years (on average), with 619 incident strokes and 2,055 deaths. They were predominantly: female (67.7%), white (66.8%), disabled (36.9%), uninsured (38.9%), in statewide MCOs (64.9%), and for-profit MCOs (51.4%). Mean age was 48.5 yrs. Baseline diagnoses incl.: diabetes (21.9%), mental illness (24.5%), substance abuse (8.4%), obesity (6.2%). Patients averaged 5.2 OP visits and 1.0 hospitalization per year. For stroke, significant variables (P < .05) in univariate analyses incl.: age (RR = 1.06), black race (1.69), statewide MCOs (.67), regional MCOs (.61), ER visits (1.06), hospitalizations (1.26), length of stay (1.01), diabetes (2.42), and substance abuse (1.57). In multivariate analysis, age, black race and hospitalizations remained significant. For death, significant variables (P < .05) in univariate analysis incl.: age (1.06), male (1.61), black race (1.13), OP visits (1.02), ER visits (1.08), hospitalizations (1.36), length of stay (1.01), diabetes (2.44), mental illness (1.22), and substance abuse (2.50). In multivariate analysis, these variables remained significant, except for mental illness and ER visits. None of the OP visit subtypes studied were found to be protective, even after adjusting for confounding factors.

CONCLUSION:  Administrative databases can be used to identify major health services exposures and cardiovascular risk factors that place hypertensive enrollees at increased risk of stroke and death. Outpatient physician visits do not protect hypertensive Medicaid managed care enrollees from stroke or death, but rather serve as a marker of increased comorbidity.


BACKGROUND:  Low socioeconomic status (SES) appears to be associated with greater cardiovascular disease (CVD) mortality in women but may also be associated with lesser breast cancer mortality. In order to determine whether the relationship between SES and disease-specific mortality risks were similar across racial groups, we examined the association between SES and mortality from CVD and breast cancer among white and African-American women.

METHODS:  Data were obtained from the National Longitudinal Mortality Study, a population-based dataset that links 10 Current Population Surveys conducted between 1978 and 1985 with National Death Index records between 1979 and 1989. Analysis included 21,303 African-American and 186,323 white women. Education (less than high school, high school, and some college or more) and annual household income (<$10,000, $10,000–$19,999, $20,000 or more) were used as markers of SES in separate models. Logistic regression models estimated the association between SES and breast cancer mortality and between SES and CVD mortality after adjustment for age (years), marital status (married/not-married), and rural/urban residence.

RESULTS:  Unadjusted cumulative incidence of CVD mortality was 4.2 percent among African-American women and 2.3 percent among white women over a period of 8.7 years. Unadjusted cumulative incidence of breast cancer mortality was 0.3 percent among African-American women and 0.4 percent among white women. After adjustment for age, marital status, and residence, less education was still associated with greater CVD mortality among African-American (OR 1.8, 95% CI = 1.03, 3.0) and white women (OR 1.4, 95% CI = 1.3, 1.6). However, after adjustment for these variables, less than a high school education was associated with lower breast cancer mortality among white women (OR 0.73, 95% CI 0.6, 0.9) but not among African-American women (OR 1.1, 95% CI 0.5, 2.3). Similar odds ratios were obtained when income was used as the SES indicator.

CONCLUSION:  The association between SES and cause-specific mortality may differ between white and African-American women for breast cancer but not CVD death. Better understanding of these interactions could guide the design of more effective interventions, particularly those aimed at improving breast cancer outcomes among women.

DOES TIME SPENT WITH THE PHYSICIAN AFFECT INPATIENT SATISFACTION?.A.R. Hoellein1; C.A. Feddock1; N.S. Becker1; J.L. Bowerman1; J.F. Wilson1; T.S. Caudill1; C.H. Griffith1. 1University of Kentucky, Lexington, KY. (Tracking ID #116798)

BACKGROUND:  The purpose of this study was to determine if the time spent in contact with the resident-physicians was reflected by their inpatients’ satisfaction with the care received.

METHODS:  Over a summer period, a convenience sample of inpatients on the internal medicine services at both the University and Veterans Affairs hospitals were surveyed by written questionnaire. The patients were asked to estimate the length of time spent with them by their doctor (resident) that day, if they were pleased with their doctor, whether they would recommend their doctor to friends and family, and overall satisfaction with the care received. Estimated time spent with doctor was categorized into 0–5, 5–15, 15–30, and over 30 minutes. Patient ratings for “I am pleased with my doctor”, “I would recommend my doctor to friends and family”, and “I was satisfied with my visit” were assessed on a 10-point, Likert-type scale. Pearson correlation matrices and multiple regression approaches were used to assess the association of these patient satisfaction domains with the time spent with the physician.

RESULTS:  Three hundred and eight patient surveys were analyzed in this study. Generally, and similar to other studies, patients were pleased with their doctor (8.7 ± 1.4), would recommend their doctor to family and friends (8.8 ± 1.7), and satisfied with their care (8.9 ± 1.3). Time spent with the patient was an independent predictor of patient satisfaction [F(3,304) = 14.3, P < .001]. Overall patient satisfaction was significantly less when the physician spent 0–5 minutes with them (7.8 ± 1.9) than 5–15 minutes (8.9 ± 1.3) (P < .05), and further improved when they spent 15–30 minutes with them (9.3 ± 1.3) (P < .05). There was a plateau effect observed as 30 or more minutes spent with the patient did not supplement their satisfaction (9.2 ± .8). Patient ratings of being pleased with their doctor and willing to recommend the physician to family and friends were similarly reduced if the physician spent 0–5 minutes (7.7 ± 2.0 and 7.1 ± 2.7), improved for 5–15 minutes (8.8 ± 1.2 and 8.9 ± 1.5), peaked for 15–30 minutes (9.2 ± 1.4 and 9.4 ± 1.3), and exhibited similar plateau effects for over 30 minutes (9.1 ± .9 and 9.5 ± .8) (P < .05).

CONCLUSION:  Patients’ perception of time spent with their resident-physician in the hospital is an important component of patient satisfaction, being pleased with the physician, and whether they would recommend the doctor to family and friends or not. Merely spending a few more minutes with the patient can increase a patient's satisfaction and may improve overall patient care.

DOES USING AN ELECTRONIC HEALTH RECORD REQUIRE MORE TIME FOR PRIMARY CARE PHYSICIANS?.L. Pizziferri1; A.F. Kittler1; M. Lippincott1; L.A. Volk1; M.M. Honour2; S. Gupta1; S.J. Wang1; T. Wang1; D.W. Bates2. 1Partners HealthCare System, Inc., Wellesley, MA; 2Brigham and Women's Hospital, Boston, MA. (Tracking ID #116292)

BACKGROUND:  Electronic Health Records (EHRs) can improve quality, safety and efficiency, but they are not yet in widespread use in the U.S. A major barrier to EHR use is the concern that the EHR will take longer for physicians to use than paper-based systems. To address this issue we performed a time motion study to evaluate how physician time utilization differed before and after EHR implementation.

METHODS:  This study was conducted in 5 urban and suburban outpatient clinics implementing an ambulatory EHR. We performed a time motion study in which trained observers directly noted and timed physician activities during a clinic session before and after implementation of the EHR. The activities were categorized and analyzed. Activity categories were divided into: Direct Patient Care (e.g., talking to or examining patient), Indirect Patient Care (e.g., recording or reviewing data in support of patient care), Administration (e.g., reviewing schedule) and Miscellaneous (e.g., idle, eating, walking).

RESULTS:  Before EHR implementation, 26 physicians were observed for a total of 110.1 hours. After EHR implementation 20 physicians were observed for a total of 84.5 hours. Both before and after implementation the average observation length was 4.2 hours and the average number of patients seen during the observations was 9.6. Table 1 summarizes data collected during pre- and post-EHR implementation. Overall time spent per patient decreased 1.1 minutes post implementation. Both before and after implementation the majority of time was spent in Direct Patient Care. Using a paired T-test (P < .05), no significant change in time was found for any activity category post-EHR implementation.

CONCLUSION:  Compared to a paper-based system, a well-designed EHR did not require more time during a primary care clinic session. Time spent with patients remained similar, and the distribution of time spent in major activities did not change.

Table 1. Time spent in analysis categories: pre and post-EHRThumbnail image of


BACKGROUND:  The American Diabetes Association has issued clinical practice guidelines to ensure optimal level of care for patients with Diabetes mellitus. Our hypothesis was that Attending Physicians and Advanced Practice Registered Nurses (APRNs) are more compliant with these guidelines than physicians in-training (Residents and Interns) . We wanted to test this hypothesis to assess the need for improvising on the education of physicians in-training.

METHODS:  We did a retrospective study of patients who were seen at the Adult Ambulatory Practices of Hartford Hospital from 5/1999 to 5/2000 with an ICD diagnosis of Diabetes mellitus.Of the 1,015 patients in the database, we excluded multiple entries of patients resulting in 646 unique identities and analyzed them. We grouped the provider class as ’Professional’ (APRN and Attending physicians) and ‘Students’ (Medical interns and residents).

RESULTS:  We analysed the data by SPSS v.11.5 software. See Table 1 for results.

Table 1. 
Provider ClassFoot ExamDM EducationLifestyle changesPneumovaxFlu shotsHbA1C monitorMicroalb screeningLipid profile
P Value<0.001<0.005<0.005<0.005<0.001Not significantNot significantNot significant

CONCLUSION:  There were no significant differences between the groups with respect to HbA1c monitoring, urine microalbumin and lipid profile screening. However, the ‘professional’ group performed better with respect to patient education and in providing preventive care to these patients. On subset analysis, this difference persisted at all levels of training. We conclude that our initial hypothesis was correct and that we need to improvise on educating the physicians in-training, focussing on patient education and preventive care.

EDUCATIONAL VALUE OF CLERKSHIP ACTIVITIES: STUDENT PERSPECTIVE.D.R. Calkins1; A.S. Peters1. 1Harvard Medical School, Boston, MA. (Tracking ID #117419)

BACKGROUND:  Despite many changes in clinical medicine in the past century, the pattern of clinical education, especially during the third year, has changed very little. There is now considerable interest in improving clinical education. To help inform this process, we surveyed third-year medical students regarding their clerkship activities and the perceived educational value of those activities.

METHODS:  We asked students to report, for the previous 24 hours, how they had allocated their time among 24 common clerkship activities. This list was based on input from a student focus group. We also asked students to rate the educational value of these activities using a 5-point, Likert scale (very low-very high). For purposes of analysis, we grouped survey items into seven categories: patient work (7 items), patient care where the student had a secondary role (2 items), patient care where the student had a primary role (4 items), teaching where the student was a secondary focus (2 items), teaching where the student was a primary focus (4 items), teaching where the student was the teacher (2 items), and personal time (3 items).

RESULTS:  We received completed surveys from 63 students. Students reported allocating their time (in hours) as follows: patient work (4.4), patient care: secondary role (3.2), patient care: primary role (2.5), teaching: secondary focus (1.2), teaching: primary focus (2.3), teaching: student as teacher (0.5), and personal time (6.0, including 3.9 hours of sleep). Few activities were judged to be of high (4.0) or very high (5.0) value. In general, the most highly rate activities were patient care: primary role (3.5–4.0), teaching: primary focus (4.1–4.5), and teaching: student as teacher (4.1–4.2). Overall, students spent only 4.4 hours in activities rated 4.0 or better.

CONCLUSION:  Third-year medical students spend relatively little time in those activities that they value most: discussion of patients with a focus on decision making, teaching activities where students are the primary focus, and personal opportunities to teach. In revising the third-year curriculum, faculty should consider increasing the time allocated to these activities.


BACKGROUND:  Sedentary lifestyle is strongly associated with increased morbidity and mortality, and increasing physical activity is a Healthy People 2,010 objective. African Americans and Hispanics are less likely to engage in recommended amounts of physical activity than whites, a finding not explained in previous studies by socioeconomic differences. Differences in activity limitations and comorbidity might explain the disparities in physical activity level.

METHODS:  Data on 30,216 adults in the 1998 National Health Interview Survey (NHIS) was used to determine the effect of differences in activity limitation and comorbidity on racial differences in physical activity levels. Race/ethnicity was restricted to non Hispanic white (White), non Hispanic black (Black), and Hispanic (Hispanic). Physical activity levels were reported as kcal/kg/day and categorized as “sedentary” (0–1.4), “moderate” (1.5–2.9) or “very active” (3.0+). These were further collapsed to “sedentary” and moderate/very active, or “active”. Activity limitation was defined as difficulty walking a quarter mile or up ten steps. Comorbidity was defined as having hypertension, heart disease, COPD, ESRD, or diabetes. Physical activity levels, activity limitation, and comorbidity were compared by race/ethnicity using χ2 statistics. Multiple logistic regression was used to determine the independent association of physical activity and race/ethnicity, controlling for activity limitation, comorbidity and covariates (age, gender, education, income, employment, perceived health and body mass index). STATA (Version 8.0) was used for hypothesis testing to account for the complex survey design.

RESULTS:  Blacks were most likely to report activity limitations (20%) compared to Hispanics (17%) or Whites (13%; P < .0001). Blacks were most likely to have chronic comorbid conditions (35%) compared to Hispanics (31%) or Whites (22%; P < .0001). Minorities were less likely to be physically active than non-minorities (Black 33%, Hispanic 32%, White 41%; P < .0001). Decreased odds of being physically active persisted for minorities when controlling for activity limitation, comorbidity, and other covariates (OR for Blacks 0.79 [95% CI 0.70–0.88]; Hispanics 0.75 [95% CI 0.68–0.82]).

CONCLUSION:  Lower levels of physical activity among minorities are not explained by differences in activity limitations or comorbidity. Cultural factors associated with decreased physical activity need to be identified to develop successful, targeted interventions to increase physical activity in these patient populations.

EFFECT OF INSURANCE AND CHANGES IN INSURANCE ON STROKE AND DEATH.A. Fowler-Brown1; G.M. Corbie-Smith1; J.M. Garrett1; N. Lurie2. 1University of North Carolina at Chapel Hill, Chapel Hill, NC; 2RAND, Arlington, VA. (Tracking ID #116667)

BACKGROUND:  Previous studies have established that the uninsured have reduced access to care. Few prospective studies have examined the effects of lack of coverage or changes in coverage on physiologic outcomes. We examined whether being uninsured or having intermittent insurance was associated with the likelihood of future stroke and death.

METHODS:  We analyzed data from the Atherosclerosis and Risk in Communities (ARIC) study, a prospective cohort study of adult African American and White residents of four communities in the U.S. Participants’ insurance status was assessed during visits every three years, for a total of 4 visits, between 1986 and 1998. Endpoints were ascertained until 2000. We used an adjusted Cox model with insurance status as a time varying variable to estimate the hazard ratios (HR) of being uninsured or having intermittent insurance on rates of stroke and death within 3 years of each visit (adjusted for age, gender, race, income). In addition, we used logistic regression to determine the incremental effect of number of visits with insurance on stroke and death (adjusted for age, income).

RESULTS:  Of the 13,540 participants included in the Cox model, those who reported being uninsured had 1.7 times the rate of stroke (HR 1.69, 95% CI 0.97–2.93) and 1.6 time the rate of death (HR 1.61, 95% CI 0.86–3.01) compared to those who were insured for the comparable period of time. Particpants who reported changes in insurance status suffered stroke at 1.7 time the rate (HR 1.73, 95% CI 1.17–2.59) and death at 2 times the rate (HR 1.97, 95% CI 1.32–2.95) as those who were insured. For the logistic regression model, we used the 11,003 participants with insurance status recorded at each of the four visits. Of these, 87% were insured at all 4 visits, and the remainder were uninsured during at least 1 visit. We found a dose response relationship with insurance status, those with fewer visits insured had a higher odds of stroke compared to persons who were insured at all 4 visits: reported insured at no visits-OR 3.00 (95% CI 1.34, 6.68), insured at 1 visit-OR 2.44 (95% CI 1.18, 5.04); insured at 2 visits-OR 1.77 (95% CI 0.96, 3.25), insured at 3 visits-OR 1.4 (95% CI 0.90, 2.20).

CONCLUSION:  Persons with no insurance or intermittant insurance suffered strokes and death at a higher rate than those with stable insurance. In addition, there was a dose response relationship between insurance and stroke, with fewer visits insured associated with higher adjusted odds of stroke.


BACKGROUND:  To treat patients with medically unexplained symptoms (MUS), data support only short-term specialty care. Primary care clinicians have no proven, long-term treatment. We hypothesized that an intensive 12-month intervention by primary care clinicians would produce improvement in the common, high-utilizing MUS patient.

METHODS:  From a staff model HMO, we identified high-utilizing patients from the information system and, among these, we then identified primary MUS patients using a systematic chart rating procedure conducted by three blinded senior medical residents. Four nurse practitioners (NP) received 84 hours of initial training and then conducted the intervention as primary caretakers, receiving supervision individually and in groups. NPs primarily relied upon an evidence-based method for the provider-patient relationship, and they integrated it with cognitive-behavioral and pharmacological principles proven effective in specialty settings. Patients had 12 scheduled visits and phone calls as well as additional visits when necessary. Collected at baseline, 6, and 12 months, the primary endpoint was a ≥4-point improvement in the Mental Component Summary of the SF-36 at 12 months. This is a clinically significant psychological change and occurs, for example, following a successful aortic/mitral valve replacement. We also collected measures of physical function, depression, anxiety, and psychosomatic symptoms. Our evaluation used intent to treat and logistic regression; variables associated with a ≥4-point improvement in MCS in univariable testing (P ≤ .20) were entered into a model in which we retained variables significant at P ≤ .10.

RESULTS:  Rater agreement with gold standard for entry into study (on primary MUS status) was 97.6%. Of 206 subjects entered, 200 completed the study (97% retention). Patients averaged 13.6 visits in the year preceding study, 79.1% were females, and the average age was 47.7 years. Forty-eight treatment and 34 control patients improved at 12 months (OR = 1.92, 95% CI: 1.08–3.40; P = .025). The relative benefit was 1.47 (CI: 1.05 to 2.07), and the number needed to treat was 6.4 (95% CI: 0.89 to 11.89). In addition to treatment (P = .019), the following baseline measures predicted improvement: severe mental dysfunction (P < .001), severe body pain (P = .039), non-severe physical dysfunction (P = .003), and advanced education (P = .022); c-statistic = 0.76. There were no adverse effects of treatment.

CONCLUSION:  A long-term, relationship-based intervention by primary care clinicians led to clinically significant improvement in MUS patients. To our knowledge, it provides the first evidence-based treatment guidelines for primary care clinicians.

EFFECTIVENESS OF DIRECT TO CONSUMER ADVERTISING AMONG THE ALREADY MEDICATED.N. Dieringer1; R.I. Shorr2. 1Methodist Healthcare, Memphis, TN; 2University of Tennessee, Memphis, TN. (Tracking ID #117439)

BACKGROUND:  Direct to consumer (DTC) marketing of pharmaceuticals is controversial, yet effective. Although previous surveys have identified demographic characteristics associated with responsiveness to DTC, little is known relating patterns of medication use to patient receptiveness to DTC.

METHODS:  We conducted a secondary analysis of data collected in national telephone survey on knowledge of and attitudes toward DTC advertisements. The survey of 1,081 U.S. adults (response rate = 65%) was conducted by the Food and Drug Administration (FDA). Our primary question of interest was the response to the item: “Has an advertisement for a prescription drug ever caused you to ask a doctor about a medical condition or illness of your own that you had not talked to a doctor about before?” Patients reported number of prescribed medicines taken as well as demographic and personal health information.

RESULTS:  771 (71%) respondents reported exposure, primarily through television, to at least one DTC prescription drug advertisement in the three months prior to the survey. Of these 688 (89%) had seen in a physician in the three months prior to the survey and 195 (25%) stated that DTC motivated them to ask their physicians about a health problem. Only 7% respondents taking no prescription medications reported that DTC advertisements motivated them, whereas 45% of respondents taking 9 or more prescription medications were motivated by DTC to discuss health problems with their doctors (p trend <.0001), see figure. This trend remained significant (p trend .003) even when controlling for age, gender, race, educational attainment, income, and self-reported health status. There was no relationship between the number of over the counter (OTC) medications taken and the propensity to discuss health related problems in response to DTC advertisements (P = .4).

CONCLUSION:  There is a strong linear relationship between the number of prescribed medications taken and the influence of DTC advertisements to prompt discussions with physicians. Whether this results in improved health outcomes is unknown.

inline image

EFFECTS OF DUCTAL CARCINOMA IN SITU ON QUALITY OF LIFE: RESULTS FROM THE NURSES HEALTH STUDY.L. Nekhlyudov1; C.H. Kroenke2; M.D. Holmes2; G.A. Colditz2. 1Harvard Medical School/Harvard Pilgrim Health Care, Boston, MA; 2Brigham and Women's Hospital, Boston, MA. (Tracking ID #116806)

BACKGROUND:  The incidence of ductal carcinoma in situ (DCIS) of the breast has been rising, attributable mostly to screening mammography, and now accounts for about 20% of breast cancer diagnoses. Some have expressed concern with the possibility of overdiagnosis and the subsequent adverse effects on quality of life. However, the effect of DCIS on a woman's quality of life is not known. We therefore undertook a study to determine the quality of life in women before and after a diagnosis of DCIS.

METHODS:  The study included 149,398 women from the Nurses Health Study I (recruited in 1976) and Nurses Health Study II (recruited in 1989) who completed the Short Form 36 (SF-36), a health-related quality of life (HRQoL) assessment, in 1992–3 and 1996–7. All women were free of cancer at the time of entry; 433 cases of DCIS were diagnosed between 1992 and 1997. Analyses were conducted using PROC GLM linear regression, adjusting for age, treatment type and/or patient characteristics.

RESULTS:  HRQoL scores at baseline and 4-year follow-up were similar among women with and without DCIS. However, women who completed the SF-36 surveys within 6 months of diagnosis of DCIS experienced clinically (though not statistically) significant declines (~5 points) in four HRQoL domains including role limitations due to physical problems, bodily pain, vitality and social functioning but not in physical functioning or mental health. Declines in HRQoL disappeared within one year of diagnosis. There were no differences in HRQoL by type of treatment received.

CONCLUSION:  Women diagnosed with DCIS may experience clinically significant declines in health-related quality of life shortly after diagnosis but rebound quickly and experience no long-term effects. This information may be useful to primary care providers counseling women about DCIS prior to screening mammography.

EFFECTS OF LEGCROSSING AND SKELETAL MUSCLE TENSING IN PATIENTS WITH VASOVAGAL SYNCOPE.N. Van Dijk1; I.J. De Bruin1; M. Linzer2; J. Van Lieshout1; W. Wieling1. 1Academic Medical Center, Amsterdam, ; 2University of Wisconsin-Madison, Madison, WI. (Tracking ID #116375)

BACKGROUND:  Physical countermaneuvers can be applied to abort or delay an impending vasovagal faint. This easy, non-invasive and cheap method would however be more beneficial as a preventive measure or to combat pre-syncope. We hypothesized that in patients with recurrent vasovagal syncope, legcrossing produces a rise in cardiac output (CO) and thereby in blood pressure (BP) with an additional rise in BP by skeletal muscle tensing.

METHODS:  Eighty-eight patients applied legcrossing after a 5 minute free standing period. Fifty-four of these patients also applied tensing of leg- and abdominal muscles. Hemodynamic effects were monitored using a non-invasive bloodpressure monitor (Finapres (R)).

RESULTS:  Legcrossing produced a significant rise in CO (+9.5%; P < .01) and thereby in mean arterial pressure (MAP) (+3.3 %; P < .01). Skeletal muscle tensing produced an additional increase in CO (+8.3 %; P < .01) and MAP (+7.8%; P < .01). These effects were larger in the elderly. The relative changes in hemodynamic variables due to legcrossing and muscle tensing from the standing position are displayed in the table below.

CONCLUSION:  Legcrossing and skeletal muscle tensing produce a rise in BP in patients prone to vasovagal syncope. Leg-crossing, as an easy, non-invasive and cheap method could be advised for all patients prone to vasovagal syncope as a preventive measure. Additional muscle tensing could be used to combat syncopal symptoms.

Table 12. Hemodynamic effects of legcrossing and muscle tensing
n = 54StandingLegcrossing (% change from baseline)Muscle tensing (% change from baseline)
SBP126.1 (15.9)104.8%*112.8%*
MAP 90.8 (11.4)103.5%*111.3%*
DBP 74.9 (10.6)101.6%*109.2%*
PP 51.2 (10.8)109.4%*118.0%*
HR 86.3 (15.7) 99.0%104.3%*


BACKGROUND:  In January 2003, people covered by the Oregon Health Plan's (OHP) Medically Needy (MN) program lost their benefits due to state budget shortfalls. The MN program is a federally-matched optional Medicaid program in which states may chose to provide Medicaid coverage and/or Medicare premium assistance to certain groups with significant health care needs not otherwise eligible for Medicaid. This population is not well characterized either nationally or locally. In Oregon the MN program mainly provided prescription drug benefits. The primary objective was to determine how the loss of benefits has affected this population's health. Secondary objectives include investigating how this population is getting their prescription medications and changes in use of prescriptions and health care.

METHODS:  A 49-question survey instrument was created to collect information about this population including demographic information, health insurance coverage, health conditions, access to care, prescription drug use and utilization of health care. A data collection contractor telephoned a statewide, random sample of 1,500 people who were enrolled in the MN program. Medication use during the MN program was obtained from administrative data and current medication use was obtained from the interview. Results of this survey were compared to the Medicare Health Outcomes Survey and the National Health Interview Survey.

RESULTS:  The 439 completed surveys included 36% men and 64% women from age 21–91 with 70–75% at less than 133% of the federal poverty level. 2/3 of respondents rate their current health as poor or fair and compared to last year 44% rate their health as worse and 39% about the same. In the six months since the MN program ended 61% have skipped doses of a medication and 64% have gone without filling a prescription. In order to pay for medications 60% of respondents have cut back on their food budget, 49% have skipped paying other bills and 47% have borrowed money. Overall, there was no significant difference in ER visits or hospitalizations by self-report comparing the six months before losing the MN program and the six months after.

CONCLUSION:  The MN program provided coverage for a low-income, chronically ill population. Since its termination, there have been significant changes in prescription drug use and financial impact in their daily lives. This population is sicker with worse perceived health status than the general population over 65 and people eligible for both Medicare and Medicaid. As states make program changes, especially to prescription drug benefits, the effects on vulnerable populations must be considered.

EFFICACY OF BILATERAL PROPHYLACTIC MASTECTOMY IN WOMEN AT MODERATE RISK FOR BREAST CANCER. A.M. Geiger1; M.B. Barton2; O. Yu3; L.J. Herrinton4; W.E. Barlow5; E.L. Harris6; J.G. Elmore5; S. Rolnick7; S.W. Fletcher2. 1Southern California Permanente Medical Group, Pasadena, CA; 2Harvard University, Boston, MA; 3Group Health Cooperative, Seattle, WA; 4Permanente Medical Group, Oakland, CA; 5University of Washington, Seattle, WA; 6Kaiser Permanente Northwest, Portland, OR; 7Health Partners Research Foundation, Minneapolis, MN. (Tracking ID #116410)

BACKGROUND:  Reports from referral clinics suggest bilateral prophylactic mastectomy reduces breast cancer incidence by 90% or more in women at high risk for breast cancer (those with a family history of breast cancer and/or a deleterious BRCA mutation) but whether this is true in community practices is unknown. We determined the efficacy of bilateral prophylactic mastectomy in reducing breast cancer incidence among women at moderate risk for breast cancer in community-based practices.

METHODS:  Eligible women were aged 18 to 80 years; enrolled in one of six health maintenance organizations during 1979 to 1998; and had one or more breast cancer risk factors (family history of breast cancer, history of atypical hyperplasia, or at least one benign breast biopsy). Using automated data and medical records we identified 276 women with bilateral prophylactic mastectomy and a random comparison sample of 196 women representing an underlying cohort of 666,800 women with one or more breast cancer risk factors, and without prophylactic mastectomy. Breast cancer outcomes were determined from cancer registries and ambulatory medical record data. Death information was obtained from medical records and state death records.

RESULTS:  Breast cancer developed in one woman (0.4%) after bilateral prophylactic mastectomy compared to an estimated 26,800 (4%) breast cancer diagnoses among women without prophylactic mastectomy. Stratifying by birth year and organization the hazard ratio for breast cancer occurrence after bilateral prophylactic mastectomy was 0.005 (95% confidence interval 0.001 to 0.044). No woman with bilateral prophylactic mastectomy died of breast cancer compared to an estimated 0.2% of the women without prophylactic mastectomy.

CONCLUSION:  Bilateral prophylactic mastectomy substantially reduces breast cancer incidence in women at moderate risk for breast cancer in community-based practices, but the absolute risk of breast cancer incidence and death in women at this risk level was found to be low.


BACKGROUND:  Emergency contraception can prevent unintended pregnancies. We studied the effect of a single educational intervention on emergency contraception knowledge, attitudes, and practice patterns of primary care providers.

METHODS:  Prior to attending one of 7 lectures on emergency contraception at different locations in Massachusetts between August and December 2002, primary care providers completed a 38-item questionnaire assessing emergency contraception knowledge, attitudes, and practice patterns. A follow-up survey was mailed to the participants 6 months following the lecture program. Knowledge, attitudes, and practice patterns regarding emergency contraception were compared before and after the intervention.

RESULTS:  Fifty-six follow-up surveys were returned from the 72 eligible participants (response rate = 78%). There were 50 physicians, 4 advanced practice nurses, and 2 physician assistants (48% General Internal Medicine, 34% Family Medicine, 9% Obstetrics Gynecology, and 9% Pediatrics/Adolescent Medicine). Baseline knowledge about emergency contraception was high, and remained high at 6-month follow-up. Compared with baseline, providers were more likely to agree that advance prescriptions should be given (P = .0006), disagree that the number of times emergency contraception is dispensed to a patient should be restricted (P = .04), and disagree that repeated use poses health risks (P < .0001). The proportion of providers who initiated counseling about emergency contraception at least sometimes during routine visits increased from 36% to 54% (P = .057) and the proportion of providers who had ever written an advance prescription for emergency contraception increased from 18% to 41% (P = .007).

CONCLUSION:  We found that a single educational lecture to primary care providers was associated with increased counseling frequency and writing of advance prescriptions for emergency contraception. Such inexpensive educational interventions can be easily instituted to primary care providers in order to increase access to this effective contraceptive method.

EMERGENCY PREPAREDNESS ON U.S. COLLEGE AND UNIVERSITY CAMPUSES.B.A. Johnson1; V.R. Koppaka2; C.S. Cors1. 1Virginia Commonwealth University, Richmond, VA; 2Centers for Disease Control and Prevention (CDC), Atlanta, GA. (Tracking ID #117064)

BACKGROUND:  Every year, more than 500,000 students enter the US to study on college and university campuses. In 2002, a worldwide outbreak of Severe Acute Respiratory Syndrome (SARS) heightened concerns about the preparedness of US campuses to deal with a deadly communicable disease outbreak.

METHODS:  To assess level of preparedness, a stratified random sample of the 3,664 US 2-year and 4-year colleges and universities was surveyed in the latter half of 2002.

RESULTS:  Of the 1000 schools surveyed, 474 (47.4%) responded to the self-administered questionnaire. Of those that responded, 302/474 (63.7%) of schools had a plan for handling an outbreak of a deadly communicable disease among students. Institutions that were more likely to have an emergency preparedness plan were those that offered a 4-year curriculum (P < .001), were affiliated with the American College Health Association (ACHA) (P < .001), had a Student Health Service (SHS) (P < .001), and offered on-campus housing (P < .001). Of schools with emergency preparedness plans, 73.5% (222/302) were actively collaborating with their local health departments. Collaboration with local health departments was more likely to involve planning for a SARS outbreak than a smallpox outbreak (47% vs. 30.1%, P < .001). Of schools actively involved in emergency planning, 158/302 (52.3%) had SARS protocols in place in the SHS, 115/302 (38.1%) had N-95 respirator masks available in the SHS, and 24/302 (7.9%) had had experience with evaluating a SARS suspect. Only 201/474 (42.4%) of campuses that responded could identify a non-hospital site for isolation of a student with a communicable disease. Among those with an emergency plan, only 154/302 (51%) had an identified site for isolation. Among the schools that responded, 267 (56.3%) planned to use an email network to quickly communicate a disease alert to students, faculty, and staff.

CONCLUSION:  While many US colleges and universities have been actively engaged in planning, additional efforts are required to prepare them to respond to an outbreak of a deadly communicable disease.


BACKGROUND:  Major depressive disorder (MDD) is still a significant cause of morbidity and mortality despite treatment advances. Existing therapies are limited in providing full remission. While emotional symptoms such as depressed mood and loss of interest are traditionally considered to constitute the core of depression, the importance of resolving painful physical symptoms such as back pain, abdominal pain, and musculoskeletal pain in patients with MDD is becoming increasingly appreciated. We report the efficacy, tolerability, and safety of duloxetine, a balanced and potent dual reuptake inhibitor of both serotonin and norepinephrine, in the treatment of MDD and associated physical symptoms.

METHODS:  Efficacy data were pooled from 2 identical, 9-week randomized, double-blind trials of duloxetine 60 mg QD (N = 244) and placebo (N = 251) in the treatment of MDD. Patients were not required to meet a minimum pain threshold at baseline, and the studies were not specifically powered for pain outcomes. The primary efficacy measure was HAMD17 total score. Painful physical symptoms were assessed using visual analog scales (VAS). The average baseline score for overall pain was 26 on the 0 (no pain) to 100 (as severe as I can imagine) VAS scale. Safety and tolerability were determined by monitoring discontinuation rates, spontaneous and solicited [Somatic Symptom Inventory (SSI)] adverse events and vital signs.

RESULTS:  Duloxetine was significantly superior to placebo (P < .001) in reducing HAMD17 total scores, starting at Week 2. The estimated probability of response and remission were 63% and 43%, respectively, for duloxetine-treated patients compared with 35% and 21% for placebo-treated patients (P < .001). Duloxetine-treated patients demonstrated significantly greater improvement in overall pain severity (P = .016) at Week 9 compared with patients receiving placebo. Discontinuation due to adverse events for duloxetine-treated patients (13.1%) was significantly higher than placebo-treated patients (13.1%, 2.5%; P = .002) in this study but comparable with the rates reported for SSRIs in other studies. Nausea, dry mouth, and headache were the most common spontaneously reported adverse events; no significant incidence of hypertension was seen. Duloxetine produced significantly greater gastrointestinal distress compared to placebo within the first week of treatment, returned to baseline by Week 2, and was not significantly different than placebo for the rest of the study.

CONCLUSION:  In these studies, duloxetine at 60 mg once daily, the expected starting and therapeutic dose, demonstrated robust efficacy and was well-tolerated in treating both the emotional and painful physical symptoms in depressed patients.

END-OF-LIFE CARE EXPERIENCES AND ATTITUDES DURING THE FIRST TWO YEARS OF MEDICAL SCHOOL.M.W. Rabow1; C.S. Hodgson1. 1University of California, San Francisco, San Francisco, CA. (Tracking ID #117343)

BACKGROUND:  Inadequate medical school end-of-life care (EOLC) education has been identified as one of the causes of poor EOLC in the United States. While some curricula are being improved, the EOLC experiences and attitudes of early medical students are not well known. This study describes student EOLC attitudes and experiences prior to beginning clinical clerkships.

METHODS:  At the end of their 2nd year, the UCSF Class of 2005 was given a confidential, self-administered questionnaire about their EOLC attitudes and experiences. Response options included 1–5 Likert scale and yes/no responses. The questionnaire was distributed by faculty at the end of a required small group session. Questionnaire responses were double-entered into an Excel database and analyzed using SAS statistical software. Simple frequencies and Chi-square statistics were calculated, with significance taken as P < .05. The study received Institutional Review Board approval.

RESULTS:  Completed surveys were returned by 127 of 141 students (response rate = 90.1%). Mean age of students was 25.5 years, with 56.7% female, 51.2% white, 28.3% Asian, 8.7% Hispanic, and 1.6% African-American. There were significant relationships between student gender and the belief that all physicians should become competent in EOLC (P = .038), between ethnicity and belief in the importance of EOLC (P < .001), and between student religiosity and worry about what medical school was doing to their humanity (P = .05). Fifty-three students (41.7%) reported having no medical school experiences with EOLC. An additional 27 students (21.3%) had only 1 experience. Ninety-seven students (76.4%) reported having experienced the death of someone close to them personally and 27 (28.1%) provided direct care to that loved one. Less than one-half of students observed supervisors break bad news (48%), manage physical symptoms at the EOL (44%), manage emotional suffering at the EOL (28%), or lead an advance directive discussion (22%). Less than 7% of students actually helped do any of these EOLC tasks. When a patient was dying or had died, clinical preceptors only sometimes discussed it with students. In these discussions, medical details were discussed 82.4% of the time; the preceptor's emotions 29.6%. Students reported that faculty response to student EOLC emotions usually involved being supportive (64.5%) but also included noticing but avoiding (16.1%), seeming oblivious (12.9%) or being critical (1.6%).

CONCLUSION:  Early medical students have relatively few medical school experiences with end-of-life care. These experiences are almost entirely observational. The frequency of discussion about emotions around EOLC is limited and faculty response to student emotions can be negative.

EPIDEMIOLOGY OF ANEMIA IN THE ELDERLY.R.S. Eisenstaedt1; R.C. Woodman2; H.G. Klein3; L. Ferrucci4; J.M. Guralnik4. 1Temple University, Philadelphia, PA; 2Ortho Biotech, L.P., Bridgewater, NJ; 3National Institutes of Health, Bethesda, MD; 4National Institute on Aging, Bethesda, MD. (Tracking ID #115914)

BACKGROUND:  While anemia seems common in the elderly and is reported to be an independent predictor of functional decline and increased mortality, prior epidemiological studies have been biased. We examined a nationally representative database of community dwelling persons with no upper age limit and adequate sample size of the oldest subset (>85); additional interview data and lab testing helped classify the etiology of anemia.

METHODS:  Data were obtained from phase 1 & 2 of the 3rd National Health and Nutrition Examination Survey (NHANES III), conducted 1988–94. Participants were sampled using a stratified, multistage probability design. Anemia and nutritional deficiencies (NutDef) were defined by WHO criteria. Creatinine clearance (CC) was calcuated, and anemia of chronic kidney disease (CKD) defined by CC < 30ml/min. Anemia of chronic disease (ACD) was defined as low serum iron without iron deficiency; unexplained anemia (UA) was anemia excluded by the above. Prevalence and distribution was estimated by appropriate sampling weights, and small subgroups were compared directly using logistic models adjusted for age, race, and ethnicity.

RESULTS:  5,252 persons age >65 were interviewed, 86% examined; blood was obtained from 85% of 2,096 persons in phase 2. The prevalence of anemia = 10.6% (11% men, 10.2% women) and approximately doubles for each decade > age 65. The prevalence among blacks, 27.8%, was 3x that in whites. Most anemia was mild, with Hgb below 10 gm in <1%. 1/3 of anemia was from NutDef, 1/3 due to CKD &/or ACD, the remaining 1/3 classisifed as UA. The cause of anemia among blacks was similar, though, among the 1/3 with NutDef, the prevalence of folate, iron and B12 deficiency was 4x, 2x, and no different, respectively, from that in whites.

CONCLUSION:  Anemia in the community based elderly was common, albeit mild, with 1/3 caused by NutDef, easily diagnosed, safely and inexpensively treated. As iron deficiency in this population is often associated with GI bleeding, additional w/u may reveal important pathology, such as polyps or cancer. 1/3 of anemia is caused by CKD or ACD, conditions of absoulute or relative erythropoieitin (Epo) deficiency, though the value of Epo therapy in the elderly with mild anemia remains uncertain. The final 1/3 of anemia is unexplained, though anemia oriented exam and lab evaluation may reveal myelodysplasia or other defined causes, and a portion of anemia may be self-limited. The higher prevalence of anemia in blacks is partly from underlying co-morbidity, such as CKD, known to be increased in that population, but also from NutDef that needs recognition as a public health concern.


BACKGROUND:  Cardiovascular disease is the leading cause of mortality in diabetics. Diabetic mortality rates are higher among African Americans compared to whites, and this disparity has increased in the past decade. Recent evidence suggests that differences in medical care may contribute to ethnic differences in health outcomes.

METHODS:  We conducted an electronic medical record review of 724 diabetic patients cared for in the academic primary-care practice at the University of Chicago in the period January 2000 through June 2003. Medications, comorbidities, lab results, clinic visits and referrals, and demographic data were retrieved.

RESULTS:  Compared to white (W) diabetic patients (n = 86, median age 64), African American (AA) diabetic patients (n = 585, median age 65), had significantly higher rates of documented hypertension, and of therapy with diuretics and calcium blockers. Rates for Hispanic (H) patients (n = 31, median age 67) were generally intermediate between those of AA and W patients. While about 90% of the 3 groups had lipid testing in the prior 2 years, and rates of cholesterol > 240 and LDL > 100 were similar in groups, a documented diagnosis of hyperlipidemia was found in only 40% of AA vs. 57% of W pts (P = .003). Moreover, of patients treated with a statin, only 47% of AA pts achieved an LDL < 100, compared to 71% of W pts (P = .008). Smoking was documented more frequently among AA pts (25 vs.14%) (P = .03). To further explore factors associated with LDL control, the charts of 35 patients with the best LDL values (mean LDL 45) and 35 patients with the highest LDL values (mean 196) were reviewed. Age, weight, race, #of chronic medications, clinic visits, and rates of missed appointments were very similar. About 95% of both groups were prescribed atorvastatin or simvastatin, and duration of therapy and dose intensity were similar. Low LDL patients were more likely to be male (40% vs 6%, P < .01) and to have been referred to cardiology (56% vs. 12%, P < .001), with trends toward less tobacco use, more frequent CAD diagnosis (47 vs. 37%, P = NS), as well as greater and more sustained LDL response to statin therapy.

CONCLUSION:  Despite higher rates of hypertension, tobacco use, and equal ascertainment of lipid values, AA patients were significantly less likely to be diagnosed with a lipid disorder, or to receive adequate statin therapy. Male gender and referral to cardiology are associated with better LDL control. Further efforts are needed to assess medication barriers and adherence, and to learn how to better characterize and reduce cardiac risk in all diabetics.

ETHNIC DIFFERENCES IN GLYCEMIC AND CARDIOVASCULAR RISK FACTOR LEVELS IN US ADULTS WITH DIABETES.L.H. Miller1; S.R. Lipsitz2; S. Natarajan1. 1New York University, New York, NY; 2Medical University of South Carolina, Charleston, SC. (Tracking ID #116924)

BACKGROUND:  Effective management of cardiovascular risk factors and hyperglycemia in patients with diabetes reduces macrovascular and microvascular disease. Lowering cardiovascular risk factors and hyperglycemia is therefore crucial to achieve better outcomes. Despite extensive effort, a large proportion of adults with diabetes do not have risk factors at therapeutic goals. While ethnic differences in risk factor levels have been described, little is known about whether these differences persist.

METHODS:  We analyzed the 1999–2000 National Health and Nutrition Examination Survey to evaluate if ethnic differences in cardiovascular risk factor levels persist in US adults with diabetes (n = 608, population estimate = 14,905,890). Ethnic categories of interest were non-Hispanic whites (NHW), non-Hispanic blacks (NHB) and Mexican-Americans (MA). Outcome measures were hemoglobin A1c (HbA1c) %, systolic and diastolic blood pressure, total cholesterol, LDL cholesterol, HDL cholesterol, triglycerides, body-mass index (BMI) and smoking status. New putative risk factors (C-reactive protein, homocysteine and fibrinogen)were also evaluated. Multivariate logistic and linear regression were performed using SUDAAN to evaluate if significant (P < .05) ethnic differences in risk factor levels exist while adjusting for age, sex and accounting for the complex sample design.

RESULTS:  The likelihood of having HbA1c levels ≥7% was higher in NHB (odds ratio [OR] 2.11, 95% confidence interval [CI] 1.08–4.14) and MA (OR 2.33, CI 1.24–4.38) than NHW. The mean HbA1c among NHW (7.48%) was significantly lower than the mean values in NHB (8.31%) or MA (8.26%). The mean systolic blood pressure (mmHg) was not significantly different in NHW (132.5) compared to NHB (137.1, P = .1) or MA (137.6, P = .08). HDL (in mg/dl) was significantly higher in NHB (50.5) than in NHW (42.4), or MA (43.6). Triglycerides were higher in MA (307.4 mg/dl) than NHB (177.9 mg/dl). C-reactive protein (in mg/dl) was also significantly higher in NHB (1.09), MA (0.81), than in NHW (0.59). Fibrinogen was significantly higher in NHB (424.5 mg/dl) than NHW (382.3 mg/dl). Diastolic blood pressure, total cholesterol, LDL, BMI and smoking status were not significantly different between racial and ethnic groups.

CONCLUSION:  Ethnic differences persist in US adults with diabetes, with NHW generally having better risk factor profiles than NHB or MA. Because achieving better risk factor control may eliminate ethnic disparities in cardiovascular outcomes, more effort should be focused on elucidating the reasons for such ethnic disparities and developing interventions to correct them.

EVALUATION OF AN INFORMATION-TECHNOLOGY INTERVENTION TO PROMOTE DIETARY BEHAVIOR CHANGE IN DYSLIPIDEMIA PATIENTS.R.H. Friedman1; K. Glanz2; H. Tim3; K. Heather3; M. Barbara3; D.C. Mitchell4; S. Charlie3; T. Sharon5. 1Boston University/Boston Medical Center, Boston, MA; 2University of Hawaii, Honolulu, HI; 3Boston University, Boston, MA; 4Pennsylvania State University, University Park, PA; 5New England Research Institute, Inc., Boston, MA. (Tracking ID #117128)

BACKGROUND:  There has been insufficient use of labor saving information technical (IT)-based aides in primary care practice other than the use of electronic medical records. We developed and evaluated an automated, telephone-based, multi-contact, patient self-care management program for decreasing saturated fat intake and increasing fruits and vegetable consumption in patients being treated for hypercholesterolemia in a primary care setting.

METHODS:  We randomized 228 primary care hypercholesterolemia patients (total cholesterol ≥240 mg/dl) to an automated intervention (n = 121) consisting of weekly automated telephone education and counseling sessions over six months versus an attention placebo control condition (n = 107) that provided weekly general health information on selected topics for six months using the same technology. We measured dietary behavior and nutrient intake by three 24-hour diet recalls at baseline and six months using the University of Minnesota Nutrition Data System. For each dietary behavior (e.g., consumption of fruits, etc.) and each nutrient (% calories from saturated fat, etc.), we compared the adjusted least square mean at six months using ANCOVA controlling for gender and baseline value of the variable. Blood samples for serum lipids where drawn at the two timepoints and where processed together and statistically analyzed using the ANCOVA method.

RESULTS:  The subjects were 58.0 ± 13.2 years of age, 68% female, 23% minority, and BMI 27.2 ± 3.6. At follow-up, intervention group subjects had higher fruit intake (1.7 v. 1.3 servings/d, P = .003), no difference in vegetable intake, less red meat consumption (0.3 v. 0.5 servings/d, P < .0001), and less use of oils/fats (3.6 v. 4.6 servings/d, P = .02). Nutrient intake was lower in the intervention for total fat (54 v. 66 g/d, P = .0005), percent calories from saturated fat (8.9 v. 10.9, P < .0001), and total calories (1625 v. 1785 kal/d, P = .01). There were no significant differences in the serum lipid values.

CONCLUSION:  A totally automated telephone-based intervention significantly improved dietary behavior and nutrient intake in primary care patients with hypercholesterolemia, however, there was no effect on serum lipids, a result that was possibly due to insufficient time to observe a dietary effect on serum lipids and/or the use of lipid-lowering medications by the physicians. These findings, nevertheless, demonstrate the efficacy of an automated program for changing important dietary behaviors in a primary care practice setting.

EVIDENCE REVIEW: DIAGNOSIS AND MANAGEMENT OF SUBCLINICAL THYROID DISEASE.E. Ortiz1; M. Surks2; G. Daniels3; C. Sawin4; N. Col5. 1AHRQ, Rockville, MD; 2Montefiore Med Ctr and Albert Einstein Coll of Med, Bronx, NY; 3Massachusetts General Hospital, Boston, MA; 4Veterans Administration, Washington, DC; 5Brigham and Women's Hospital, Boston, MA. (Tracking ID #116724)

BACKGROUND:  Subclinical hypo- and hyperthyroidism are common conditions, with prevalences of 4–8.5% and 2%, respectively. Improvements in thyroid-stimulating hormone (TSH) assays have resulted in increased detection of subclinical disease, and controversy surrounds the definition, clinical importance, and necessity of prompt diagnosis and treatment. Our objective was to develop evidence-based recommendations for diagnosing and managing subclinical thyroid disease.

METHODS:  Representatives from 3 endocrinology organizations planned a consensus development conference and convened a panel of 13 experts in thyroid diseases, primary care, epidemiology, evidence-based medicine, cardiology, and other clinical and research domains. A systematic literature review was conducted on all English language articles from 1995–2002; seminal articles published before 1995 were also included. 195 articles were reviewed. Panel members also attended a 3-day conference, where 12 invited experts presented on various issues relating to subclinical thyroid disease. Panelists then sequestered themselves to sort through the evidence and develop recommendations.

RESULTS:  There is insufficient evidence to recommend population-based screening for subclinical hypo- or hyperthyroidism. Among patients with subclinical hypothyroidism, there is no single TSH level at which clinical action is always indicated or contraindicated. The evidence does not support routine treatment when serum TSH is <10 mU/L. As the TSH concentration rises above 10 mU/L, the rate of progression to overt hypothyroidism is increased, and treatment may prevent the manifestations of hypothyroidism and ameliorate increases in cholesterol. Levothyroxine therapy is a reasonable consideration for these patients, although evidence that therapy will reduce symptoms or improve outcomes is inconclusive. For patients with subclinical hyperthyroidism whose TSH is 0.1–0.45 mU/L, the panel recommends against routine treatment as there is insufficient evidence to establish a clear association between this degree of hyperthyroidism and adverse clinical outcomes, including atrial fibrillation. If the TSH concentration is <0.1 mU/L, treatment should be considered for patients with Graves’ or nodular thyroid disease; over age 60; with symptoms suggestive of hyperthyroidism; or at risk for heart disease, osteopenia, or osteoporosis.

CONCLUSION:  There is insufficient evidence to recommend routine screening or treatment for subclinical thyroid disease. Given the paucity of evidence, clinicians are encouraged to use the evidence, together with clinical judgment and patient preferences, to make informed and shared decisions together with their patients.

EVOLVING CONCEPTIONS OF INDIVIDUALIZATION OF CARE IN OLDER TYPE 2 DIABETES PATIENTS.E.S. Huang1; G. Sachs1; M.H. Chin1. 1University of Chicago, Chicago, IL. (Tracking ID #116084)

BACKGROUND:  The care of type 2 diabetes patients over 65 years of age is controversial, with some guidelines recommending that older patients be treated by general population standards, while others recommend that risk factor goals be individualized. We examine the policy and clinical implications of diabetes practice guidelines emphasizing individualization for a cohort of older patients.

METHODS:  From 12/2000–1/2003, we recruited patients, at least 65 years of age, living with type 2 diabetes, attending the University of Chicago clinics. Patients were asked about their preferences regarding complications and treatments of diabetes, as well as health status (e.g. self-care behaviors and functional status). Physicians of patients were asked to provide an assessment of health status (e.g. life expectancy(LE)). Medical records were abstracted for comorbidities, risk factor control, and medications. Outcomes of interest included the proportion of patients with ideal risk factor control under general population and individualized guidelines. We also compared current intensity of care for patients with and without different indicators of frailty: LE ≤ 5 years, age ≤ 85, 4–6 activities of daily living requirements, modified Charlson Comorbidity Index Score ≥ 5, and 3 microvascular complications.

RESULTS:  The mean age was 74 and the mean duration of diabetes was 13.4 years (N = 556). The majority were female (62%) and African-American (79%). Under general population guidelines, only 24–42% had adequate glucose control (glycosylated hemoglobin (HbA1C) ≤ 6.5% or <7%). Under individualized guidelines, less intense targets would be applied to frail patients (e.g. HbA1C ≤ 8%). For each frailty definition, guidelines emphasizing individualization increased the proportion of frail patients with adequate control (50–84%). For example, in patients with LE ≤ 5 years, the proportion with ideal glucose control shifted from 35% to 70%. Frail and non-frail patients had similar risk factor levels, except in the case of microvascular complications where glucose control was worse in patients with complications than in those without. Frail patients were generally treated with a larger number of diabetes-related medications. They were more likely to receive insulin (e.g. 40% in LE ≤ 5 years vs. 21% in LE >10 years, P < .05) and aspirin (e.g. 61% in LE ≤ 5 years vs. 41% in LE >10 years, P < .05).

CONCLUSION:  Guidelines advocating individualization of care dramatically alter the definition of quality of care for frail older diabetes patients. Current patterns of risk factor levels and medication use indicate that physicians and patients may be defining treatment plans by responding to risk factor levels or the presence of complications rather than actively predetermining the goals of care. Future efforts are needed to help physicians identify frail older diabetes patients and tailor preventive therapies.

EXERCISE COUNSELING AMONG U.S.- AND FOREIGN-BORN ADULTS.M.S. Goel1; E.P. McCarthy1; R.S. Phillips1; C.C. Wee1. 1Beth Israel Deaconess Medical Center, Boston, MA. (Tracking ID #116503)

BACKGROUND:  Guidelines recommend physicians counsel all patients about exercise. The foreign-born are the fastest growing population in the U.S. and have substantial rates of obesity, yet little is known about the influence of birthplace on exercise counseling.

METHODS:  We examined 25,365 respondents from the 2000 National Health Interview Survey reporting a provider visit in the past year. We compared exercise counseling among U.S.-born whites (n = 17,010), blacks (n = 3,369), Hispanics (n = 1,642), and Asians (n = 129), and foreign-born (FB) whites (n = 801), blacks (n = 326), Hispanics (n = 2,128), and Asians (n = 514). Multivariable logistic regression models adjusted for demographic factors, smoking, self-reported health, comorbid illnesses, body mass index (BMI), physical activity, hospital stays in past year, and access to care. All analyses used SUDAAN and results were weighted to reflect national population estimates.

RESULTS:  Overall, the mean age was 46 years, 12% were foreign-born, 22% obese (BMI 30+ kg/m2), and 40% sedentary. Compared with the U.S.-born, foreign-born adults were less likely to be obese (17% vs. 23%) and generally more likely to be sedentary. U.S.- and foreign-born blacks were less likely to be counseled after adjustment (Table).

CONCLUSION:  Rates of exercise counseling are low, but blacks, regardless of birthplace, are less likely to receive exercise counseling. Given the high prevalence of obesity and sedentary behavior among blacks, interventions to improve exercise counseling are needed.

Table 13. Obesity, Activity Level, and Exercise Counseling By Race/Ethnicity and Birthplace
 Obesity (%)Sedentary (%)Counseled (%)Odds Ratio (95%CI)
US White213623Reference
US Black3353220.8 (0.7–0.9)
US Hispanic3245241.0 (0.9–1.2)
US Asian1326251.1 (0.6–2.1)
FB White1843221.0 (0.8–1.2)
FB Black2240170.7 (0.5–0.9)
FB Hispanic2159170.9 (0.7–1.0)
FB Asian 738201.1 (0.8–1.4)

EXPECTATIONS AND OUTCOMES OF GASTRIC FEEDING TUBES.T.S. Carey1; L.C. Hanson1; J.M. Garrett1; C. Lewis1; N. Phifer2; C.E. Cox3; A.M. Jackman1. 1University of North Carolina at Chapel Hill, Chapel Hill, NC; 2Moses Cone Hosptial, Greensboro, NC; 3Duke University, Durham, NC. (Tracking ID #116139)

BACKGROUND:  Background: The use of gastric feeding tubes has increased over the past decade. Substantial doubt remains regarding the efficacy of the procedure in patients with serious illnesses. We compared expected outcomes by patients’ decision-making surrogates with actual outcomes.

METHODS:  Methods: Prospective cohort study. We enrolled consecutive patients > 17 yrs of age who had received placement of an initial gastric feeding tube at one university and one community North Carolina hospitals, excluding those with metastatic non-ENT cancer or trauma. Surrogate (generally spouse or adult child) respondents were used since almost all patients were either incompetent or unable to verbalize. We assessed ADL's ; expectations of benefit from the feeding tube such as decreased risk of pneumonia; complications; and expectations for recovery. Follow-up surveys were conducted at 3 months and at 6 months for patients who were alive and still had a feeding tube in place. Death certificate review was used to ascertain cause of death.

RESULTS:  Results: 288 patients were enrolled. Mean age was 65 yrs, 54% male, and 61% white. The patients were generally poor, with 50% of incomes < $10,000/year, 36% had resided in a nursing home prior to the procedure. 30% had a primary diagnosis of stroke, 16% neurodegenerative disorder, 20% head and neck cancer, and 34% other. Surrogate expectations for patient improvement were very high at baseline: 95% expected improvement in nutrition; 87% a better quality of life; 66% decreased risk of pneumonia; 90% to live longer; 56% greater independence; 64% less pain; 79% more comfortable. Expectations remained essentially unchanged when re-queried at 3 and 6 months as to whether expectations were fulfilled. Patient outcomes demonstrated that 30% of patients died by 6 months. At 3 months, 38% were residing in a nursing home and 27% had the gastrostomy removed. Perceived global quality of life was poor at 4.6(0–10 scale). Patients were impaired in most ADLs with little change over time. Medical complications were common: 25% suffered from decubiti at 3 months, 24% had at least one additional episode of treatment for pneumonia since tube feeding initiation.

CONCLUSION:  Conclusion: Surrogates currently have unrealistically high expectations of benefit from gastric tube feeding. Mortality is high while functional status and perceived quality of life is poor and improves little over time. Providers and families need better information about the outcomes of this common procedure to educate them regarding the likelihood of functional improvement.

FACTORS ASSOCIATED WITH BLOOD PRESSURE CONTROL AMONG ADHERENT PATIENTSK. Taneda1; C.L. Bryson1; M.B. McDonell1; S.D. Fihn1. 1HSR&D VA Puget Sound Health Care System, Seattle, WA. (Tracking ID #116318)

BACKGROUND:  Hypertension control is often inadequate even among patients that are adherent to their drug therapy. The characteristics of patients with poor hypertension control who are adherent are not well described. We assessed associations between hypertension control and provider and patient characteristics among “adherent” patients.

METHODS:  As part of the Ambulatory Care Quality Improvement Project involving 7 VA general internal medicine clinics, we collected medication data and self-reported comorbidities during a 2-year study period. We defined adherence according to self-report and a pharmacy record-based measure (ReComp). Blood pressure (BP) was measured at clinic visits within 90 days of the self-report. Among those adherent patients, we assessed association between BP control (<140/90) and characteristics of providers and patients.

RESULTS:  6,768 patients who completed questionnaires and had corresponding BP records were analyzed. 4,950 cases (73%) were found to be adherent. Of these, 39.3% had controlled BP (under 140/90). The relative odds of having controlled BP were associated with patients’ age in decade (OR 0.80, 95% CI 0.75 to 0.86). The variables not associated with BP control were: geographic distance from a clinic, provider's sex, type of provider, the number of hypertensive patients in a provider's panel, duration of care by the same provider, patients’ sex and ethnicity, marital status, education, income, or work status.

CONCLUSION:  BP was not well controlled even among adherent patients. Poor control was associated with older age but not with other characteristics of patients or providers. BP should be controlled more aggressively, particularly among older patients.

FACTORS ASSOCIATED WITH DEPRESSIVE SYMPTOMS IN PATIENTS WITH HIV/AIDS.M.S. Yi1; J.M. Mrus2; T.J. Wade3; M.L. Ho1; R.W. Hornung1; A.H. Peterman4; C.M. Puchalski5; J. Tsevat2. 1University of Cincinnati Medical Center, Cincinnati, OH; 2University of Cincinnati and Veterans Affairs Medical Centers, Cincinnati, OH; 3Brock University, St. Catharines, Ontario; 4Northwestern University, Evanston, IL; 5George Washington University, Washington, DC. (Tracking ID #116651)

BACKGROUND:  Depression has been linked to health status, immune suppression, and mortality in patients with chronic illness. Since depression appears to be a marker for poorer health outcomes among patients with chronic illness, we sought to determine which factors may relate to depressive symptoms in patients with HIV/AIDS.

METHODS:  Patients were recruited from 4 medical centers in 2002, and trained interviewers administered the questionnaires. The level of depressive symptoms was measured with the Center for Epidemiologic Studies Depression Scale-10 (range 0–30; a score of ≥10 indicates significant depressive symptoms [SDS]). Predictor variables included patient socio-demographics, clinical characteristics, risk attitudes, health status (HAT-QoL; HIV Symptom Index), social support, self-esteem, optimism, spirituality (FACIT-SpEx), religiosity (Duke Religion Index), and religious coping (Brief RCOPE). We performed univariate analyses and used logistic regression for multivariable analyses.

RESULTS:  We collected data from 450 subjects. Their mean (SD) age was 43.3 (8.4) years; 86% were male; 45% were white; and their mean CD4 count was 420 (301). Two hundred forty-one (54%) met criteria for SDS. In univariate analyses, having SDS was associated (P < .05) with site of care, being unemployed (66% having SDS if unemployed vs. 42% if employed), being single (58% vs. 43%), not having stable housing (77% vs. 52%), having a high school education or less (63% vs. 47%), having no health insurance (62% vs. 51%), feeling that life was not better since HIV diagnosis (64% vs. 30%), having lower CD4 counts (379 if having SDS vs. 468 if not having SDS), having poorer health status, having more somatic HIV-related symptoms, having less social support, participating less frequently in organized religion, having lower spiritual well-being, coping through religion poorly, having lower self-esteem, being less optimistic, and being less risk-seeking. In multivariable analyses, having SDS was associated with having lower life satisfaction, greater health worries, more somatic HIV-related symptoms, less social support, being less optimistic, and lower spiritual well-being (c statistic = 0.92).

CONCLUSION:  A majority of patients with HIV/AIDS reported SDS. Less social support and optimism, poorer health status, and poorer spiritual well-being were related to SDS in patients with HIV/AIDS. Because patients with HIV/AIDS have a high rate of SDS, and because depression has been linked to negative health outcomes, unhealthy behaviors, and non-adherence to medical regimens, further study of the determinants of SDS and ways to optimize mental health is essential for patients living with HIV/AIDS.


BACKGROUND:  To date, there has been little examination of how frequently providers prescribe lifestyle modifications to manage hypertension (HTN) and subsequent patient adherence rates to these recommendations. We examined rates of adherence to suggested lifestyle modifications for HTN management and of associated blood pressure control.

METHODS:  Using data from the NHANES 1999–2000 survey, we examined participants aged 20 or older who were told they had HTN at least twice by a doctor or other health professional to assess whether their provider had recommended lifestyle modification only, medication only or both to manage their HTN. Using logistic regression model we determined the odds ratio (OR) of receiving and adhering to each therapy adjusting for each participant's demographic and clinical characteristics. Confidence interval (CI) adjusted for clustering and weights were obtained using jackknife-1 procedure in SUDAAN statistical software. In a secondary analysis, we examined rates of hypertension control (<140/90 and <130/85 for diabetics) among adherent patients in each arm.

RESULTS:  Of aware hypertensive individuals (n = 1244), 146 (13.1%) were only advised to take medications and 81 (7.3%) to alter lifestyles, while 846 (76.0%) were advised to do both. Patients advised to take medication only had higher adherence rates (84.2%), than those advised to modify their lifestyle only (54.3%), or advised to do both (53.9%). Hypertension is controlled in 36.6% of those taking medications, 56.8% of those following lifestyle changes, and 46.1% doing both. Individuals are less likely to receive lifestyle recommendations if they are aged 60 or greater (OR 0.25; CI 0.13–0.48), are white (OR 0.46; CI [0.23–0.89]), or have completed high school (OR 0.39; CI [0.19–0.81). Individuals who are obese (BMI > 30) (OR 0.25; CI [0.15–0.42]), diabetic (OR 0.42; CI [0.20–0.89]), or hypercholesterolemic (OR 0.49; CI [0.30–0.80]), have lower odds of adhering to lifestyle recommendations, while higher odds of adhering to lifestyle recommendations were associated with being white (OR 1.76; CI [1.04–2.96]) or age 60 or older (OR 2.29; CI [1.33–3.96]).

CONCLUSION:  Despite higher adherence rates, prescribing medications alone has lower rates of hypertension control combined to prescribing both medications and lifestyle modification. Clinicians should recommend and encourage adherence to both medications and lifestyle changes for optimal management of hypertension, and future research examining patterns of care for hypertension should examine rates of advising lifestyle modification as well as medication prescribing.

FACTORS INFLUENCING SAFER SEX PRACTICES IN OLDER URBAN WOMEN.A. Paranjape1; L. Bernstein1; D. St. George2; J. Doyle1; S. Henderson1; G.M. Corbie-Smith3. 1Emory University, Atlanta, GA; 2Walden University, Minneapolis, MN; 3University of North Carolina at Chapel Hill, Chapel Hill, NC. (Tracking ID #117147)

BACKGROUND:  Infection with the human immunodeficiency virus (HIV) is an important health issue for older urban women, with most women in this population contracting the virus through heterosexual transmission. Primary prevention of HIV depends on safer sex practices, such as abstinence, consistent condom use or monogamy, however little is known about the factors that influence these practices in older women. Appropriate condom use not only depends on knowledge of condom efficacy, but also on relationship dynamics such as ability to negotiate condom use, dependence on one's partner for condoms, taking personal responsibility for obtaining condoms and trusting one's partner to disclose HIV sero-positivity. Therefore, we sought to determine the effect of relationship factors on older urban women's safer sex practices.

METHODS:  We conducted a cross-sectional study in 514 women, age 50 and older, attending an ambulatory medicine clinic of a large urban hospital located in a region with a high HIV incidence. Using face-to-face interviews, we assessed knowledge of condom efficacy by responses to the question ‘How effective do you think using a condom during sex is at preventing someone from getting HIV or AIDS?’ Partner trust, obtaining condoms and dependence on partner were assessed by affirmative responses to the questions: ‘Do you believe your current partner would tell you if he had HIV or AIDS?’, ‘In your current relationship have you ever bought or obtained condoms?’ and ‘Do you rely on your sexual partner to provide condoms?’ respectively. Safer sex, was defined as frequent condom use or abstinence. We conducted bivariate analyses using the chi-square test and the t test and used logistic regression modeling to estimate the effect of individual relationship factors on safer sex; 95% confidence intervals (CI) were calculated for all odds ratios.

RESULTS:  Of the 514 women, 155 (30%) were in a relationship at the time of the interview. Of those currently in a relationship, 81% were sexually active and only 13% used condoms frequently. There were no significant differences between those who did and did not practice safer sex by age, ethnicity, employment status and marital status. Trust in one's partner was associated with lower odds of safer sex (OR 0.3, CI 0.08–1.06) while obtaining condoms (OR 9.2, CI 1.9–44.2) and dependence on partner (OR 12.3, CI 3.0–50.3) were associated with higher odds of safer sex. Knowledge of condom efficacy was not associated with safer sex decisions.

CONCLUSION:  Despite the increase in HIV infections in older women, alarmingly few of these women practice safer sex. Safer sex practices in these women were not associated with knowledge of condom efficacy, but rather with relationship factors such as trust in one's partner, obtaining condoms and dependence on partner for condoms. Therefore, HIV prevention messages to older urban women should include references to these relationship factors and the role they may play in safer sex practices.


BACKGROUND:  Obesity is increasing among US adults with a disproportionate effect on minorities and underserved populations. There are few multi-disciplinary obesity programs in urban public hospitals serving these vulnerable populations. In order to meet this need and demonstrate its feasibility and effectiveness, a group of primary care residents with expertise in obesity management created a group-based, multi-disciplinary weight management clinic within a large urban public hospital serving primarily underserved and minority patients.

METHODS:  Due to treatment requirements eligibility was restricted to literate patients who had a BMI ≥30 and <50, were capable of mild exercise (>2 METs), and were without untreated psychiatric conditions. Patients were individually screened for eligibility by residents. Consenting patients were enrolled in a 16 week group with weekly group sessions and home assignments. Each session was conducted by the residents, a behavioral psychologist, and a nutritionist. The program focused on lifestyle change using established cognitive behavior techniques, nutrition and exercise counseling along with problem solving sessions.

RESULTS:  Residents screened 34 patients and 14 were enrolled (50% of ineligible patients were excluded due to a BMI ≥50). The mean age (standard deviation (SD)) was 53 (9) with 3 males and 11 females. The group was 36% Hispanic, 36% white, 14% African American, and 7% other ethnicities. Approximately 36% had less than a high school education. Among participants, 11 had Medicaid (71%), 3 had no insurance (21%) and 1 had private health insurance (8%). The mean BMI was 38 (SD = 6). Coexistent comorbidities were diabetes (29%), hypertension (29%), hyperlipidemia (36%), osteoarthritis (28%), depression (29%) and symptoms of obstructive sleep apnea (21%). At baseline 21% were smokers and 36% were taking aspirin. Patient adherence varied with 20% patients completing 0 visits, 20% completing 1–7 visits and 60% completing 8 or more visits. The median number of visits was 9. The mean weight loss (SD) among the group completing 8 or more visits was 5 kg (12) while those completing <8 visits gained 0.5 kg (2).

CONCLUSION:  This resident initiated weight management clinic in an urban public hospital provides a way for underserved adults with obesity to obtain comprehensive obesity treatment. We found that patients completing ≥8 sessions had significant weight loss. Such a multi-disciplinary group-based model may be an effective means for reducing obesity in vulnerable populations. However, non-adherence is a continuing challenge and methods for improving adherence to obesity programs must be developed and implemented in order achieve optimal results.

FINDING RESEARCH SUBJECTS IN THE AGE OF HIPPA.R. Hess1; M.A. McNeil1; C.J. Chang1; C.L. Bryce1. 1University of Pittsburgh, Pittsburgh, PA. (Tracking ID #116077)

BACKGROUND:  Privacy regulations associated with the Health Insurance Portability and Accountability Act (HIPAA) have complicated the conduct of outcomes research. To facilitate such research, we created the Functional Assessment System Tablet (FAST), a computerized touch-screen form that collects general patient information, facilitates longitudinal follow-up, and systematically queries all patients regarding willingness to be: 1) included in the research registry project (RRP) and 2) contacted about future research studies (the prospective subject list (PSL)). Within our implementation cohort, we investigated systematic bias in RRP and PSL participation.

METHODS:  All patients presenting to our outpatient practice self-complete the FAST. Real time information is provided to clinicians and placed in a de-identified research database. Upon completion, the computer asks patients to participate in the RRP and PSL. The RRP identifiably links the FAST with patients’ complete medical records for use in research. Using the PSL, researchers of IRB approved studies may query the RRP, identify study eligible patients, and approach them regarding participation. We examined the impact of clinical variables (i.e., gender, age, marital status, educational attainment, hazardous drinking, social support, self-reported disease, and health related quality of life via RAND-36) on RRP and PSL participation using Fisher's exact and Student's t tests.

RESULTS:  54 sequential patients completed the FAST; none refused. These patients included: 93% women; average age of 48 years (range 19–76); 53% married or in a committed relationship; 58% with at least a college degree; 7% hazardous drinkers; RAND-36 mental health composite (MHC) of 43 ± 12.7 and physical health composite (PHC) of 45±12.6. The most common self reported diseases included: hypertension (45%); arthritis (32%); and depression (38%). 31 patients (57%) participated in the RRP. We were unable to detect any significant association between RRP participation and clinical characteristics (p-values: 0.13–0.86). 87% of RRP subjects also consented to the PSL. We examined association between PSL consent and: age; educational attainment; arthritis; and depression and detected no significant differences (p-values: 0.16–0.67).

CONCLUSION:  In our preliminary sample, over half of sequential patients seen in a University based outpatient practice consented to participate in a research registry. We were unable to detect any systematic difference in this novel cohort inception method between those who participated and those who did not, suggesting that participants were representative of the overall patient population. Use of a computerized data collection instrument, such as the FAST, along with a research registry and PSL, represents a promising mechanism for facilitating both outcomes research and enrollment in clinical trials while adhering to current privacy regulations.

FIREARM SAFETY SCREENING AND COUNSELING IN PRIMARY CARE: A SURVEY OF RESIDENTS.J. Kahn1; M.M. Davis1. 1University of Michigan, Ann Arbor, MI. (Tracking ID #115808)

BACKGROUND:  Morbidity and mortality from firearms are a significant public health issue affecting adults and their children. In an attempt to lessen the toll of firearm-related injuries, several medical professional organizations including the ACP recommend physician screening and counseling regarding firearms in the primary care setting. To date, there are no data on the firearm screening and counseling practices of internal medicine residents. The objective of this study was to determine and compare the beliefs, attitudes and practices of house officers in different specialties with regard to firearm safety screening and counseling.

METHODS:  House officers in the Departments of Internal Medicine, Pediatrics and Family Practice at the University of Michigan were sent an anonymous, web based survey via e-mail. The survey included questions regarding firearm screening and counseling behavior in the residents’ primary care clinics, personal experience with firearms, attitudes about firearms and screening practices, and demographic information. Responses to 5-point-scale Likert-type items were dichotomized and analyzed using chi-square tests. The study protocol was approved by the University of Michigan Medical School Institutional Review Board.

RESULTS:  145 of 216 residents completed the survey (response rate=67%). Respondents were split among internal medicine (IM; 52%), med-peds (MP; 15%), pediatrics (PD; 22%), and family medicine (FM; 11%) programs, with variable response rates by specialty (PD 76%, MP 75%, IM 64%, FM 57%). Most residents (90%) believe that firearm-related violence is an important public health issue. 81% believe that physicians should be involved in firearm injury prevention, however this belief differed by specialty (IM 72% vs all others 90%; P < .01). By contrast only 56% of residents overall reported that they routinely screen their patients for firearm ownership, with IM residents less likely to screen (25%) than residents in other specialties (90%; P < .001). Firearm screening behavior was strongly associated with perceived clinic preceptors’ expectation regarding screening. 77% of IM residents perceived no expectation to screen from their clinic preceptor compared with 6% for other residents (P < .001). Firearm screening behavior was also associated with prior medical school training regarding screening (78% screening rate with training vs. 46% without; P < .001).

CONCLUSION:  Residents’ stated beliefs do not match their reported behaviors regarding screening for firearms, especially among internal medicine residents. Further training of medical students and of clinic preceptors may increase the number of residents who screen their patients for firearm ownership.

FOLLOW-UP OF ABNORMAL SCREENING MAMMOGRAMS AMONG WOMEN UTILIZING A COUNTY-SPONSORED MOBILE MAMMOGRAPHY VAN.M. Peek1; J. Han2. 1Rush University / Rush-Presbyterian-St. Luke's Medical Center, Chicago, IL; 2Rush Medical College, Chicago, IL. (Tracking ID #117385)

BACKGROUND:  Mobile mammogram vans have been successfully utilized to increase access to screening mammography for medically underserved women. However, there is often difficulty with tracking results and arranging for patient follow-up of abnormal mammograms. We sought to characterize the follow-up patterns for abnormal screening mammograms in a cohort of women utilizing a public county-sponsored mobile van.

METHODS:  We prospectively studied women aged 40 years and older presenting for screening mammography on mobile units from 1/1/01–7/01/02. Using electronic records and medical chart review, demographic, radiological and clinical variables were collected; women with abnormal mammogram results were followed for 12–24 months for evidence of diagnostic imaging within the city's public hospitals. We then surveyed those patients without documented follow-up to determine whether they had received follow-up outside of the safety net system and assess reasons for lack of diagnostic evaluation. Three rounds of mailed letters and telephone calls were conducted in an attempt to reach women with no documented follow-up.

RESULTS:  During the study period, 659 women had abnormal screening mammograms on the mobile van; 63.5% had BIRADS 3 mammograms, 31.4% BIRADS 0, and 5% BIRADS 4 or 5. The mean age was 54.5 years and patients were ethnically diverse: 54.8% African American, 18.6% Hispanic, 12% Caucasian, and 11% Asian. Only 20.2% of the screened women had documented diagnostic imaging within the public healthcare system. The mean follow-up interval was 152.1 days. We determined the follow-up status of 55.6% (n = 366) of the initial sample, of whom 76.2% received subsequent evaluation (either imaging or physician follow-up) after their abnormal mammograms, 153 within the safety net system and 126 outside of the system. Reported reasons for lack of follow-up included fear of more tests (10.1%), concerns of cost (6.7%), misunderstanding the report letter (9.0%), having forgotten (6.7%), other illnesses (5.6%), being too busy (6.7%), being unaware that the mammogram was abnormal (46.1%) or other reasons (22.5%). Race was not associated with inadequate diagnostic follow-up (P = .78), but younger women (less than 47 y/o) were less likely to have further evaluation (P = .0006), a factor not accounted for by BIRADS category.

CONCLUSION:  Successful follow-up of women screened in mobile mammography units is difficult, particularly when patients belong to multiple healthcare systems. One-fourth of women whose follow-up status was determined had no diagnostic follow-up. The most common reason was the lack of knowledge about the abnormal results, making aggressive tracking by experienced staff a crucial component of such outreach programs.

FOURTH YEAR MEDICAL STUDENTS’ STANDARDIZED PATIENT EXAMINATION PERFORMANCE AND SPECIALTY CHOICE.K.E. Hauer1; C. Hodgson1. 1University of California, San Francisco, San Francisco, CA. (Tracking ID #115716)

BACKGROUND:  National Residency Match data demonstrate a trend in medical students’ career choice away from specialties focused on patient interactions toward technology-based specialties. Few studies have examined the relationship between medical school clinical performance and specialty choice.

METHODS:  We analyzed scores from a 7-station comprehensive standardized patient examination for 124 beginning fourth-year medical students to determine whether there are differences in clinical performance based on intended specialty type: personal care (family practice, internal medicine, obstetrics and gynecology, neurology, pediatrics, psychiatry, dermatology) versus technology-oriented (anesthesiology, radiology, emergency medicine, pathology, radiation oncology, general surgery, surgical subspecialties). We used t tests to characterize differences in MCAT total scores, USMLE Step 1 scores, and CPX scores based on specialty. MCAT and USMLE scores were included as measures of baseline academic performance. Correlations between MCAT scores, USMLE scores and the dependent variables (History, Physical Exam, Physician-patient Interaction, and case total scores) were performed controlling for specialty. A maximum score of 100% was possible in each skill domain assessed in the exam. We performed Multiple Analyses of Variance (MANOVAs) with specialty and gender as the independent variables and MCAT and USMLE scores as covariates. The institutional review board approved the study.

RESULTS:  Among personal care students, 48 (79%) applied in primary care specialties (internal medicine, pediatrics, or family practice). Students anticipating careers in personal-care specialties versus technology-oriented specialties performed better in History Taking [67.9% (Standard deviation 7.08) vs. 64.3% (SD 7.12), P = .01] and Physical Examination [52.2% (SD 10.6) vs. 48.2% (SD 9.3), P = .02], although there were no differences in Physician–patient Interaction [68.6% (SD 6.8) vs. 67.2% (SD 7.4), P = .24]. Physician-patient Interaction scores were significantly higher for women than men [69.3% vs. 66.2%P = .05]. There was no indication of case specificity.

CONCLUSION:  The higher scores among students pursuing primary care and other personal-care specialties may reflect differential acquisition of clinical skills in the core clerkships. Alternatively, students with stronger clinical skills may gravitate toward personal-care specialties. These findings have implications for fourth-year curricula and for students’ preparation for the USMLE Step 2 Clinical Skills Examination to ensure that all graduates have achieved similar clinical proficiency.

GENDER AND ETHNIC DIFFERENCES IN OVERNIGHT URINARY STRESS HORMONES.C. Masi1; E. Rickett1; L. Hawkley1; J. Cacioppo1. 1University of Chicago, Chicago, IL. (Tracking ID #115602)

BACKGROUND:  Heart disease mortality varies by gender and ethnicity in the United States. In 2001, the age-adjusted death rate for heart disease was 50% higher among males compared to females and 30% higher among blacks compared to whites. Differences in stress hormone response may contribute to higher rates of cardiovascular disease among men and blacks. Spot testing of overnight urinary epinephrine (E), norepinephrine (NE), and cortisol (Cort) permits a time-integrated assessment of the stress response system.

METHODS:  Overnight specimens were obtained from a population based sample of 230 adults aged 50 to 65. Specimens were assayed for E, NE, and Cort and corrected for creatinine concentration and fat free mass. ANOVA was used to assess raw and corrected hormone differences by demographic variables.

RESULTS:  Significant gender differences were noted in raw E, NE, Cort, and creatinine concentrations, with males displaying higher levels of each. When corrected for creatinine excretion and fat free mass, males demonstrated higher levels of E and NE but not Cort. Blacks had significantly higher raw values for urinary creatinine, E, and NE compared to whites and Hispanics. When corrected for creatinine excretion and fat free mass, the hormone differences remained but were no longer statistically significant.

CONCLUSION:  Higher overnight E and NE may be a risk factor or marker for cardiovascular disease among men and blacks. To account for differences in urine concentration and muscle mass, spot urine assessment of stress hormones should include correction for creatinine excretion and fat free mass.

Table *. Urinary Creatinine (Cr), Fat Free Mass (FFM), and Corrected E,NE, & Cort (ng/mg creat/kg) (P < .05 for male-female) (**P < .05 for white-black)
 Male (n = 109)Female (n = 121)White (n = 83)Black (n = 81)Hispanic (n = 66)
Cr (mg/dL)  112.73   68.72*   82.05  103.65**   84.05
FFM (kg)   69.57   44.13*   56.86   57.17   54.21
E   94.58   68.76*   77.46   82.35   78.03
Cort  840.32  711.02  809.04  692.50  809.18

GENDER DIFFERENCES IN PRIMARY CARE PHYSICIANS’ SELF-ASSESSMENT OF TEACHING AND EVIDENCE-BASED MEDICINE SKILLS.S.G. Chheda1; C. Gjerde1; K.M. Hla1; P. Kokotailo1; B. Anderson1. 1University of Wisconsin-Madison, Madison, WI. (Tracking ID #116942)

BACKGROUND:  Recent studies have shown gender differences in physician-scientists’ self-assessed abilities to perform clinical research. Self-assessment of abilities is a measure of self-efficacy, an individual's belief in their ability to accomplish tasks and reach goals. Self-efficacy is an important variable in learning and career outcomes. It is unknown if such gender differences in self-assessed skills exist in physicians in other academic areas. The purpose of this study was to examine gender differences in the self-assessed skills of primary care physicians participating in a year-long faculty development program focusing on enhancement of clinical teaching and evidence-based medicine (EBM)skills.

METHODS:  Participants’ in the HRSA supported University of Wisconsin Primary Care Faculty Development Program (PCFDP) attended five weekend sessions over a one-year period. Program content included skills in teaching and EBM. Participants completed a self-assessment of their skills using a 6 point scale (0 = no skills, 5 = high level skills) at the start (pre-assessment) and conclusion (post-assessment) of the program. Statistical analysis was completed using the two-sample t test.

RESULTS:  Between 1999 and 2003, all 38 female and 34 male participants completed pre-assessment, and 32 female (84%) and 31 male (91%) participants completed the post-assessment. On pre-assessment women ranked themselves lower than men rated themselves on all seven teaching assessment items, with four of these reaching statistical significance. Similarly, women ranked themselves lower than men rated themselves on all five EBM pre-assessment items, with four of these reaching statistical significance. Though both men and women collectively ranked themselves higher at post-assessment compared with pre-assessment, and women tended to have larger gains than men, women rated themselves lower than men rated themselves on all items. These differences reached statistical significance (P < .05) for all five EBM items (use EBM concepts in clinical practice use likelihood ratios, use terms like AAR, RRR and NNT, critically appraise trials, use pre-appraised EBM resources); and three of seven teaching items (use adult education principles, give feedback to learners and solicit feedback on teaching).

CONCLUSION:  Female primary care physicians consistently rated their teaching and EBM skills lower than male participants both before and after participation in a faculty development program focused on enhancing these skills. Decreased self-efficacy may indicate a barrier in women's participation in academic roles requiring these skills. Attending to these gender differences may enable faculty development programs to better meet the needs of all participants.


BACKGROUND:  Community-acquired pneumonia (CAP) results in expenditures of over $9 billion yearly. Such is primarily due to the need for inpatient, parental antibiotic therapy. In an effort to decrease length of stay (LOS) in patients with CAP while maintaining quality, the American Thoracic Society (ATS) guidelines for CAP management include criteria for early intravenous-to-oral switch therapy. We hypothesized that knowledge of and adherence to the ATS guidelines for switch therapy was low at our institution, resulting in unnecessarily long hospitalizations.

METHODS:  University Hospital (UH) adopted a CAP clinical guideline and pre-printed an order set with recommendations similar to the ATS guidelines for switch therapy. Prior to implementation of the UH CAP guideline, 130 patients admitted with CAP were retrospectively identified to determine if “switch therapy” was initiated in an appropriate fashion as indicated in the ATS guidelines. Additionally, physicians were surveyed on ATS guideline awareness before and subsequent to adoption of the UH CAP clinical guideline.

RESULTS:  Of 130 patients reviewed, 71 were excluded because of comorbidities that extended hospitalization. Seventy-six percent of the remaining candidates were eligible for switch therapy. Fewer than 60% were indeed switched when deemed appropriate by ATS guidelines. When patients were treated in accordance to ATS guidelines, LOS was one day less (2.4 vs. 3.6, P < .05). Of physicians surveyed, just 5% were aware of guidelines for switch before adoption of the UH guidelines, versus over 40% after guideline adoption.

CONCLUSION:  Early-switch therapy for CAP had not been utilized adequately prior to implementation of the UH CAP guideline. Such utilization is apt due in part to a lack of awareness. Adoption of guidelines by UH increased guideline awareness and thereby may also reduce LOS for patients with CAP.

GLYCEMIC AND LIPEMIC TESTING AND CONTROL IN ETHNIC MINORITIES WITH DIABETES.J.H. Halanych1; H. Lin2; F. Wang2; D.R. Miller2; D.R. Berlowitz2; K.M. Skinner2; L. Pogach3; S.M. Frayne4. 1University of Alabama at Birmingham, Birmingham, AL; 2CHQOER, VA Hospital, Bedford, MA; 3VA New Jersey Health Care System, East Orange, NJ; 4VA Palo Alto Health Care System, Menlo Park, CA. (Tracking ID #116135)

BACKGROUND:  Ethnic minorities have higher rates of diabetes and its complications. Differences between ethnic minorities and whites in medical care processes and outcomes are not fully described. We examined disparities in glycemic and lipemic testing and control in a large national sample of African American, Hispanic and white diabetics.

METHODS:  Our study included non-institutionalized patients with diabetes from Veterans Health Administration (VA) facilities that submitted lab data and procedure codes to central VA databases in 1999. We compared African American and Hispanic to white veterans on 4 of the national Diabetes Quality Improvement Project measures: annual glycosylated hemoglobin (HbA1c) testing, annual low density lipoprotein (LDL) testing, poor glycemic control (HbA1c9.5), and poor lipemic control (LDL130). Logistic regression was used to adjust for age, gender, and comorbidity, and random effects modeling accounted for facility-level clustering.

RESULTS:  Of 322,215 diabetics, 4% were Hispanic, 17% African American, and 79% white. In adjusted analyses, fewer minorities received annual lipemic testing and more had poor lipemic control than whites (Table). African Americans had similar annual glycemic testing, but more had poor glycemic control. Hispanics had more annual glycemic tests, and still had more individuals with poor glycemic control.

CONCLUSION:  Ethnic disparities still exist, especially in outcomes. While there is room for improvement in rates of lipid testing, more notable are the increased rates of poor glycemic and lipemic control. Thus, quality improvement efforts for minorities should emphasize better outcomes (rather than processes) of care, for example through treatment intensification and prioritization of clinical goals to maximize cardiovascular risk reduction.

Table 15. Odds Ratios (95% C.I.) in minorities vs whites (adj. for age, sex, comorbidity, & facility effects)
 HispanicAfrican American
Received glycemic testing1.14 (1.09, 1.20)0.99 (0.97, 1.02)
Received lipemic testing0.89 (0.85, 0.93)0.74 (0.73, 0.76)
Poor glycemic control1.14 (1.09, 1.19)1.35 (1.32, 1.39)
Poor lipemic control1.13 (1.08, 1.17)1.49 (1.46, 1.53)

GRADUATES’ PERCERPTIONS OF THE VALUE OF AN “EDUCATION FOR LIFE” REQUIREMENT IN AN INTERNAL MEDICINE RESIDENCY.M. Panda1; T. Ball1; N. Desbiens1. 1University of Tennessee, Chattanooga, Chattanooga, TN. (Tracking ID #115732)

BACKGROUND:  The ACGME defines Internal Medicine as a discipline requiring evidence-based (EB) decision making and commitment to lifelong learning. Literature on life-long learning recommends that it should form a part of the curriculum of every profession and EB practice implies life-long learning. To our knowledge, there is little literature about specific curricula that promote life-long learning during internal medicine residency. The Chattanooga Unit of the University of Tennessee, College of Medicine has required third year residents to submit a report on their plans for keeping up with their medical education (“Education for Life”) as a requirement for graduation since 1999. We have done a follow-up survey of the first three graduating classes that completed this requirement. We hypothesized that graduates would find the requirement useful, and that graduates who sought further specialization would use more modalities to remain current with advances than generalists.

METHODS:  A survey containing questions regarding modalities used by the graduates to keep up with their education, their current careers and demographics was sent to 27 graduates 2 to 4 years after graduation. These were compared with the “Education for Life” reports they had submitted prior to graduation.

RESULTS:  Twenty-four surveys were returned giving a response rate of 89%. The current age of the graduates ranged from 29–42 years with a mean of 33 years. Sixty-seven percent were male, and 83% were married at the time of graduation. Twenty-nine percent had pursued fellowship training; 71% practiced as internists. All but one respondent remembered the “Education for Life” requirement; 57% thought it beneficial. Though the graduates intended to keep up with a mean of 3.4 methods, they reported keeping up with a mean of 4.2 methods. Residents usually kept up with one or two more modalities than they originally estimated (fig). In the multivariable analysis we found that the number of modalities graduates used was significantly associated with the number they planned to use before graduation (P = .04) but not with the choice of subspecialty versus generalism as a career.

CONCLUSION:  As hypothesized, graduates find an “Education for Life” requirement useful. Graduates, regardless of specialty orientation, used more modalities for continuing medical education than they thought they would, perhaps reflecting a void they felt after their departure from a structured education during their residency training. Though a larger multi-institutional study is needed, our study helps identify one method of promoting life-long learning.

GUIDELINE ADHERENCE AND BLOOD PRESSURE CONTROL AMONG SPECIAL POPULATIONS.L.S. Hicks1; D. Fairchild2; E. Orav3; D.W. Bates3; J.Z. Ayanian4. 1Brigham and Women's Hospital and Harvard Medical School, Boston, MA; 2New England Medical Center, Boston, MA; 3Brigham and Women's Hospital, Boston, MA; 4Harvard University, Boston, MA. (Tracking ID #116764)

BACKGROUND:  There has been little examination of adherence rates to the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure (JNC) special populations guidelines and whether adherence to these guidelines is associated with improved blood pressure (BP) control among these high-risk patients.

METHODS:  We reviewed 15,768 visits to 14 general internal medicine clinics between 7/1/01–6/30/02. We classified each visit as adherent or non-adherent by identifying medications selected for therapy for HTN and comparing them with JNC VI racial/ethnic-specific and disease-specific guidelines. We also classified each visit as BP controlled (BP of <130/85 for patients with diabetes or renal failure or <140/90 for other patients) or not. We compared adherent and BP controlled visit rates by patient demographic and clinical characteristics using chi-square tests. Logistic regression was used to determine adjusted odds of JNC adherence and BP control by patient race/ethnicity and clinical characteristics.

RESULTS:  Adherent visits more frequent among Black (83.4%) and Latino (83.0%) patients, compared to Whites (78.4%) (P < .001). Adherent visits was also more frequent among women (79.0% for men, 82.2% for women, P = .02). After adjustment for demographic characteristics and comorbid disease, adherent visits remained more common among Blacks (O.R. 1.36[1.14–1.64]) compared to Whites, and less common among men (O.R. 0.80 [0.68–0.95]) compared to women. BP was controlled more often among White patients (41.6% compared to 35.3% for Black and 20.1% for Latino, P = .05), and among the privately insured (42.3% compared to 13.7% for Medicaid and 35.9% for Medicare, P = .006). After adjustment for demographic characteristics, JNC adherence, and comorbid diseases, diabetic patients were less likely to have BP control (O.R. 0.45 [0.40–0.51]) than non-diabetics, and patients with coronary artery disease (CAD) more likely to have BP control (O.R. 1.28 [1.07–1.53]) compared to those without CAD. JNC adherence was not associated with BP control.

CONCLUSION:  We found high rates of physician adherence to JNC guidelines among special populations, however blood pressure control rates were low, particularly among diabetics. The drug class prescribed may be less predictive of obtaining BP control than other processes of care such as intensity of therapy. Providers should focus their efforts on more aggressively managing HTN among diabetics.

GUIDELINE ADHERENCE TO DIURETIC USE IN MULTI-DRUG ANTIHYPERTENSIVE REGIMENS.H. Singh1; M.L. Johnson1. 1Baylor College of Medicine, Houston, TX. (Tracking ID #116542)

BACKGROUND:  Recent hypertension guidelines acknowledge that majority require two or more agents to achieve target blood pressure (BP). The Joint National Committee. Sixth Report (JNC VI, 1997) states the need to use a diuretic as a first or second line agent in absence of compelling indications for other drugs or contraindications, and emphasized need for a diuretic based regimen when using three or more drugs to avoid resistance in control. Inadequate use of diuretics as first line agents occurs despite evidence of high level of patient tolerability at lower doses. Data about use in multi-drug regimens is limited but of critical value. The objective of our study was to assess adherence to JNC VI guidelines in use of diuretics in multi-drug regimens for hypertension.

METHODS:  A retrospective design was used to collect computerized data from all patients seen for any reason (N = 84,369) at a tertiary care Veterans Affairs hospital longitudinally from October 1, 1998 to March 1, 2001.Patients with hypertension were identified for inclusion based on a combination of : 1) diagnostic information from ICD-9-CM codes 2) prescription drug use from pharmacy dispensing records, and 3) clinical parameters from vital signs. Any patients with diagnosis of chronic heart failure (CHF) and diabetes were excluded. Prescriptions of seven classes of anti-hypertensive drugs were identified from the computerized pharmacy records and the number being prescribed was tabulated.The proportion of patients with any thiazide or loop diuretic use among patients with two, three, or four or more drugs was tabulated.

RESULTS:  A total of 25,052 patients (29.7%) were studied. Average age(sd) was 59.8(13.2) years; 96% of patients were male, and 31.5% were ethnic/racial minorities. About 44% (n = 11,049) of patients were receiving two or more anti-hypertensive drug classes. Among patients on two drug regimens (n = 5,717) 50% were on a diuretic. Among patients on three drug regimens (n = 3,327) where diuretic use according to guidelines is almost mandatory, 73% of patients were on diuretics, increasing to 89% in regimens of 4 or more drugs (n = 2,005). Therefore, over one-third of all patients on multi-drug regimens (n = 3,970 or 35.9%) were not receiving diuretics even when compelling indications for other drugs (CHF and diabetes) were excluded.

CONCLUSION:  Our study found substantial non-use of low dose diuretics in multi-drug regimens to control blood pressure. Even though the medical literature has been propagating diuretic use for several years, better strategies to educate practitioners about increasing their use in multi-drug regimens and dispelling some of the myths about diuretic adverse effects may still be needed. Further research is needed to qualify how these rates of use compare outside the VA setting, and to determine patient and provider factors associated with variation in diuretic use.

HEALTH CARE COSTS BEFORE AND AFTER ANTIDEPRESSANT INITIATION: DOES ADEQUATE TREATMENT REDUCE COSTS?R.S. Stafford1; K.M. O’Leary2; S.N. Finkelstein3; J.B. Weilburg2. 1Stanford Prevention Research Center, Program on Prevention Outcomes and Practices, Stanford, CA; 2Massachusetts General Hospital, Boston, MA; 3Massachusetts Institute of Technology, Cambridge, MA. (Tracking ID #116279)

BACKGROUND:  Health care costs are known to be higher in patients with depression. We sought to evaluate whether the adequacy of newly initiated antidepressant (AD) treatment affected health care costs. We hypothesized that in those patients receiving adequate AD trials increased AD pharmacy costs would be offset by a reduction in other health care costs.

METHODS:  Data from a large HMO plan in Massachusetts were used to identify a cohort of 4,407 patients where health care costs could be followed for 9 months before and 9 months after the initiation of ADs. Using pharmacy claims data, we defined an adequate AD trial as involving at least 6 months of continuous therapy at an average daily dose of at least 20 mg of fluoxetine or the drug-specific equivalent. We tracked health care costs by quarter using medical and pharmacy claims data.

RESULTS:  Of the 4,407 patients initiated on ADs, average health care costs were $910 per person per quarter over the 18 months of follow-up, compared to typical mean quarterly costs of $400 per member. Only 38% of the patients initiated on ADs had an adequate AD trial. Costs per patient rose in the 9 months prior to AD initiation ($660 per quarter 6–9 months before vs. $730 3–6 months before vs. $990 1–3 months before). There were no differences in pre-treatment costs between patients later treated adequately vs. inadequately. Costs peaked in the quarter following initiation ($1,230) with adequately treated patients having greater costs ($1,340) compared to inadequately treated patients ($1,160). In subsequent quarters, health care costs continued to differ depending on AD trial adequacy. For adequately treated patients, costs did not decrease considerably over the ensuing quarters ($1,190 per quarter 6–9 months after initiation), contrary to our hypothesis. Patients lacking adequate treatment, however, had a significant (P < .01) reduction in their health care costs to $790 per quarter at 6–9 months after initiation. Of this mean $400 difference between adequate and inadequately treated patients, antidepressant pharmacy costs accounted for $160, other medications for $70, outpatient costs for $90, and inpatient costs for $80 per quarter.

CONCLUSION:  Patients ultimately prescribed ADs had higher health care costs that were observed even prior to AD initiation. Although we found no evidence of a cost-offset associated with adequate treatment among the patients we studied, such an effect cannot be excluded. The observation of higher costs incurred by adequately treated patients could potentially be explained by a greater severity of illness among these patients.

HEALTH CARE WORKERS’ EXPECTATIONS OF & EMPATHY TOWARD PATIENTS IN ABUSIVE RELATIONSHIPS.C. Nicolaidis1; M.A. Curry1; M. Gerrity1. 1Oregon Health & Science University, Portland, OR. (Tracking ID #116528)

BACKGROUND:  Most surveys of health care workers’ attitudes about domestic violence (DV) have focused on screening rates and barriers to screening, but primary care providers must often continue to care for patients in abusive relationships long after the initial screening or disclosure takes place. Less is known about attitudes that may affect a provider's ability to provide appropriate long-term care or to empathize with patients who choose to remain in abusive relationships.

METHODS:  We surveyed 235 healthcare workers in 25 community-based primary care practices prior to their participating in an educational intervention. Associations were assessed by Chi2 tests, Pearson's correlations, and t tests using STATA Software.

RESULTS:  A majority (56–61%) agreed with statements regarding a provider's responsibility to screen for DV in a a variety of situtions where DV screening would be recommended, though fewer (11–35%) stated they always or nearly always screened for DV in such situations. Participants’ sense of responsibility regarding screening correlated with their self-reported screening practices (P < .0001). A majority (57–59%) agreed with statements meant to assess unreasonable expectations or overly controlling behavior. (Eg. “A provider's responsibility includes making sure a patient gets to a shelter right away if he or she discloses abuse.”) Most participants (50–58%) found it easy to empathize with a patient who decided to stay in an abusive relationship in situations where the patient was described as poor or disabled, regardless of patient gender. Few (24–38%) could empathize with a patient who made such a decision in situations where the patient was described as educated or financially secure. Participants who agreed with statements meant to assess unreasonable expectations or overly controlling behavior had more difficulty empathizing with patients who decided to remain in abusive relationships (P = .01). Those who stated that they, a close friend, or relative had experienced DV (64%) found it easier to empathize with patients who chose to remain in abusive relationships than those who had not had personal experience with DV (P = .009). Gender, prior DV training, sense of responsibility regarding screening, and screening practices did not correlate with unreasonable expectations or with ability to empathize with patients who choose to remain in abusive relationships.

CONCLUSION:  Domestic violence training efforts must not only focus on increasing screening rates, but also on helping providers have more reasonable expectations about the natural history of abusive relationships. Health care workers need to better understand why patients may choose to remain in abusive relationships, even in the absence of economic or health limitations.

HEALTH INSURANCE COVERAGE, RACE, AND MORTALITY AMONG NEAR-ELDERLY ADULTS.J.M. McWilliams1; A.M. Zaslavsky2; E. Meara2; J.Z. Ayanian2. 1Brigham and Women's Hospital, Boston, MA; 2Harvard University, Boston, MA. (Tracking ID #115752)

BACKGROUND:  Uninsured near-elderly adults may be particularly at risk of adverse health outcomes, including premature mortality, due to insufficient treatment of chronic illness and modifiable risk factors. Therefore, we compared mortality of insured and uninsured near-elderly adults with stratification by race, income, and the presence of diabetes, hypertension or heart disease.

METHODS:  Among 8,736 adults age 50–61 in the nationally representative Health and Retirement Study, we assessed the relation between insurance coverage in 1992 and subsequent mortality through 2000. Employing rigorous propensity-score methods to adjust for numerous sociodemographic, health, and behavioral characteristics, we used Cox proportional hazards models to compare mortality between privately insured and uninsured adults weighted according to their propensity to be insured.

RESULTS:  Adjusted mortality was significantly greater among uninsured adults than privately insured adults (8-year rates: 10.5% vs. 7.5%; hazard ratio: 1.43 (95% C.I.: 1.10,1.85)). This finding was evident among white adults (11.2% vs. 7.3%; 1.57 (1.16,2.12)) but not among black adults (11.3% vs. 10.4%; 1.08 (0.67,1.75)) or Hispanic adults (2.7% vs. 7.5%; 0.35 (0.13,0.96)). Additionally, being uninsured was associated with greater adjusted mortality among adults who reported low incomes (14.1% vs. 9.4%; 1.53 (1.11,2.12)) or diabetes, hypertension, or heart disease in 1992 (18.8% vs. 12.5%; 1.56 (1.15,2.10)), but not among those with moderate to higher incomes (8.1% vs. 6.4%; 1.27 (0.78,2.06)) or without these conditions (5.4% vs. 4.5%; 1.22 (0.82,1.80)), respectively.

CONCLUSION:  Lacking health insurance was associated with substantially higher mortality among near-elderly adults who were white, had low incomes, or had diabetes, hypertension or heart disease. Increased mortality among both insured and uninsured black adults may indicate that health insurance is insufficient to overcome lifelong risk factors for ill health or that insured black adults face greater barriers to care and receive lower quality health care than insured white adults. Unexpectedly lower mortality among uninsured Hispanic adults may reflect unmeasured differences in acculturation between insured and uninsured Hispanic adults that affect health outcomes. Expanding coverage to uninsured near-elderly adults may improve health outcomes for some vulnerable groups but be insufficient to reduce the increased mortality experienced by near-elderly black adults.

HEALTH LITERACY IN THE SOUTH BRONX: PREVALENCE AND PICTORIAL PRESCRIPTION INTERVENTION.S. Nasr1; C. Napier2; I. Sharif1; G. Sacajiu2; S. Villanueva1. 1Montefiore Medical Center, Bronx, NY; 2Albert Einstein College of Medicine, Bronx, NY. (Tracking ID #116755)

BACKGROUND:  More than 40 million Americans are functionally illiterate. Functional health literacy (FHL) is a set of skills required to operate in a health care environment. Low literacy has been documented in urban settings and studies have found it to be related to poor health outcomes. However, no studies have measured health literacy in the South Bronx, containing the poorest congressional district in the nation. Furthermore, pictorial prescriptions have been proposed as a way to improve FHL. No studies have evaluated the effectiveness of pictorial prescriptions. The objectives of this study were 1) to measure FHL in a South Bronx population and 2) to test the feasibility of pictorial prescription labels as an intervention in that population.

METHODS:  We conducted a cross-sectional survey of randomly selected adult patients presenting for routine appointments at an urban clinic from February to August 2003. All patients received the Short Test of Functional Heath Literacy in Adults to measure FHL. Depending on the patient's primary language the test was given in English or Spanish. Next, patients completed a questionnaire evaluating the understanding of pictorial versus written prescription labels.

RESULTS:  Of 148 approached, 114 patients completed the survey. FHL was found to be inadequate or marginal in 31.6%. Of Spanish speakers (n = 26), 53.8% had inadequate or marginal FHL. Of English speakers (n = 88), 25% had inadequate or marginal FHL. The overall prevalence of health illiteracy found in our random sample is comparable to that found in other urban communities (Power = 93%). Overall, pictorial prescription labels were understood by 35 % of patients; written prescriptions were understood by 93%. Amongst those with inadequate or marginal FHL, only 12.9% understood the pictorial prescriptions versus 76.7% of the written prescriptions.

CONCLUSION:  We found the level of inadequate FHL in this poor urban population substantial. This proportion of health illiteracy is likely to have a significant impact on patients’ well being. While our intervention does not appear to significantly improve health literacy, we should continue to look for patient-centered solutions to better the understanding of health information.

HEALTH LITERACY OF PATIENTS ENROLLED IN DIABETES EDUCATION.S. Kim1; F. Love1; D. Quistberg1; J.A. Shea1. 1University of Pennsylvania, Philadelphia, PA. (Tracking ID #115361)

BACKGROUND:  Poor health literacy, common in patients with diabetes, has been associated with worse health outcomes. Diabetes education can improve health outcomes by teaching patients self-management skills. However, little is known about health literacy levels of patients that receive diabetes education. The objectives of our study were to assess the correlates of health literacy in patients receiving diabetes education and examine the association of literacy with diabetes knowledge, glycemic control, and self-management behaviors.

METHODS:  In 2003, all patients enrolled monthly in multiple session diabetes education classes at the Hospital of the University of Pennsylvania were recruited. At the first class health literacy was measured by the short-form Test of Functional Health Literacy in Adults (s-TOFHLA). Diabetes knowledge and self-management behaviors were assessed using previously published instruments. Glycemic control was assessed by obtaining most recent hemoglobin A1C (HbA1C) levels from medical records.

RESULTS:  Fifty eight percent (N = 92) of patients participated. There were no differences between participants and non-participants in age, sex, race, education, and HbA1C. Overall, 77% had adequate and 23% had marginal or inadequate health literacy skills. Lower literacy was associated with being older (mean 67 vs 58 years, P = .001), having less education (mean 10 vs 14 years, P < .001), lower annual income (P = .001), no commercial insurance (P = .002), and more diabetes complications (P = .042). There were no differences in years with diabetes (P = .59). Patients with adequate health literacy skills had better knowledge of diabetes (mean out of 24, 17 vs 14, P = .014), but health literacy was not associated with HbA1C (8.4 vs 8.2, P = .68) or self-management behaviors (mean days of adherence out of the last seven) such as diet (4.3 vs 4.7, P = .47), exercise (2.7 vs 2.3, P = .52), and medication adherence (6.0 vs 6.6, P = .24).

CONCLUSION:  In this select sample of diabetic patients enrolled in diabetes education, the prevalence of low health literacy was lower than that reported among diabetics in primary care settings. Low health literacy was associated with less diabetes knowledge, but was not associated with worse glycemic control or self-management behaviors. Regardless of literacy level, all patients had room to improve on self-management behaviors. Efforts should focus on improving educational outreach for patients with low literacy skills.


BACKGROUND:  Although insurance status is often measured as a static condition, individuals may transition from insured to uninsured status and vice-versa. The prevalence of such transitions for nonelderly adults and corresponding patterns of healthcare utilization and expenditures have not been well characterized.

METHODS:  We analyzed the 1998 Medical Expenditure Panel Survey to characterize the insurance status and insurance transitions for individuals aged 19–64 years old (unweighted n = 12,856). We used multivariate negative binomial regression to examine the association of insurance transitions with measures of healthcare utilization, and used 2-step logistic-linear multiple regression to analyze patterns of expenditures related to transitions. All models were adjusted for age, gender, race/ethnicity, presence of chronic condition(s), receipt of SSI due to disability, household income, marital status, education, employment status, region of residence, and presence of child dependents. Analyses accounted for the complex sampling design of MEPS.

RESULTS:  In 1998, 24.34 million adults (15%) were continuously uninsured, compared to 118.54 million (73.1%) who were continuously insured through private or public sources. Over 19 million adults experienced 1 or more transitions in insurance status that involved a period of uninsurance: 8.36 million (5.2%) experienced uninsured-to-insured (U-I) transitions, while 10.98 million (6.8%) experienced insured-to-uninsured (I-U) transitions. In multivariate models, utilization of emergency visits was significantly higher among adults with U-I transitions than among those continuously uninsured (incidence rate ratio = 1.51; 95% CI 1.11, 2.04); emergency visit patterns did not otherwise differ by insurance status. Compared to continuously uninsured persons, adults who were continuously insured or had I-U or U-I transitions had significantly more frequent inpatient discharges and office-based provider visits. Adjusted mean annual healthcare expenditures per person were only slightly higher for continuously insured adults ($2,013; 95% CI $1,882, $2,144) than for adults with U-I transitions ($1,870; $1,475, $2,266) and I-U transitions ($1,861; $1,437, $2,286), all of whom had significantly higher mean total expenditures than continuously uninsured adults ($1,147; $922, $1,372). Prescription medication expenditures revealed a similar pattern.

CONCLUSION:  Our findings suggest that classification of health insurance based on current status alone may obscure distinct patterns of health services use among adults with insurance transitions. Moreover, examination of the effects of transitions may inform evaluations of programs designed to facilitate and maintain insurance coverage for US adults.

HEPATITIS C PATIENTS: WHAT DO THEY REALLY MEAN WHEN THEY SAY, “I QUIT DRINKING”?S.W. Kanuch1; N.V. Dawson2; R.C. McCormick2; E.P. Stoller2; C.E. Blixen3; A.T. Perzynski2; J.J. Terchek2. 1MetroHealth Medical Center, Cleveland, OH; 2Case Western Reserve University, Cleveland, OH; 3Cleveland Clinic Foundation, Cleveland, OH. (Tracking ID #115979)

BACKGROUND:  Greatly reducing or eliminating alcohol use may importantly enhance the prognoses of patients with Chronic Hepatitis C virus (HCV), even if they are not candidates for HCV treatments. Little is known about factors related to alcohol use among patients who are not dependent on alcohol and who have a disease or receive medication for which alcohol is contraindicated. To gain insight into these factors, we asked alcohol non-dependent patients who are positive for HCV about their alcohol use and found that there is considerable variability in what a patient may mean when he says, “I quit”.

METHODS:  Subjects (N = 30) were black (16), white (10), and Hispanic (4) males (17) and females (13) aged 18 years or older, who were diagnosed with HCV at an urban teaching hospital. To be eligible for the study, patients had to be non-dependent drinkers, determined by having an AUDIT (Alcohol Use Disorders Identification Test) score less than 11. Patients were given a pencil-and-paper questionnaire with the question, “Did you stop drinking due to your diagnosis?” and with the AUDIT questions. Eligible patients who consented were interviewed using a semi-structured, open-ended format. Interviews were audiotape recorded, transcribed, and coded. Transcript-based techniques were used to analyze the findings.

RESULTS:  During the interview 20 patients said that they quit drinking or referred to themselves as a non-drinker. During the course of the interview, 6 of these respondents reported drinking on “special occasions” and 5 of them revealed that they had merely changed their drinking pattern to reduce the amount and/or type of alcohol consumed. 8 respondents indicated on the questionnaire that they had quit drinking with diagnosis. 2 of these respondents revealed during the interview that they had reduced their alcohol intake rather than remaining abstinent. In all, 12 of the 30 respondents indicated in some format that they quit, yet continued to consume some alcohol.

CONCLUSION:  These findings suggest that the lay definition of “quitting” is variable. Quitting may imply a reduced pattern of drinking behavior. The sample size reported here is not sufficiently large to generalize about the prevalence of these findings to the general HCV population. Studies with larger samples are being carried out to further examine alcohol use behaviors in the face of an HCV diagnosis.

HEPATITIS C PREVALENCE AND RISK FACTORS IN AN URBAN PRIMARY CARE PRACTICE.T.M. Sturm1; A. Federman1; J. Wisnivesky1; N. O’Connor1; D. Gardenier1; T. McGinn1. 1Mount Sinai School of Medicine, New York, NY. (Tracking ID #117559)

BACKGROUND:  Hepatitis C virus (HCV) prevalence has been estimated at 1.8 % nationwide, and 3–5% in primary care settings. It is estimated that half of patients with HCV have not been identified. Since routine screening for HCV has not shown to be cost effective, it is essential to be able to identify high-risk patients in primary care settings and initiate effective screening practices. Our goals are to determine the prevalence of HCV antibody among adults in an East Harlem hospital-based primary care practice, and to identify risk factors for infection.

METHODS:  A 27-item risk questionnaire was developed to identify potential risk factors, and was divided into 5 exposure domains: occupational, medical, general exposure, personal care, and risk behavior situations. Adult patients waiting to be seen by a provider in a hospital-based primary care practice were randomly recruited to complete the questionnaire and have blood tested for HCV antibody.

RESULTS:  A total of 1,000 patients were recruited, with a 33% refusal rate. Those who refused participation were older (mean age 58 vs. 51, P < .05), more likely to be African-American (48% vs. 34%, P = .000) and have Medicare (38% vs. 17%, P = .000). Patients agreeing to participate were more likely to have Medicaid (55% vs. 45%, P < .001) or private insurance (18% vs. 12%, P = .004) and be Hispanic (53% vs. 40%, P = .000). There was no difference between participants and non-participants in gender or type of appointment. Among study participants the mean age was 51, 74% were female, 55% self-reported Hispanic ethnicity, 34% African-American, and 55% had Medicaid insurance. HCV prevalence was 8.3%, with 30% of those being a new diagnosis. Stepwise logistic regression identified six variables significantly associated with HCV positivity: puncture/injection with a needle that may have been used before [OR 13.7, 95% CI (5.2 to 36.6)], work as police/fire/EMS/correction officer [OR 9.0, CI (1.9 to 42.5)], history of abnormal liver blood tests [OR 6.2, CI (2.9 to 13.3)], cocaine use ever [OR 3.6, CI (1.8 to 7.3)], history of a positive test for HIV or hepatitis B [OR 3.8, CI (1.5 to 9.4)], and Medicaid insurance [OR 2.4, CI (1.2 to 4.8)]. Individuals under age 40 had a lower odds of infection [OR 0.1, CI (0.03 to 0.54)].

CONCLUSION:  HCV prevalence in this largely Hispanic and African-American urban primary care practice is 8.3%, significantly higher than reported in other primary care settings. Among patients with HCV, 30% were unaware of their infection. The patient characteristics identified here can be used to derive an instrument to guide screening practices in similar primary care environments.

HEPATITIS C SEROPREVALENCE IN GENERAL MEDICINE AND TRAUMA SERVICES.B.J. Turner1; K. Brady1. 1University of Pennsylvania, Philadelphia, PA. (Tracking ID #117433)

BACKGROUND:  Hepatitis C virus (HCV) infection warrants efforts to screen patients in high prevalence settings for both primary and secondary prevention purposes. We conducted an anonymous seroprevalence study in inpatients in two urban hospitals to determine the frequency and predictors of undiagnosed HCV in general internal medicine (GIM) and trauma services.

METHODS:  From 8/02–10/02, we obtained 789 left over sera for 998 unique patients (79%) who were aged 17–65 and admitted to 2 GIM services (university hospital (UH) and affiliated hospital (AH)) or the trauma service at the UH. We obtained demographics and clinical histories from in- and out-patient administrative files for up to two years before admission. After removing all identifiers, an outside lab performed anti-HCV EIA and, among those with a positive test, HIV EIA with Western Blot confirmation. We estimated a logistic regression model to identify predictors of undiagnosed HCV.

RESULTS:  Of the 789 tested, 65 (8.2%) had known HCV and, of those with HCV, 8 were HIV+ (12.3%). Of the remaining 724 patients without known HCV, the demographics were: mean age 44, 65% Black, 35% White, 50% male, 25% married, 63% single, and 20% trauma. Of these 724 patients, 63 (8.7%, 95% CI 6.7, 11%) had undiagnosed HCV. Of the 63 with undiagnosed HCV, 7 were HIV+ (11.1%). Among the 724 patients without a prior HCV diagnosis, the adjusted odds ratio (AOR) of undiagnosed HCV was increased for: single vs married [2.24, CI 1.05, 4.81] and trauma vs AH GIM service [2.84, CI 1.04, 8.40]. A trend was observed for a greater AOR of HCV for UH GIM vs AH GIM service [2.18, CI 0.96, 5.63]. The AORs of unknown HCV were lower for age 17–39 vs >50 [0.27, CI 0.12, 0.57] and tended to be lower for women vs men [AOR 0.60, CI 0.34, 1.06]. Race was not associated. Of 2,850 persons admitted yearly to GIM and trauma without known anti-HCV+, we estimate that 250 are anti-HCV+ but are likely to be unaware of this infection.

CONCLUSION:  Roughly equal numbers of admissions to the study services had diagnosed HCV and apparently undiagnosed HCV. Undiagnosed HCV was more likely in older, single, male patients who were admitted to the university hospital's trauma and GIM service. Since nearly 10% of admissions had apparently undiagnosed HCV, these data support focused efforts to screen patients for HCV risk and offer testing to those with HCV risk factors.

HEPATITIS C VIRUS KNOWLEDGE AND PRACTICE PATTERNS AMONG PHYSICIANS WHO TREAT DRUG USERS: A NATIONWIDE SURVEY.A.H. Litwin1; H. Kunins1; K. Berg1; J. Arnsten1; I. Soloway1; A.D. Federman2; M.N. Gourevitch1. 1Albert Einstein College of Medicine / Montefiore Medical Center, Bronx, NY; 2Mount Sinai School of Medicine, New York, NY. (Tracking ID #116977)

BACKGROUND:  Screening for and management of hepatitis C virus (HCV) infection are critical for the health of patients with substance use disorders, yet little is known about HCV-related knowledge or practice patterns among physicians who care for drug users.

METHODS:  From 6/03–9/03 we conducted a survey of substance abuse treatment physicians currently providing care to drug users. The survey was mailed to a random national sample of members of the American Society of Addiction Medicine (N = 810), and to all physicians listed as addiction specialists on the rosters of the New York State Office of Alcoholism and Substance Abuse Services and the Medical Society of the State of New York (N = 123).

RESULTS:  The response rate was 54% (N = 506), of whom 375 reported currently providing care to substance users. Of these, 25% were internists, 24% were family practitioners, and 80% were board-certified in their specialty. Respondents reported a mean HCV prevalence of 32% among their patients. Physician knowledge of treatment eligibility criteria and efficacy was poor when compared to the recently revised National Institutes of Health 2002 recommendations. Sixty-five percent believed that patients actively using illicit drugs were not eligible for treatment, 56% believed that patients with normal LFTs were not eligible for treatment, and 84% did not understand that HCV genotypes 2 and 3 were associated with sustained viral response rates of over 75%. Physician knowledge of the natural history of HCV was also inaccurate; only 38% understood that a minority (fewer than 30%) of chronically infected patients develops cirrhosis after 20 years of infection. Despite these gaps in knowledge, physicians reported screening 75% of injection drug users for HCV antibodies, and counseling 92% of HCV-infected patients about alcohol consumption. Hepatitis A and B vaccinations were recommended to 53% and 59% of non-immune HCV-infected patients, respectively. Eighty-three percent of physicians reported referring at least one patient to a HCV specialist over the last 12 months, with a median of 10 referrals per physician.

CONCLUSION:  Although physicians who treat drug users report performing HCV-related screening, secondary prevention of liver disease and referral to treatment, we identified substantial gaps in knowledge regarding eligibility for and assessment of HCV treatment. To improve the health of HCV-infected persons, active efforts to improve HCV-related knowledge among physicians providing care to drug users are urgently needed.

HIGH PREVALENCE OF OSTEOPENIA IN OLDER MEN IS NOT ASSOCIATED WITH HIV INFECTION OR PROTEASE INHIBITOR USE.J.H. Arnsten1; R. Freeman1; A.A. Howard1; M. Floris-Moore1; R.S. Klein1. 1Montefiore Medical Center, Albert Einstein College of Medicine, Bronx, NY. (Tracking ID #115973)

BACKGROUND:  Osteopenia is a recently described adverse event in HIV-infected patients and a sequela of normal aging in both men and women. In men, aging accelerates bone loss, with a prevalence of osteoporosis in men older than 50 of 19%, approximately half that of women in the same age group. To date, no studies have described the prevalence of reduced bone mineral density (BMD) in older, HIV-infected men.

METHODS:  We analyzed BMD of the lumbar spine, hip, and total body in 343 older (age > 50) men, including 219 HIV-infected and 124 HIV-uninfected men. We also analyzed the impact of protease-inhibitor (PI) use on BMD.

RESULTS:  The median age of subjects was 54 y (range 50–81). 53% were Black, 27% were Hispanic, and 16% were white. 88% were unemployed, and 71% had graduated high school. Mean body weight was 79.6 kg (range 45–135) and mean BMI was 26.7 (range 18–55); only 31% reported strenuous exercise at least once per week. 90% of subjects had ever smoked cigarettes, and 70% were current smokers. 67% had a history of injection drug use, and this proportion did not differ by HIV status. Among HIV-infected men, 94% had taken antiretroviral therapy (ART) and 77% had taken PIs. The median duration of NRTI use was 36 months, and the median duration of PI use was 24 months (range 0–216 for both). Median CD4 count was 390 cells/mm3. The prevalence of osteopenia (lumbar spine, femoral neck, or total body T-score less than −1.0) among all men was 56% (n = 184), and the prevalence of osteoporosis (T-score less than –2.5) was 16% (n = 53). In univariate analysis, factors associated with osteopenia included lower weight (OR = 1.02, 95% CI = 1.00, 1.03, per kg decrease) and non-Black race (OR = 2.5, 95% CI = 1.6, 3.9). In addition, osteoporosis was associated with BMI (OR = 1.11, 95% CI = 1.03, 1.19 per kg/m2 decrease) and ever use of prednisone (OR = 4.0, 95% CI = 1.2, 13.3). In a linear regression model, non-Black race, lower BMI, and ever use of prednisone remained associated with reduced BMD of the spine after controlling for age, HIV infection, and smoking. We further found that neither HIV infection nor PI use was significantly associated with reduced BMD at any site.

CONCLUSION:  Osteopenia is highly prevalent among older men with or at risk for HIV infection, but is not associated with HIV infection or protease inhibitor use.

HIV PATIENT WORRIES AND SOURCES OF INFLUENCES ON INTENT TO TAKE NEWLY PRESCRIBED MEDICATIONS: ANALYSIS BY AGE.P.G. Clay1; R. Campo2; J. DeHovitz3; D. Johnson4; K. Squires4; W. Jordan5; S. Sajjan6; S. Narayanan6. 1University of Missouri-Kansas City, Kansas City, MO; 2University of Miami School of Medicine, Miami, FL; 3SUNY Health Science Center, Brooklyn, NY; 4USC, Los Angeles, CA; 5Drew University / Oasis Clinic, Los Angeles, CA; 6Merck & Co., Inc., West point, PA. (Tracking ID #115246)

BACKGROUND:  Recent data has been published demonstrating the increased likelihood for regimen failure due to non-adherence as age increases. As the population of new and established HIV infected persons continues to age with 21% of all new AIDS cases being age 45 or greater at year-end 2001, it is critical to better understand the influences and critical issues impacting the aging population's willingness to adhere to antiretrovirals.

METHODS:  HIV~PRACTICE Cooperative is a multi-center observational study of 304 adult HIV patients (pts) considered for a change in antiretroviral therapy. Pts completed a questionnaire on: worries about treatment, HIV disease knowledge, medication (med) adherence (missing at most one dose within the last three days), provider relationship, and influences on intent to take new meds. Pts were stratified by age <50 & ≥50. Demographic characteristics were examined using Chi-square tests, and differences in worries and sources of influences were ascertained using individual regression models, controlling for demographics, risk behavior, knowledge and physician effects.

RESULTS:  Forty five pts (15%, out of 298 pts with eligible data) were of age ≥50 yrs: 18% female; 51% self-identified as African-Americans; 27% Hispanics; and 18% as Caucasians. Fifty nine percent of these older pts reported annual income <$15000 and 37% reported not finishing high school. Compared to pts of age < 50 years, older pts were more likely to: report injection drug risk behavior (12% vs. 27%, P = .01), live alone (30% vs. 44%, P = .049) and lack adequate HIV knowledge (60% vs. 78%, P = .02). There were no significant differences in other pt demographics. In the regression models, older patients were more likely to report higher HIV disease related worries (P = .003), and less likely to be influenced by provider on their intent to take new meds (P = .04). There were no statistically significant differences observed in med worries, past med adherence and level of influence from media and family/friends on their intent to take new meds.

CONCLUSION:  Older patients reported higher HIV related health worries and lower influence of provider on intent to take newly prescribed medications. HIV providers may need to communicate more with these patients—focusing on disease knowledge and establishing a trusting relationship if maximal adherence to antiretrovirals is to be achieved.

HORMONE THERAPY USE AMONG INNER CITY WOMEN AFTER THE WHI.I.M. Helenius1; D. Korenstein1; E.A. Halm1. 1Mount Sinai School of Medicine, New York, NY. (Tracking ID #116680)

BACKGROUND:  In 7/02, the Women's Health Initiative (WHI) showed that long term risks of combination hormone therapy outweigh benefits. Preliminary work by others demonstrates a major shift in the use of hormone therapy (HT) among majority women since then. Less is known about how the WHI results have affected minority populations. This study measures patterns of HT use after the WHI among inner city women.

METHODS:  We conducted a cross-sectional survey of women in an inner city primary care clinic in New York City. All women who were taking HT (estrogen alone or in combination with progesterone) at the time of the release of the results of the WHI were eligible to participate. Subjects were identified by chart review and approached during regularly scheduled appointments. The survey was administered by trained interviewers, and designed to describe factors involved in the decision to continue or discontinue HT and to explore the experience of women who attempted to discontinue HT.

RESULTS:  Between 8/03 and 12/03, 53 of 75 approached patients agreed to participate (75%). Patients’ mean age was 59 years (range 45–73), 45% were Hispanic, 45% were African American, 43% had less than a high school education, 77% were unemployed, 17% lived with a partner. 83% had used estrogen therapy (ET) and 32% used combined estrogen-progesterone therapy (EPT). The mean duration of HT use was 10 years (range 2–31). Two-thirds said they had heard new information about the risks of HT. Sources of information were: television 41%, friends and family 24%, newspaper or magazine 23%, doctor 20%, internet 6%. Since the WHI, 83% patients attempted to stop HT (88% in the ET group and 80% in the EPT group) and 17% continued without attempting to stop. 64% of those who attempted to stop were successful and 36% restarted HT. Patterns were similar among ET and EPT users (P = .6). Among patients who attempted to stop, 48% said stopping was their doctor's idea and 41% said that it was their own idea. Three-fourths of those who attempted to stop, did so by stopping “cold turkey,” and one-fourth cut down over time. Increased symptoms were common after stopping: hot flashes 71%, fatigue 52%, insomnia 50%, memory loss 39%, mood swings 36%, depression 34%, vaginal dryness 32% and decreased libido 16%. Of those who attempted to stop, 23% used another prescription medication (primarily SSRIs), and 16% used a complementary or alternative medicine (soy products, black cohosh, and combination products) to relieve symptoms. 60% of successful stoppers wished they could continue HT.

CONCLUSION:  Since the WHI, the rate of discontinuation of ET and EPT has been high among minority women. However, one-third of women who attempted to stop HT restarted, due primarily to hot flashes. Use of alternative medications for symptom relief was low.

HOSPICE USE AMONG ASIAN AMERICANS DYING WITH CANCER.Q. Ngo-Metzger1; R.S. Phillips2; E.P. McCarthy2. 1University of California Irvine, Irvine, CA; 2Beth Israel Deaconess Medical Center, Boston, MA. (Tracking ID #116733)

BACKGROUND:  Our previous work showed that Asian Americans and Pacific Islanders (AAPI), especially the foreign-born, are less likely to use hospice than white patients, but little is known about hospice use among different AAPI ethnic groups.

METHODS:  We used the SEER-Medicare database to study white and AAPI patients diagnosed with cancer at age 66 or older in the Surveillance, Epidemiology, and End Results (SEER) program who died between Jan 1, 1988 and Dec 31, 1998. We linked demographic and cancer characteristics to hospice claims for patients diagnosed with lung, colorectal, prostate, breast, gastric, and liver cancer. Our sample consisted of 175,467 white, 2,145 Chinese, 3,510 Japanese, 1,781 Filipino, 856 Hawaiian, and 322 other Asian patients. We used proportional hazards regression to examine the relation between hospice enrollment and ethnicity after adjusting for patient demographics, place of birth, managed care insurance, year of diagnosis, type of cancer, stage at diagnosis, and tumor registry.

RESULTS:  Overall, 20% of patients enrolled in hospice, and use varied by race/ethnicity. The following table shows unadjusted enrollment rates and adjusted hazard ratio (aHR) and 95% confidence intervals for rates of hospice enrollment, comparing whites and AAPIs. aHRs <1.00 indicate lower rates of hospice enrollment among AAPIs, adjusting for demographic and clinical factors.

CONCLUSION:  Compared to white patients, AAPI subgroups have substantially lower rates of hospice enrollment.

Table 16. Hospice Use Comparing AAPI and White Patients
Race/EthnicityHospice Enrollment (%)aHR (95% CI)
Chinese18%0.62 (0.55–0.69)
Japanese20%0.67 (0.60–0.73)
Filipino15%0.61 (0.54–0.70)
Hawaiian24%0.78 (0.67–0.91)
Other Asian21%0.70 (0.55–0.90)
Foreign-Born16%0.97 (0.93–1.01)

HOSPITAL DISCHARGE TO A HOMELESS MEDICAL RESPITE PROGRAM PREVENTS EARLY READMISSION.S.G. Kertesz1; S. Swain2; M. Posner3; J. O’Connell2; M. Shwartz3; A. Ash3. 1University of Alabama at Birmingham, Birmingham, AL; 2Boston Health Care for the Homeless Program, Boston, MA; 3Boston University, Boston, MA. (Tracking ID #115295)

BACKGROUND:  Clinicians and care managers are often stymied by discharge planning for hospitalized homeless persons. In many cities, a medical respite program (unrecognized by most payors) receives the homeless who no longer need hospital care, but are not ready for the streets or shelters. We used data from the largest U.S. respite program to test whether discharge to respite was associated with reduced odds of early Readmission or Death (R/D), inpatient charges and inpatient days for the hospitalized homeless.

METHODS:  We examined R/D occurring within 90 days of discharge from a medical/surgical hospitalization 7/98–6/01, for 784 homeless adults. The predictor of interest was discharge destination, in 4 categories: Respite, Own Care (shelters/streets), Other Care (e.g. nursing home), and Left Against Medical Advice (AMA). We assessed the effect of discharge destination on R/D using logistic regression, and on inpatient days and charges with multiple linear regression. We adjusted for confounders including: comorbidity using Diagnostic Cost Groups (DCG), index hospitalization length of stay (LOS), alcohol and drug abuse, & sociodemographics. Data came from crosslinking 7 administrative sources.

RESULTS:  15% of Respite subjects experienced R/D within 90 days, versus 19% of those discharged to Own Care, 20% of those who Left Against Advice and 22% of those discharged to Other Care (P = .60). However, Respite subjects had longer index hospital LOS than Own Care subjects (mean = 6 vs. 3.5 days), and longer LOS predicted greater odds of early R/D. Risk-adjusted results are:

Compared to Own Care, Respite patients had 50% lower odds of early R/D (P = .03), 1 less inpatient day (P = .008), and $1,740 less inpatient charges (P = .03).

CONCLUSION:  For acutely hospitalized homeless persons, discharge to a homeless medical respite unit was associated with reduced odds of early readmission or death and reductions in inpatient days and charges. This homeless-customized service model may reduce utilization after discharge, and payors should consider reimbursing for it.

inline image

inline image

inline image

HOSPITALIZED PATIENTS’ COMFORT WITH AND PARTICIPATION IN MEDICAL ERROR PREVENTION. A.D. Waterman1; T.E. Burroughs2; J. Garbutt1; V.J. Fraser1; W.C. Dunagan1; B.M. Waterman3; W. Levinson4; T.H. Gallagher5. 1Washington University in St. Louis, St. Louis, MO; 2Saint Louis University, St. Louis, MO; 3Waterman Research Solutions, St. Louis, MO; 4University of Toronto, Toronto, Ontario; 5University of Washington, Seattle, WA. (Tracking ID #117091)

BACKGROUND:  Although published guidelines recommend that patients participate in medical error prevention, it is unknown whether patients are comfortable with and willing to take error prevention actions while hospitalized.

METHODS:  Two weeks after discharge, we conducted a telephone survey asking patients how comfortable they were taking 8 error prevention actions and whether they took these actions during their last hospitalization. We surveyed a random sample of 2,346 patients admitted to a Midwestern hospital system from August–November, 2003. Patients were predominately Caucasian (78%), between 40–79 years (63%), and female (58%), with non-emergency room hospitalizations (91%) requiring a 1–4 day hospital stay (62%).

RESULTS:  90% of patients agreed that they could help prevent errors and most were very comfortable asking what a medication is (91%), asking general medical questions (89%), asking staff to confirm their identity (84%), bringing someone to watch for errors (76%), telling staff that an error happened (74%), and helping staff mark the location of a surgical procedure (71%). Patients were not as comfortable asking a doctor or nurse to wash their hands (51% and 57% very comfortable, respectively). During their hospitalization, of the 161 patients who reported that a medical error occurred, 79% reported it to medical staff. For the other behaviors, 75% of patients (N = 2,346) asked general medical questions or what a medication was, but less than 38% took any of the other error prevention behaviors. Only 4% of patients asked doctors or nurses if they washed their hands. In a multivariate model, patients who were not Caucasian (P = .003) and patients who were in the hospital longer than 11 days (P = .04) were more likely to ask doctors if they had washed their hands than other patients.

CONCLUSION:  Hospitalized patients’ interest in and comfort with taking error prevention actions currently does not translate into action. Interventions to increase patient comfort with specific behaviors, particularly those that may seem rude to medical staff, are needed.

HOW “COMPLETE” ARE MEDICINE RESIDENTS’ EFFORTS AT INFORMED CONSENT FOR PROSTATE-SPECIFIC ANTIGEN SCREENING?M.H. Farrell1; J.M. Stein1; L.K. Ladouceur1; E.C. Chan2. 1Yale University, Waterbury, CT; 2University of Texas Health Science Center at Houston, Houston, TX. (Tracking ID #116573)

BACKGROUND:  Cancer screening with prostate-specific antigen (PSA) is controversial. Professional guidelines recommend informed consent before screening, but are unclear what to include in counseling. Chan et al. (Am J Med 1998;105:266–274) used experts and couples to define the “reasonable person” standard of consent for PSA screening to include 3 bare-minimum, 7 conversation and 7 brochure facts. We adapted this list into a 34-item structured implicit tool to assess “completeness” of counseling.

METHODS:  As part of a project to develop communication quality assessment tools, we taped visits between 39 internal medicine residents and 5 male 50-yo standardized patients trained to ask about PSA. Interviews were done without any preceeding teaching. We reviewed PSA transcripts using the structured implicit tool derived from the Chan study. To assess reliability, 2 authors reviewed 55% of transcripts and 3 authors reviewed 24%. We constructed a “completeness” score with weighted points as shown in figure 1.

RESULTS:  The 3 most common facts included were: an elevated PSA result may lead to other tests to see if cancer is present (61% residents), the bare-minimum fact false positive PSA result can occur (53%), and PSA screening is controversial (47%). The other 2 bare-minimum facts (false negative results and uncertainty about mortality benefit) were included by <20% of residents. The mean completeness score (reflecting all 3 categories) was 7.2 (SD 5.7) with a peak of 19. Of the 39 PSA transcripts, only 2 included all 3 bare-minimum facts; 57% included at least one. The mean numbers of conversation and brochure facts cited were 1.8 (SD 1.5) and 1.0 (SD 0.9) respectively, but 46% of residents cited at least one conversation fact and 59% at least one brochure fact.

CONCLUSION:  Completeness of PSA counseling varied in this resident sample, but did not consistently cover content considered bare-minimum for the reasonable person standard for informed consent. Educational programs and quality assessment tools may be needed to improve the way residents counsel patients about PSA screening.

inline image

HOW GOOD IS THE AGREEMENT BETWEEN PROSPECTIVELY AND RETROSPECTIVELY COLLECTED DATA COMPRISING THE PNEUMONIA SEVERITY INDEX?D.A. Aujesky1; R.A. Stone1; D.S. Obrosky1; D.M. Yealy1; T.E. Auble1; M.J. Fine1. 1University of Pittsburgh, Pittsburgh, PA. (Tracking ID #115790)

BACKGROUND:  Although the Pneumonia Severity Index (PSI) is a validated prediction rule used prospectively to guide the admission decision and retrospectively for risk stratification, no prior studies have compared the prospective to the retrospective application of the PSI. We sought to compare the agreement for individual prognostic variables and PSI risk class using data collected prospectively and by retrospective chart review in a large cohort of patients (pts) with community-acquired pneumonia.

METHODS:  We analyzed data from a randomized clinical trial to compare 3 PSI implementation strategies of low [LOW], moderate [MOD], and high [HIGH] intensity in 32 hospital emergency departments (EDs), to assess the agreement between data collected prospectively by ED personnel and retrospectively by trained chart abstractors in the same patients. Within each implementation arm, we examined the agreement for the 20 variables that comprise the PSI: 3 demographics, 10 history and physical examination findings, and 7 laboratory and radiographic variables. We also assessed the agreement for the 5 PSI risk classes into which pts were classified and whether pts were defined as low-risk (classes I-III) versus high-risk (classes IV and V). Concordance was measured using total percent agreement (% agree) and the unweighted kappa statistic for dichotomous variables. For comparison of PSI risk classes, we used the weighted kappa statistic.

RESULTS:  A total of 744, 1,171, and 1,305 pts were in the LOW, MOD, and HIGH intervention arms, respectively. The mean difference in prospectively and retrospectively collected pt age was 0 years (95% CI: −3 to 3 years). For all other prognostic variables, the % agree ranged from 84.8% to 100% and kappa ranged from 0.31 to 1.0 for the LOW arm; the corresponding ranges were 86.1% to 100% (% agree) and 0.36 to 1.0 (kappa) for the MOD arm, and 84.7% to 100% (% agree) and 0.21 to 1.0 (kappa) for the HIGH arm. For the 5 PSI risk classes, % agree was 92.3% (LOW), 93.7% (MOD), and 92.0% (HIGH) and the weighted kappa was 0.78 (LOW), 0.82 (MOD), and 0.78 (HIGH). In classifying pts as low versus high-risk, % agree was 86.6%, 89.3%, and 88.4%, and kappa was 0.70, 0.73, and 0.75 for the LOW, MOD, and HIGH arms, respectively.

CONCLUSION:  Individual variables comprising the PSI exhibit fair to substantial agreement between prospectively and retrospectively collected data. More importantly, there is a substantial agreement for assignment to the 5 PSI risk classes and classification as low versus high-risk, methodologically validating both prospective and retrospective application of the PSI.

HOW MANY REVIEWS DO YOU NEED? MEASURING THE QUALITY OF CARE AT CLINIC SITES WITH IMPLICIT PHYSICIAN REVIEW.T.P. Hofer1; S. Asch2; R.A. Hayward1; M.M. Hogan3; L.V. Rubenstein4; J.L. Adams5; E.M. Kerr1. 1University of Michigan, Ann Arbor, MI; 2West Los Angeles Veteran's Administration, Los Angeles, CA; 3Ann Arbor VAMC Center for Practice Management and Outcomes Research, Ann Arbor, MI; 4RAND, Sepulveda, CA; 5RAND, Santa Monica, CA. (Tracking ID #117044)

BACKGROUND:  We sought to develop a more reliable structured implicit chart review instrument for use in assessing the quality of care for chronic disease and to estimate how many reviews would be needed to provide a reliable estimate of quality of care for a clinic site.

METHODS:  We conducted a reliability study with records of both outpatient and inpatient care as the objects of measurement. 12 reviewers conducted a total of 1,666 reviews of 621 patient records selected from 26 VA clincal sites in in two regions of the country. Each patient had between one and four conditions specified as having a highly developed evidence base (diabetes and hypertension) or a less developed evidence base (chronic obstructive pulmonary disease or a collection of acute conditions). Multilevel analysis that accounts for the nested and cross-classified structure of the data was used to estimate the signal and noise components of the measurement of quality and the reliability of implicit review.

RESULTS:  For COPD and a collection of acute conditions the reliability of a single physician review was low (intra-class correlation = .16−.26) but comparable to most previously published estimates for the use of this method. For diabetes and hypertension the reliability is significantly higher at 0.46. The higher reliability is a result of the reviewers collectively being able to distinguish more differences in the quality of care between patients (P < .007) and not due to less random noise or individual reviewer bias in the measurement. The variance at the level of the clinical site was one third of that at the patient level within site and implicit reviews of between 24 and 33 records would give a quality rating with a reliablity of 0.8 at the level of the clinical site.

CONCLUSION:  For conditions with a well-developed quality of care evidence base, such as hypertension and diabetes, structured implicit review to assess the quality of care over a period of time could reliably detect differences in quality across practice settings with as few as 25 reviews per site of care and could be a reasonable complement or alternative to explicit indicator approaches for assessing and comparing quality of care. Structured implicit review, as well as explicit quality measures, must be used more cautiously for illnesses for which the evidence base is less well developed, such as COPD and acute, short-course illnesses.

HOW MUCH SHOULD CONVENIENCE COST?: A COST-EFFECTIVENESS ANALYSIS OF XIMELAGATRAN ANTICOAGULANT THERAPY.A.D. Auerbach1; M.C. Fang1; T. Minichiello1; G. Sanders2. 1University of California, San Francisco, San Francisco, CA; 2Duke University, Durham, NC. (Tracking ID #117140)

BACKGROUND:  Aside from pharmacodynamic advantages, new oral thrombin inhibitors such as ximelagatran have substantial proposed benefits for patients in that they will not require frequent monitoring of International Normalized Ratios (INR’s), thereby potentially improving patient quality of life and reducing laboratory costs. For the prevention of atrial fibrillation-related stroke and prevention of recurrent DVT, data suggest that ximelagatran may be as effective as warfarin without additional bleeding risk. However, ximelagatran will likely be considerably more expensive than warfarin.

METHODS:  We performed a cost-effectiveness analysis by developing 2 separate but similar state-transition Markov models; one posited a hypothetical cohort who required lifelong anticoagulation for atrial fibrillation (AF) and the other studied a cohort who required lifelong anticoagulation after recurrent deep vein thrombosis (DVT). In each model, patients received either warfarin or ximelagatran. Our base case assumed that these agents were similar in terms of efficacy and risk for bleeding, but patients receiving ximelagatran were assumed to require less frequent lab testing, and to have a normal quality of life while anticoagulated. We then modeled what level of incremental annual cost or potential efficacy differences would support widespread use of these agents in terms of quality adjusted life expectancy (QALY) and marginal cost-effectiveness ratios (MCER).

RESULTS:  Using base-case assumptions of an annual drug cost of $1,440 ($4/day) for ximelagatran and $237 ($0.65/d) for coumadin, use of ximelagatran in patients with AF produced longer quality-adjusted life expectancy (0.21 QALY), but at substantial marginal cost ($113,600/QALY); similar findings were observed in patients with DVT ($109,400/QALY). Assuming equal efficacy, MCER's for ximelegatran were less than $100K/QALY when the agent's annual costs were less than $1,300/yr (approximately $3.60/d) in both models; MCER's for ximelagatran were below $50K/QALY when drug costs were less than $900/yr. For every 1% relative improvement in efficacy or bleeding risk there was a drop of $5–7,000/QALY in marginal cost-effectiveness of ximelagatran for AF and DVT patients, giving MCER's < $50K/QALY if new agents were at least 14% more effective or safe than warfarin.

CONCLUSION:  Quality of life improvements and cost reductions related to less frequent INR monitoring may not make ximelagatran a cost-effective alternative to warfarin if ximelagatran costs substantially more than current therapy. Clinicians should feel comfortable prescribing these agents at higher cost, however, if ximelagatran is proven to be even moderately more effective or safe than warfarin.

HOW PRIMARY CARE PROVIDERS TALK TO PATIENTS ABOUT ALCOHOL: A QUALITATIVE STUDY.K. McCormick1; K.A. Bradley1; N.E. Cochran2; A.L. Back1; J.O. Merrill1; E. Williams1. 1University of Washington, Seattle, WA; 2Dartmouth College, White River Junction, VT. (Tracking ID #116949)

BACKGROUND:  Hazardous and problem drinkers have been shown to benefit from alcohol counseling by a primary care provider. However, most at-risk drinkers do not receive such counseling, and little is known about alcohol-related discussions that do occur. The purpose of this study was to describe the style and content of alcohol-related discussions between at-risk drinkers and their providers in order to inform future educational efforts.

METHODS:  We conducted qualitative analyses of 39 audiotaped visits to a VA General Medicine Clinic by patients who screened positive for at-risk drinking and discussed alcohol use with their providers. Both patients (N = 29) and providers (N = 14) were unaware of the study's focus. One investigator coded all audiotapes, and two others coded a subset of the tapes. To enhance trustworthiness, 4 other general internists evaluated the thematic coding scheme as it was developed, and reviewed the coding periodically.

RESULTS:  Three major themes have emerged. First, providers did not have a consistent method for assessing patients’ drinking practices, providing feedback, or offering advice. A second overarching theme was awkwardness on the part of the provider, as suggested by vague, fractured language, unclear advice, and abrupt, premature closure. Finally, differences between smoking- and alcohol-related discussions were observed. When addressing smoking, providers elicited specific information, and provided explicit feedback and advice. The smoking-related discussions also lacked the awkward elements that were noted during discussions about alcohol.

CONCLUSION:  Our findings suggest that providers are uncomfortable during alcohol-related discussions, and lack a standardized approach to address at-risk alcohol use. Smoking-related discussions differ in both regards, as providers appear better prepared to discuss smoking in a coherent and comfortable manner. This study highlights important elements to address in educational interventions that seek to increase rates of evidence-based alcohol-related counseling.

HOW USEFUL IS FOLLICLE STIMULATING HORMONE IN ASSESSING MENOPAUSAL STATUS?J.B. Henrich1; J. Hughes2; S. Kaufman3; D. Brody4. 1Yale University, Hamden, CT; 2The Orkand Coporation, Falls Church, VA; 3The National Institute of Child Health and Human Dev., Bethesda, MD; 4National Center for Health Statistics, Bethesda, MD. (Tracking ID #116807)

BACKGROUND:  Follicle stimulating hormone (FSH) is the most commonly used marker of the menopause; however, its usefulness has been questioned. We used data from the National Health and Nutrition Examination Survey (NHANES 1999–2000) to identify factors associated with FSH levels, and to assess the efficacy of FSH in distinguishing between women in the reproductive, menopausal transition and postmenopause stages.

METHODS:  NHANES measures the health and nutritional status of the civilian, non-institutionalized U.S. population using a stratified, multistage sampling procedure. Data are collected through a standardized in-person household interview, and physical and laboratory examinations. Participants eligible for our study were non-pregnant, non-lactating women aged 35 to 60 years without surgical menopause. Standard staging criteria were used to classify women into reproductive stages based on self-reported menstrual history. Independent variables included demographic data, Body Mass Index, cigarette smoking, alcohol use, physical activity, and menopausal hormone use. Outcome measures were: 1) factors associated with FSH in multivariable analysis; and 2) Receiver Operator Characteristic (ROC) curve-derived FSH cutoff points between stages.

RESULTS:  609 women comprised the study sample. Age, reproductive stage, BMI and menopausal hormone use were significantly associated with FSH in multivariable analysis. ROC curves identified optimal FSH cutoff points between the reproductive and menopausal transition stages [FSH = 13 mIU/mL, Likelihood ratio (LR) += 5.8], and between the menopausal transition and postmenopause stages (FSH = 45 mIU/mL, LR += 2.6); the former value was a better indicator of the distinction between stages.

CONCLUSION:  When combined with the menstrual history, an FSH of 13 mIU/mL may be useful in distinguishing between women in the reproductive and menopausal transition stages; an FSH of 45 mIU/mL is less useful in separating the menopausal transition and postmenopause stages. Clinicians should consider age, menstrual history and FSH collectively when assessing a woman's reproductive stage.

inline image

HOW WELL DO CLINIC-BASED BLOOD PRESSURE MEASUREMENTS AGREE WITH RIGOROUS STANDARDS?J. Kim1; H.B. Bosworth2; C. Voils2; M.K. Olsen3; T.K. Dudley3; M. Gribbin3; E.Z. Oddone2. 1Duke University School of Medicine, Durham, NC; 2Duke University Medical Center, Durham, NC; 3Durham VA Medical Center, Durham, NC. (Tracking ID #116445)

BACKGROUND:  There is variability in blood pressure measurements. To optimize the accuracy of blood pressure readings, clinical trials implement a rigorous protocol to reduce measurement bias. However, for outcome studies that evaluate effectiveness of interventions and for physicians who make hypertensive treatment decisions, the blood pressure readings are clinic-based, with less rigorous standards. This study was designed to compare three methods of blood pressure measurements: random-zero sphygmomanometer, digital oscillometric device, and clinic-based methods.

METHODS:  Prospective study of 100 patients with an ICD-9 diagnosis of hypertension and an upcoming appointment with their Duke primary care physician. Patients were screened over the phone and scheduled to meet with a research assistant 60 minutes before their doctor visit, or if unable to meet before, directly after their doctor visit. Following strict protocol, four blood pressure measurements were obtained: two with the Hawksley Random Zero (RZ) sphygmomanometer and two with the BPM-100 digital device in an order that was randomized. The clinic-based blood pressure was taken by the clinic RN and entered into the Duke computerized medical record system during the same doctor visit within an hour of the standard assessments.

RESULTS:  The patients were 77% female, 78% White, 20% Black, 2% Asian, with a mean age of 64. There was excellent agreement between the RZ readings and the digital readings: the intra-class coefficient (ICC) for systolic blood pressures (SBP) = 0.974 (95% confidence interval: 0.965–0.982), diastolic blood pressures (DBP) = 0.947 (0.923–0.964). There was also excellent agreement between the RZ readings and the clinic-based readings for SBP, ICC = 0.906 (95% CI: 0.832–0.944), but less so for DBP, ICC = 0.774 (0.619–0.860). Clinic-based readings overestimated RZ readings, with a mean overestimation of 8.3 mm Hg for SBP and 7.1 mm Hg for DBP.

CONCLUSION:  The RZ and digital devices showed excellent agreement for both systolic and diastolic blood pressures. However, the agreement between the RZ and clinic-based measures was not as good, particularly for DBP. Measurement differences of this magnitude may have implications for outcomes research and quality of care studies that generally rely on clinic-based measurements.

HUNGER AND FOOD INSECURITY IN A VETERAN PATIENT POPULATION.R. Bobba1; E.L. Arsura2. 1Salem VAMC, University of Virginia, School of Medicine, Salem, VA; 2Salem VAMC, Salem, VA. (Tracking ID #116726)

BACKGROUND:  Although previous clinical observations suggest that some patients in an urban setting experience hunger and food insecurity, there are limited data on the prevalence of hunger in a veteran patient population.

METHODS:  To determine the prevalence of hunger and food insecurity in an ambulatory veteran patient population we performed a cross-sectional survey. We surveyed a sample of patients who were seen in either ambulatory general medicine or medical subspecialty clinics over a four week period. Main outcome measures were rates of hunger, food insecurity and associated factors.

RESULTS:  Of 400 interviewed patients, 301 (participation rate, 75%) completed the survey either in general medical clinics (n = 145) or in medical subspecialty clinics (n = 156). Of the patients surveyed 49 (16.3%) respondents reported some measure of hunger or food insecurity. 24 (7.9%) reported hunger and missing a meal or abstaining for a whole day or longer. 11 (3.6%) reported not eating because they could not afford food. 15 (4.9%) cut down the size of meals in the previous 12 months and 17 (5.6%) did not have the kind of food they wanted. 16 (5.3%) either worried about whether food would run out before they obtained money to replenish food supplies or put off paying other bills so they would have money for food. 9 (2.9%) used emergency services to obtain food. The population of individuals who suffered from hunger or food insecurity were on average younger than those without hunger (61.8 ± 10.2 years vs. 68.8 ± 10.1 years P < .008) and had a lower mean income ($11,000 vs $13,800 P < .05). There was no difference in other assessed parameters between the 2 groups including race, sex, home owner ship or level of education attained.

CONCLUSION:  Hunger and food insecurity are common among patients seeking care at our veterans facility. Affected individuals are younger and have a lower mean income. There are mental health implications and it needs to be seen if hunger and food insecurity impact physical health.


BACKGROUND:  To identify individuals who are at high-risk for colon and breast cancer, an adequate and complete family history assessment is required. Guidelines for mammogram and colon cancer screening specify initiating screening at younger ages for individuals with a positive family history; and criteria for genetic testing for both hereditary nonpolyposis colon cancer (HNPCC) and hereditary breast-ovarian cancer (HBOC) syndromes rely almost exclusively on family history information. Little data exists evaluating the adequacy of the family history interview in primary care for these purposes.

METHODS:  A retrospective chart review of 995 consecutive new patient visits to 28 primary care physicians. Visits were identified by CPT codes and 93% of visits included family history information. Data was abstracted from chart review regarding the family cancer history information obtained for colon, breast, ovarian, or any HNPCC cancer (endometrial, stomach, small bowel, hepato-biliary, renal-pelvis, ureter, brain).

RESULTS:  Family history information on cancer diagnoses was collected on 68% (679/995) of the patients. Four hundred and twelve individuals identified 679 affected first- or second-degree relatives. Age at diagnosis was documented for first and second-degree relatives in only 51% of relatives affected with colon cancer, 38% of relatives affected with breast cancer, 26% of relatives affected with ovarian cancer, and 5% of relatives affected with a HNPCC-associated cancer. No patients (ages 40–49 years with a high-risk family history for colon cancer) were identified whom might be candidates for initiating colonoscopy screening before age 50. Nine women who were ages 30 to 39 years with a high-risk family history for breast cancer were identified as candidates for initiation of mammography screening before age 40. Fifty-six percent were recommended for mammography screening. No patients fulfilled Amsterdam II criteria for genetic testing in HNPCC, although two patients had pedigrees that were suggestive yet lacked data on age at diagnosis. Six women had pedigrees that met published criteria for referral for BRCA1 testing and one patient (17%) was referred.

CONCLUSION:  Age at cancer diagnosis was frequently missing from family history assessments, which could have a potential impact on identification of high-risk individuals. When pedigree information does identify high-risk individuals, slightly more than half of high-risk women eligible for early mammography screening undergo screening and few patients are referred for genetic testing.

IDENTIFYING LEARNER NEEDS IN GERIATRIC PRESCRIBING.A.M. Wolf1; N.B. May1; B.R. Huot1; J.M. Schectman1; J.D. Voss1; M.L. Plews-Ogan1; J.M. Evans1. 1University of Virginia, Charlottesville, VA. (Tracking ID #116849)

BACKGROUND:  The high incidence of adverse drug events (ADEs) in the elderly, together with the burgeoning geriatric population, highlights the importance of assuring that internists are competent in geriatric prescribing. To date no geriatric prescribing needs assessment has been nationally disseminated to inform internal medicine educators.

METHODS:  A focus group of internal medicine residents (PGY1–3) was assembled to address the topic of geriatric prescribing. Results of the focus group were used to develop a survey, distributed to all categorical and primary care track residents in a university residency program. Survey questions measured level of preparedness to prescribe for the elderly in 7 domains (5-point scale, 1 = unprepared, 5 = very prepared), and asked residents to rate their agreement (5-point scale) with statements about prescribing confidence, concern regarding missing ADEs, and polypharmacy.

RESULTS:  Nine residents participated in the focus group and 65 (79%) responded to the survey. The focus group identified several major areas of perceived need: (1) strategies to reduce polypharmacy (“we’re taught how to add, not subtract”) (2) how to incorporate metabolic changes of aging into prescribing (3) psychotropic drug prescribing in the elderly, especially Alzheimer's medications (4) better utilization of outpatient pharmacists for both education and quality assurance (5) access to geriatricians in residents’ primary care training setting (6) strategies to better detect ADEs. Survey respondents reported the following levels of preparedness (from least to most prepared): prescribing dementia medications (mean response 2.7); integrating age-related changes into prescribing (3.0); prescribing pain medications (3.3); prescribing antidepressants (3.5); reducing polypharmacy (3.6), and awareness of high-risk drugs (3.7). Thirty-six percent of residents were concerned they were missing serious ADEs in their elderly patients. Most (74%) felt that polypharmacy was a significant problem among their patients. Only 3% of residents felt very confident in their overall geriatric prescribing ability. Prescribing confidence improved significantly with level of training in all domains except analgesic and antidepressant prescribing. Concern regarding missing serious ADEs declined with training level (P = .04), but concern regarding polypharmacy did not vary by training level ( = .5). Residents who felt that there is little difference between caring for elderly and non-elderly patients were less likely to identify polypharmacy as a significant problem in their elderly patients (P = .02).

CONCLUSION:  Curricular efforts in the domain of geriatric prescribing should focus on prescribing dementia and pain medications, adjusting prescribing to age-related metabolic changes, reducing polypharmacy, and detecting ADEs. Geriatricians and pharmacists should be better integrated into residents’ usual practice settings.

IMPACT OF A MASS MEDIA CAMPAIGN TO IMPROVE ANTIBIOTIC USE.R. Gonzales1; K. Corbett2; S. Wong3; A. Deas2; B. Leeman-Castillo2; J. Maselli1; A. Sebert-Kuhmann2; J. Glazner4; R.S. Wigton5. 1University of California, San Francisco, San Francisco, CA; 2University of Colorado, Denver, Denver, CO; 3Denver Health and Hospital Authority, Denver, CO; 4University of Colorado Health Sciences Center, Denver, CO; 5University of Nebraska System, Omaha, NE. (Tracking ID #116754)

BACKGROUND:  Office-based educational interventions are effective, but costly, strategies for reducing unnecessary antibiotic use in community practices. We evaluated the impact of a low-cost ($88,628 for media placement) mass media campaign—“Get Smart: Use Antibiotics Wisely”—on office visits and antibiotic prescriptions for acute respiratory tract infections (ARIs).

METHODS:  The campaign was launched Nov 1, 2002, and included billboards (n = 5), bus tails (n = 40), bus stop posters (n = 20), interior bus signs (n = 150), National Public Radio spots (n = 60), and airing (n = 112) of a Spanish language public service announcement. Office visit and pharmacy claims data were provided by 4 large managed care organizations (MCOs) prior to (Nov 01–Feb 02) and following (Nov 02–Feb 03) the intervention. Office visits for ARIs were based on ICD-9 codes for nonspecific URI, pharyngitis, sinusitis, bronchitis, otitis media and pneumonia. Separate sample, random-digit-dialed telephone surveys asked respondents (in English or Spanish) if they had made a medical office visit for any reason during the past 3 months. Tests of significance included mixed-effects models for comparing office visit and antibiotic prescription rates, and multivariable logistic regression for comparing proportions of telephone survey respondents with a recent office visit before and after the campaign.

RESULTS:  Following the campaign, pediatric office visits for ARIs declined to a greater extent in the intervention area (−28%) compared with the control area (−17%) (P = .002) (intervention-attributable MCO visits avoided = 2,594). Conversely, adult office visit for ARIs declined the same amount in the intervention area (−22%) compared with the control area (−24%) (P = .61). A similar pattern was detected by telephone survey. Among parents of young children (age < 6 yrs), there was a decrease in office visits reported by respondents residing in the intervention area (−9%) compared with the control area (+21%) (P = .04; adjusting for age, gender, Spanish language and insurance status). There was no significant difference between groups in office visit rates reported by respondents who were not parents of young children (+5% vs. +9%) (P = .64). Pediatric and adult antibiotic prescription rates for ARIs ranged between 37% and 45%, and did not differ significantly between areas.

CONCLUSION:  A low-cost mass media campaign reduced pediatric office visits for ARIs—the primary target of the campaign—but not antibiotic prescribing behavior. Analysis of pharmacy-level antibiotic dispensing data is needed to confirm that the decrease in office visits corresponds to a decrease in community-level antibiotic use.

IMPACT OF A TEACHING TOOL: A STRUCTURED DEBATE ABOUT SCREENING MAMMOGRAPHY FOR WOMEN IN THEIR FORTIES.I.M. Helenius1; C.E. Goldstein2; D. Korenstein1. 1Mount Sinai School of Medicine, New York, NY; 2University of Calgary, Calgary, Alberta. (Tracking ID #115584)

BACKGROUND:  In order to effectively counsel women in their 40's about the decision to undergo screening mammography, physicians must be able to discuss the potential risks and benefits of the test. We designed and implemented an evidence based educational program using structured debate to 1) improve residents’ ability and willingness to discuss controversial issues regarding mammography with their patients, and 2) increase residents’ confidence in clinical decision making regarding screening mammography.

METHODS:  We conducted a prospective cohort study on the effect of a teaching intervention during ambulatory block month on 2nd year medical residents between 1/2003 and 9/2003. The intervention consisted of a 3-hour structured debate on the merits of mammography for women in their 40’s. Six debates were held over the course of 6 months, with approximately 6 residents participating in each debate. Each debate was structured around three questions: 1) What does the data show? 2) What are the properties of the test and the disease that impact screening? 3) What do patients expect and prefer? Questionnaires utilizing 4- and 5-point Likert scales were administered 2 weeks prior to and 3 months after each debate. Pre- and post-debate answers were compared using the Wilcoxon paired rank test using each resident as his/her own control.

RESULTS:  Thirty-three of 36 residents participated in the debates. Of those, 100% completed the pre- and post- questionnaires. 97% agreed that the debate was an effective teaching tool. Almost all reported routinely recommending screening mammography to women in their 50's before and after the debate (100% pre-debate, 94% post-debate, P = .67). They were divided on the question of whether they routinely recommend screening mammography to average risk women in their 40's both before and after. (56% agreed and 44% disagreed pre-debate, 51% agreed and 49% disagreed post-debate, P = .19). After the debate, 18 residents (54%) became more confident that their practices regarding screening mammography for women in their 40's were supported by the literature, and only 5 (15%) became less confident (P = .03). When asked about their comfort discussing mammography screening with their patients 16 (48%) reported an increase, 14 (42%) no change, and 3 (9%) a decrease (P = .02). 25 (76%) became more likely to discuss screening mammography with their patients (P < .001).

CONCLUSION:  An evidence based, structured debate on the pros and cons of screening mammography for women in their 40's effectively improves residents’ attitudes about discussing this controversial topic with their patients.

IMPACT OF ADHERENCE, KNOWLEDGE, AND QUALITY OF LIFE ON ANTICOAGULATION CONTROLN.J. Davis1; H.H. Billett1; H.W. Cohen2; J.H. Arnsten1. 1Montefiore Medical Center, Albert Einstein College of Medicine, Bronx, NY; 2Albert Einstein College of Medicine, Bronx, NY. (Tracking ID #115048)

BACKGROUND:  An estimated one-third of patients attending anticoagulation clinics are poorly controlled and at risk for thromboembolic events. However, it is unknown how adherence to warfarin, patient knowledge about warfarin therapy, and perceived impact of warfarin therapy on quality of life affect anticoagulation control.

METHODS:  We conducted a cross-sectional survey of patients attending two anticoagulation clinics serving inner city communities in the Bronx, NY. Adherence to and knowledge about warfarin therapy were assessed using multi-item scales, and scale responses were dichotomized to create categories of good and poor. Good knowledge of anticoagulation was defined as ≥70% of correct responses on an 18-item knowledge questionnaire, and good adherence was defined as 100% of adherent responses on a 4-item scale. Good anticoagulation control was defined as having ≥70% of INR values (obtained within 60 days of survey completion) in therapeutic range. Associations of anticoagulation control with adherence, knowledge and quality of life were tested using chi square analyses.

RESULTS:  Among 52 respondents, mean age was 50.9; (SD = 16.9) and 64% were female. Only 14% had good anticoagulation control. Mean number of INR values obtained within 60 days was 4.9 (SD = 2.1), and the mean percentage of INR values within therapeutic range was 42% (SD = 27). Good adherence was reported by 50% of participants and was significantly associated with good anticoagulation control (OR = 9.6, 95% CI = 1.1, 84.5, P = .01). 37% of participants had good knowledge of anticoagulation. 19% of participants responded that warfarin therapy had a negative impact on their quality of life. Knowledge and quality of life were not significantly associated with good anticoagulation control.

CONCLUSION:  The percentage of patients with good anticoagulation control in this inner city anticoagulation clinic was lower than that reported in the general population. Patients who self-reported good adherence were more likely to have good anticoagulation control. Using a simple self-report measure may assist in identifying patients at risk for poor anticoagulation control, and should lead to closer follow-up of non-adherent patients. Patient knowledge of therapy and perceived impact on quality of life were not associated with good anticoagulation control. Interventions to increase adherence are needed to improve anticoagulation control and better realize the benefits of warfarin therapy among inner city, indigent patients.


BACKGROUND:  Although the incidence of invasive cervical cancer decreased in past years in the United States, cervical cancer continues to disproportionately affect Hispanic women. Hispanic women have one of the lowest Papanicolaou (Pap) smear screening rates. Low English proficiency among Hispanics may impose additional barriers, like poor communication and time constraints, on health care providers to recommend screening.

METHODS:  We analyzed data from the 2000 National Health Interview Survey. We performed two separate analyses. Our first analysis assessed the impact of English language proficiency on Pap smear use; we included Hispanic women age 18 or older and excluded women who have had a hysterectomy. Our second analysis assessed the impact of patients’ English language proficiency on physician recommendation of Pap smears. We limited this sample to Hispanic women, age 18 or older, without a Pap smear in the past three years and excluded women who have had a hysterectomy and those who had not seen a physician in the past year since they could not have possibly received a recommendation by the health care provider during that period.

RESULTS:  Our analysis of Pap smear use in the past three years included 2,337 women; 55.9% were highly proficient English speakers. In a logistic regression model, highly English proficient Hispanics were more likely to report having a Pap smear in the past three years (OR = 1.52, 95% CI 1.10–2.10) as compared to the low proficient. Other factors significantly associated with having a Pap smear were household income of $20,000 or higher (OR = 1.50, 95% CI 1.15–1.95), having a usual source of care (OR = 1.92, 95% CI 1.38–2.63), and having health insurance (OR = 1.72, 95% CI 1.25–2.38). A total of 314 Hispanic women were included in the analysis of health care provider recommendation for Pap smear in the past year. In the adjusted analysis, highly English proficient Hispanics were more than two times more likely to receive a recommendation for a Pap smear during the past year as compared to less proficient Hispanics (OR = 2.08, 95% CI 1.01–4.27).

CONCLUSION:  Low English language proficiency is a barrier to receive cervical cancer screening and physician recommendation for Pap smears among Hispanics. Intervention aimed at health care providers such as cultural sensitivity training, educational campaigns identifying populations at greater risk of no recommendation and remainder systems prompted by low proficiency patients may increase Pap smear recommendation in Hispanic women.


BACKGROUND:  Healthy people given a new diagnosis may develop a “labeling” phenomenon, manifested by increased anxiety, disability and doctor visits despite being no more ill than their peers. Such behavior has been documented in women following a false positive mammogram (FP). We sought to determine if this phenomenon is generalizable to the US population at large by employing a novel use of the Medical Expenditures Panel Survey (MEPS).

METHODS:  MEPS is a nationally representative survey of health care use and expenditures. Information available includes demographics, diagnoses, health status, use of medical services, charges and payments. All women who received a bilateral screening mammogram in 1996–98 were identified. Those with breast cancer and those with records that did not extend at least 6 mo prior to and 12 mo after the index mammogram were excluded. A FP was defined as any mammogram followed by a breast related procedure within 3 mos or a repeat mammogram within 9 mo. All others were defined as a true negative (TN). Non breast related office visits and charges for the 6 mo preceding and two 6 mo periods following the index mammogram were tallied and compared using a multiple regression model adjusting for visit rate prior to the index mammogram, age, race, perceived health status, census region and income.

RESULTS:  1828 individuals met the inclusion criteria of whom 206 (11.3%) met the definition for a FP. Average age for the group was 55 (range 17–90) and 86% were white. Prior to the index mammogram, the visit rate for both the FP and TN groups was 3.25 visits/yr. Following the mammogram, the mean visit rate for the FP group was 4.50 and for the TN group was 3.49. (P < .0001 in the adjusted model). Median charges for the FP and TN groups prior to the index mammogram were $219 and $240. Following the mammogram, charges for the FP group were $540 versus $390 in the TN group. (P < .017).

CONCLUSION:  In this nationally representative sample, women who had a FP were substantially more likely to visit their physicians in the year following the test than women with a negative test, above and beyond the need to evaluate the abnormal test. This is consistent with the hypothesis that false positive tests can lead to a labeling phenomenon and that this can have significant clinical impact. This phenomenon should be recognized by clinicians and potentially play a role in physician-patient interactions before and after a screening mammogram.

IMPACT OF LOW NUMERACY ON PATIENTS WITH HUMAN IMMUNODEFICIENY VIRUS INFECTION.E. McNeill1; C.A. Estrada1; C. Ransdell1; J.M. Spillane1; J.C. Byrd1. 1East Carolina University, Greenville, NC. (Tracking ID #116108)

BACKGROUND:  Numeracy, the ability to use basic probability and mathematical concepts, has not been studied in patients with human immunodeficiency virus infection (HIV). We assessed numeracy skills in patients with HIV infection and explored the relationships between numeracy, patients’ general knowledge of HIV infection, and ability to accurately recall their own CD4 counts and viral loads.

METHODS:  In a cross-sectional study of 101 patients with HIV infection, we measured numeracy using a modified Schwartz questionnaire (6 questions) and knowledge of HIV infection with the HIV KQ-18 (a validated questionnaire). The numeracy and the knowledge questionnaires were verbally administered. We used the kappa statistic to compare reported CD4 counts and viral loads with chart abstracted values.

RESULTS:  Eighty-one percent of the patients were African American, and 51% were women. Twenty-four percent were employed, and 79% had incomes less than $15,000 per year. Most patients had a low numeracy score, Table. Patients with lower numeracy scores had lower scores of HIV knowledge (P < .0001) and a trend towards decreased ability to accurately recall their own CD4 counts and viral loads (lower concordance, kappa).

CONCLUSION:  Low numeracy is prevalent in patients with HIV infection. Patients with lower numeracy scores had poorer understanding of their disease process and a trend towards poorer recall of their CD4 counts and viral loads.

Table 17. Table
Numeracy ScorePatients N (%)HIV Knowledge, KQ-18CD4 KappaViral Load Kappa
Low 0–252 (51.5)11.80.430.22
Medium 3–440 (39.6)15.00.580.63
High 5–6 9 (8.9)15.40.56

IMPACT OF MULTIDISCIPLINARY CASE MANAGEMENT ROUNDS ON INPATIENT COST OF CARE.D.G. Bell1; R. Stroebel1; E. Tangalos1; P. Targonski1. 1Mayo Clinic, Rochester, MN. (Tracking ID #117193)

BACKGROUND:  Effective health care team communication and care plan management are critical to providing optimal care for hospitalized patients. We expanded our multidisciplinary approach to care by formally instituting case management “rounds” on a daily basis for patients admitted to an academic primary care internal medicine (PCIM) hospital service. Goals were to improve clinical outcomes by enhancing communication and coordination of care by members of the multidisciplinary team. Participants in the case management rounds included attending physicians and unit nurses, physical therapists, social workers, and the consulting pharmacist. Brief case management rounds were conducted daily on the patient unit to discuss each patient's current and expected plan of care and potential barriers affecting that plan, including dismissal planning issues. The objective of this study is to assess the effect of this strategy on length of stay and cost of care.

METHODS:  A retrospective analysis was performed utilizing hospitalization data on all 2,486 admissions to the PCIM hospital service during the year 2002. 1,492 patients (60%) were admitted to nursing units utilizing the multidisciplinary case management strategy (CMS). 994 patients (40%) were admitted to other general internal medicine nursing units which were equivalent in staffing and structure, but did not utilize the CMS. These admissions were designated as controls. The same physician teams attended patients across all nursing units, and all other elements of care were uniform between units. Outcomes including length of hospital stay (LOS) and cost of care (net operating income, NOI) were assessed for each management group and controlled for diagnosis related group (DRG) utilizing weighted 2-sample t tests.

RESULTS:  After adjusting for admission diagnosis, weighted mean LOS was 0.51 days shorter (SE = 0.5, t = 9.2, P < .0001) and NOI was $445 greater (SE = $54.13, t = 8.2, P < .0001) for nursing units utilizing the multidisciplinary CMS. The greatest LOS decreases and NOI gains were seen for congestive heart failure, stroke, COPD, and musculoskeletal related DRGs.

CONCLUSION:  This study demonstrates the positive financial impact of a strategy utilizing daily multidisciplinary case management rounds on a general internal medicine teaching service at a major academic medical center. Results suggest that optimizing communication and coordination of care reduces length of stay and leads to improvement in cost of care. Future studies will assess the effect of this strategy on quality of care.

IMPACT OF ORGANIZATIONAL CLIMATE AND HECTIC OFFICE ENVIRONMENT ON PHYSICIAN STRESS AND ERROR IN PRIMARY CARE.M. Linzer1; L. Baier Manwell1; J. Bobula1; M. Mundt1; E. Williams2; B. Horner-Ibler1; A. Maguire3; J. McMurray1; B. Man4; M. Plane1; M. Schwartz5; A. Varkey4. 1Univ of Wisconsin, Madison, WI; 2Univ of Alabama, Tuscaloosa, AL; 3Medical College of Wisconsin, Milwaukee, WI; 4Rush Univ, Chicago, IL; 5New York Univ, New York, NY. (Tracking ID #117240)

BACKGROUND:  The impact of organizational climate on medical outcomes is not well understood. The MEMO study (Minimizing Error, Maximizing Outcome) investigates this question through a conceptual model relating working conditions to quality of care.

METHODS:  MEMO is a longitudinal study of physicians and patients from NYC, Chicago, Milwaukee, Madison and rural Wisconsin. Physicians completed a survey with: 1) scales measuring stress, satisfaction, and mental health; 2) scales modified from those of Kralewski to assess organizational climate (OC); 3) questions on past errors; and 4) the OSPRE (Occupational Stress and PReventable Error) measure to assess likelihood of future error. Clinic managers provided data on office structure. We used split sample exploratory and confirmatory factor analyses to identify OC domains, and regression analysis to determine predictors of stress, past errors, and future error. Simple correlation analyses assessed association of key structural variables with physicians’ perception of office environment.

RESULTS:  Of 420 physicians from general medicine and family medicine, 38% described their office environment as busy tending toward chaotic, and another 10% as hectic or chaotic. Although 79% indicated job satisfaction, 61% agreed work was stressful, 27% noted burnout symptoms, and 31% were moderately or more likely to leave their jobs within 2 years. Domains of OC included leadership values (alpha = .86), quality emphasis (.86), organizational trust (.79), information/communication (.68) and cohesiveness (.66). Chaotic atmosphere predicted stress, dissatisfaction, and poorer mental health (all P = .001). Lack of quality emphasis was linked to past errors (P < .005). Likelihood of future error (OSPRE alpha = .87) was higher in practices with less emphasis on information (P < .02). Physician outcomes were also associated with age, gender, ethnicity, work hours, work control, inadequate resources, and a lesser emphasis on diversity. Strong correlates of a hectic or chaotic office included larger numbers of vulnerable (uninsured and Medicaid) patients (r = .41, P < .001) and fewer exam rooms (r = 39, P < .001), but not lower staffing ratios (r = −0.18, P = .12).

CONCLUSION:  Physician stress is prevalent in primary care and is increased in offices with hectic environments. Additional exam rooms may improve hectic offices. Safety and quality in primary care could also be enhanced by emphasizing quality over quantity of care, and with improved communication and information systems.

IMPACT OF PAST PHYSICAL ACTIVITY ON CURRENT PHYSICAL ACTIVITY AND RISK OF CORONARY HEART DISEASE.M. Conroy1; N. Cook2; J.E. Buring3; I. Lee3. 1Massachusetts General Hospital, Boston, MA; 2Harvard University, Boston, MA; 3Brigham and Women's Hospital, Boston, MA. (Tracking ID #115049)

BACKGROUND:  Middle-aged and older women who are physically active have decreased risk of coronary heart disease (CHD). However, little is known about how physical activity during youth influences physical activity and CHD risk later in life.

METHODS:  We addressed these questions using data from the Women's Health Study, which comprises 39,876 healthy US health professionals, age ≥ 45 years. At baseline, women reported their physical activity and other variables influencing physical activity or CHD risk. At 24 months, they also reported their physical activity in high school and age 18–22 years. During follow-up over 9 years, 477 women developed CHD. We used multivariable logistic regression to determine the association between physical activity in young adulthood and meeting physical activity recommendations (i.e., 30 minutes of moderate activity on most days of the week) at middle-age. We used proportional hazard regression to estimate the multivariate-adjusted relative risks of CHD according to physical activity during: 1) middle-age; 2) young adulthood; and 3) both middle-age and young adulthood.

RESULTS:  Among 37,169 eligible participants, the most active women (vigorous physical activity 10–12 months/year) during young adulthood were more than twice as likely to meet physical activity recommendations at middle-age than the least active (no vigorous activity) (multivariate-adjusted OR = 2.43; 95% C.I. 2.24, 2.63). At baseline, the most active women (≥1,500 kcal/week) had a 39% lower risk of CHD than the least active (<200 kcal/week) (multivariate-adjusted RR = .61; 95% C.I. 0.46, 0.81). However, the most active women during young adulthood did not have a lower risk of CHD occurring during middle-age and older when compared to women who were the least active during young adulthood (multivariate-adjusted RR = 0.81; 95% C.I. 0.58, 1.14). A combined measure of young adulthood and middle-aged physical activity showed no additive effect.

CONCLUSION:  Women who are physically active during young adulthood are more likely to be active when middle-aged and older. Physical activity during middle-age predicts lower risk of CHD. Public health initiatives and health care providers should encourage women at all ages to become more physically active to establish healthy habits in younger women and reduce risk of CHD in middle-aged and older women.

IMPACT OF PERCEIVED DISCRIMINATION ON USE OF PREVENTIVE SERVICES.A.N. Trivedi1; J.Z. Ayanian1. 1Brigham and Women's Hospital, Boston, MA. (Tracking ID #116120)

BACKGROUND:  Prior research has documented disparities in the use of health services based on race, gender, and insurance status. We examined the frequency of perceived discrimination in the health care system, and the relation between perceived discrimination and the receipt of preventive services.

METHODS:  We analyzed the 2001 California Health Interview Survey, a random–digit telephone survey, to assess the relation between the perception of having been discriminated against in a health care setting over the past 12 months and receipt of the following preventive health services over this time period: aspirin use and cholesterol testing among people with heart disease or hypertension, foot examination and HbA1C testing among people with diabetes, colon cancer screening for adults >age 50, PSA testing among men >age 50 , flu vaccination among adults >age 65, mammography among women age 50–79, and Pap testing among women age 18–65. We used multivariate logistic regression to control for multiple demographic and clinical characteristics with SUDAAN software.

RESULTS:  Of the 54,968 subjects, 2,830 (4.7%) reported experiencing discrimination in a health care setting over the past 12 months, most frequently related to insurance type (27.4), race (13.7%), or income (6.6%). Female gender, non-white race, lack of insurance, rural residence, low income, lack of high school education, limited English proficiency, presence of comorbid medical conditions, increasing number of physician visits, fair or poor health status, current tobacco use, unmarried status, and high BMI were significantly associated with reporting discrimination (all P < .05). Persons who reported discrimination were less likely than those not reporting discrimination to have had HbAIC testing (78.5% vs. 86.1%P = .035), cholesterol testing (71.0% vs. 79.2%P < .01), mammography (78.4% vs. 83.6%P = .03) and PSA testing (37.1% vs. 45.7%P = .03.) There was no significant difference in aspirin use (33.4 vs. 35.2%P = .38), foot examination (60.8% vs. 67.6%P = .12), colon cancer screening (41.8% vs. 44.0%P = .26), flu vaccination (62.5% vs. 67.5%P = .10) and Pap testing (87.7% vs. 87.1%P = .61). After adjustment, there was no significant relation between perceived discrimination and use of preventive services except for cholesterol testing (OR 0.69 (0.55–0.88), P < .01).

CONCLUSION:  The adjusted association between perceived discrimination and decreased use of preventive health services was relatively small in magnitude and significant only for cholesterol testing. Given the low prevalence of reported discrimination, this factor is unlikely to account for a large portion of observed disparities in receipt of appropriate preventive services.

IMPACT OF REDUCED DUTY HOURS ON EDUCATIONAL SATISFACTION OF INTERNAL MEDICINE HOUSESTAFF.A.R. Vidyarthi1; P. Katz1; R.M. Wachter1; A.D. Auerbach1. 1University of California, San Francisco, San Francisco, CA. (Tracking ID #117371)

BACKGROUND:  In July 2003, the Accreditation Council for Graduate Medical Education (ACGME) limited resident duty hours to 80 per week. The impact of these duty hour restrictions on educational satisfaction of housestaff is unknown.

METHODS:  This cross-sectional study surveyed 125 Internal Medicine Residents rotating through the three hospitals affiliated with the University of California, San Francisco. The main outcome measures included: self report of perceived value of educational activities, frequency of administrative/clerical tasks interfering with educational activities, and educational satisfaction after the implementation of duty hour reduction.

RESULTS:  Overall response rate was 76%. Residents rated educational activities such as Morning Report, teaching others, work rounds and attending rounds most highly. Answering pages, paperwork, waiting on hold on the telephone, and scheduling tests and appointments were the most frequent barriers to engaging in the educational activities. After restriction of duty hours, the amount of time spent in educational and administrative activities did not change. Sixty-eight percent of residents stated that decreased duty hours had no impact or a negative impact on education. In multivariable models, residents with the lowest overall work satisfaction were PGY-1s (P = .007), those who reported feeling overwhelmed at work (P < .0001), and those who reported working more than 80 hours per week (P = .015). However, only PGY-1 residents (P < .05) and those who felt overwhelmed with work (P = .01) were less satisfied with education.

CONCLUSION:  In this academic internal medicine training program, we found that duty hour reduction did not improve educational satisfaction or time spent in educational activities. This may be due, in part, to the unchanged burdens of administrative tasks that compromise residents’ ability to engage in educational activities. Systematic changes to resident work-life will be needed if duty hour reductions are to improve educational satisfaction.

IMPACT OF RELIANCE ON HELICAL CT ON THE DIAGNOSIS OF PULMONARY EMBOLISM: A BAYESIAN ANALYSIS.S.R. Ranji1; K.G. Shojania1; R.L. Trowbridge2; A.D. Auerbach1. 1University of California, San Francisco, San Francisco, CA; 2Maine Medical Center, Portland, ME. (Tracking ID #116774)

BACKGROUND:  Helical CT scan has become the primary test used to investigate pulmonary embolism at many institutions, despite limited evidence to support its use in this fashion. We sought to determine the probability of false-negative and false-positive diagnoses of pulmonary embolism by comparing clinicians’ decisions to withhold or institute anticoagulation to the calculated post-test probability of pulmonary embolism.

METHODS:  We reviewed the charts of 322 adult patients who underwent helical CT as the initial diagnostic test for suspected PE between 1998 and 2000 at two affiliated teaching hospitals. Pre-test probability of PE was calculated using a validated clinical prediction rule. Post-test probabilities were calculated using published estimates of CT test characteristics and pre-test probabilities; sensitivity analyses within ranges of these estimates were also performed. The post-test probabilities were then compared to clinicians’ treatment decisions.

RESULTS:  66 patients were treated for PE (20.5%). Most patients had either a low (n = 184, 57.1%) or moderate (n = 101, 31.4%) pre-test probability of pulmonary embolism. 96.5% of patients who had a positive CT scan were treated for PE, and 95.4% with a negative scan were not treated. Using published estimates of CT test characteristics, only 0.4–3.9% of patients with a negative CT scan likely had an untreated pulmonary embolism (i.e. a false-negative scan). However, approximately 28% of patients with a CT scan read as positive for PE may have been unnecessarily anticoagulated due to a false-positive scan.

CONCLUSION:  Clinicians at our institution regarded helical CT results as definitive, anticoagulating virtually all patients with a positive CT scan and withholding anticoagulation in virtually all patients with a negative CT. Our analysis reveals that such reliance on helical CT results may lead to an acceptable false-negative rate, but an unacceptably high false-positive rate. Failure to appropriately apply Bayesian reasoning when interpreting helical CT results may lead to unnecessary anticoagulation of a substantial proportion of patients.

IMPACT OF RESIDENT DUTY HOUR REFORM IN A MEDICINE CORE CLERKSHIP: A WORK-SAMPLING STUDYJ.R. Kogan1; L.M. Bellini1; J.A. Shea1. 1University of Pennsylvania, Philadelphia, PA. (Tracking ID #116223)

BACKGROUND:  In July 2003, residency programs were required to make significant structural changes to comply with the ACGME's resident duty hour regulations. While the impact most effects residents, it is important to know the impact on medical students’ educational experiences. This study's purpose was to determine, using random work sampling, if there are differences in medicine core clerkship students’ daily activities, who they interact with, and the perceived educational value of their experiences before and after implementation of resident duty hour regulations.

METHODS:  Almost all students enrolled in our inpatient medicine core clerkship pre (n = 38, 93%) (3/31–6/20/03) and post (n = 37, 95%) (9/29 -12/19/03) duty hour reform agreed to participate. Participants wore a random reminder pager in the hospital, set to signal approximately every 90 minutes for one week. For each signal, students completed a survey asking their activity (direct patient care, indirect patient care, education, other), who they were with and the activity's educational value (1 = low; 5 = high). Students recorded the times they arrived to and left the hospital daily. The proportion of events spent in each activity (event proportion) was calculated and then averaged across all students pre and post reform. Analysis of variance was used to assess group differences.

RESULTS:  Sixty-nine students (92%) returned surveys. 804 and 912 surveys were completed pre and post reform respectively. The mean number of surveys completed per student per week was 24.9 (s.d. 7.6); the mean number per day was 4 (s.d. 2.5). No significant differences existed pre and post reform in event proportions for direct patient care (.13 vs .15, P = .15), indirect patient care (.35 vs .32, P = .13), and education (.38 vs .36, P = .47). Students pre and post reform were equally likely to be alone (.38 vs .34, P = .12), with their resident (.31 vs .32, P = .72), their attending (.14 vs .16, P = .34), another resident (.11 vs .12, P = .83) and another intern (.16 vs .15, P = .81), but more likely to be with their own intern post reform (.33 vs .39, P = .01). No differences between groups existed in the perceived educational value of direct patient care activities (3.7 vs. 3.8, P = 42 ), education activities (4.0 vs. 4.1, P = .16), or time in the hospital (9.4 vs 9.1 hours/day; P > .05) pre and post reform. Indirect patient care activities were more valued pre reform (3.4 vs. 3.2, P = .03).

CONCLUSION:  Despite significant changes in the structure of our residency program in the post work hour reform era, there were few differences, as assessed by random time sampling, in the time students spent in the hospital, who they interacted with, the proportion of events spent in various activities and their perceived educational value. Generalizability of these findings requires further study.

IMPACT OF THE WOMEN´S HEALTH INITIATIVE RESULTS ON PATIENT´S TRUST IN MEDICAL RECOMMENDATIONS.M.A. Schonberg1; R.B. Davis1; C.C. Wee1. 1Beth Israel Deaconess Medical Center, Boston, MA. (Tracking ID #116696)

BACKGROUND:  Contrary to prior evidence, the Women's Health Initiative (WHI), published in 7/02, found that the overall health risks of hormone replacement therapy (HRT) exceeded the benefits. We examined how women on HRT felt about the WHI results and whether it affected their trust in medical recommendations.

METHODS:  From 7/03 to 9/03, we perfomed a telephone survey of women randomly selected from a large academic practice who were >50 years and taking HRT as of 7/09/02. We obtained data on sociodemographics, HRT use, WHI awareness, beliefs and emotions about its findings, and confidence in medical advice before and after the WHI. To assess trust, we asked how the HRT experience affected women's willingness to take a new drug recommended by their doctor for prevention of heart disease, osteoporosis, and disease in general, and whether women were more likely to ask about side effects of medications. Using bivariable and multivariable analyses, we identified whether factors such as age, race, education, stopping HRT, or WHI awareness, were associated with being less likely to take a new drug and more likely to ask about the side effects of drugs.

RESULTS:  Of 315 eligible women, we interviewed 204 pts (response rate 65%). The mean age of the 204 pts was 61 years, 70% were white, 56% had at least a college degree, 75% had been on HRT for >5 years, 47% were on estrogen alone, and 70% stopped using HRT. Almost all were aware of the WHI (94%), primarily through the media (86%); 82% believed the results of the study. Although 27% reported experiencing a negative emotion such as disappointment or anger after learning the trial's results, 88% remained confident in medical advice. While considering their HRT experience, 31% of pts were less likely to take a new drug to prevent heart disease, 28% were less likely to take one to prevent osteoporosis, and 15% were less likely to take one to prevent disease in general; 45% were more likely to ask about drug side effects than before. After adjustment, having stopped HRT was significantly associated with being less likely to take a new drug to prevent heart disease (aOR 3.1, 95% CI [1.4–6.6]) and to prevent disease in general (4.2 [1.2–14.7]). Pts with at least a college degree were more likely to ask about drug side effects than less educated pts (2.0 [1.1–3.5]). We did not identify any factors associated with being less likely to take a new drug to prevent osteoporosis.

CONCLUSION:  Although the WHI reversed recommendations regarding HRT use in postmenopausal women, most women believed the results, were not upset by its findings, and remained confident in medical advice. However, stopping HRT was associated with being less likely to take a new drug to prevent heart disease or disease in general. For these women, their experience with HRT may have affected their trust in medical recommendations for the use of preventive medications.

IMPACT OF VIDEOCONFERENCING ON NIGHTTIME, ON-CALL MEDICAL DECISION-MAKING IN THE NURSING HOME.M. Weiner1; G. Schadow2; D. Lindbergh2; J. Warvel2; G. Abernathy2; S. Perkins2; J. Daggy2; P. Dexter2; C. McDonald2. 1Indiana University Center for Aging Research, Indianapolis, IN; 2Regenstrief Institute, Inc., Indianapolis, IN. (Tracking ID #117387)

BACKGROUND:  Nursing-home residents with medical problems at night often have no direct access to physicians. Off-site physicians order needless tests or referrals based on limited information relayed via telephone by nurses. We developed a videoconferencing system for physicians to evaluate directly residents with acute incidents. We assessed physicians’ ratings of decision-making and satisfaction with videoconferencing and determined medical conditions conducive to videoconferencing in the nursing home. We have reported preliminary findings and now report final results. We hypothesized that physicians would rate decision-making as easier in most of the video encounters.

METHODS:  Residents of a county-managed nursing home were enrolled and studied between June 2001 and June 2003. We assembled a portable videoconferencing workstation for use in the nursing home and provided 7 attending physicians with home workstations and commercial broadband Internet service. We assessed reasons for calls from the nursing home to on-call physicians between 05:00 PM and 01:00 AM, when a research assistant was available. Physicians received automated prompts to conduct videoconferencing, which they could decline when they felt certain it would not aid clinical management (e.g., simple clarification about an order). Following each video, physicians rated technical difficulties, ease of medical decision-making, and overall satisfaction with the video (scale of 1 to 5, 5 = very satisfied).

RESULTS:  We studied 461 video-eligible participants (59% female, 55% African-American, mean age 61 years), 386 of whom generated 1,399 calls from nursing staff to physicians after business hours. Physicians judged 87 of 99 video encounters. Using a scale of 1 to 5, physicians rated videos with a 3 or greater in 75% of cases; the most common rating was 4 (49%). Calls most often occurred for questions about orders (25%), the laboratory's report of abnormal lab values (22%), admission to the facility (14%), or nurse's or aide's note of worrisome examination findings (11%). Lab values and questions about orders accounted for only 8% (each) of videos. Falls or injuries accounted for 3.9% of calls but 11% of videos, and pain or dyspnea accounted for 4.4% of calls but 16% of videos. Technical difficulties were noted in 42% of cases, but physicians indicated that decisions were easier with videos in 71%.

CONCLUSION:  Despite technical difficulties, nighttime videoconferencing between nursing-home residents and physicians leads to easier medical decision-making when the alternative is no direct contact with residents. In this setting, videoconferencing is useful to evaluate falls, injuries, pain, and dyspnea. More reliable and portable technology is now available and may provide even greater usefulness of this application.

IMPAIRED HEALTH FUNCTION AMONG WOMEN EXPERIENCING CURRENT INTIMATE PARTNER VIOLENCE.J.M. McCauley1; J.C. Campbell2; H. Abushomar3; L. McNutt4; D.E. Ford2. 1Johns Hopkins Community Physicians, Baltimore, MD; 2Johns Hopkins University, Baltimore, MD; 3University of Albany (State University of New York), Albany, NY; 4University at Albany, SUNY, Albany, NY. (Tracking ID #117320)

BACKGROUND:  Though Intimate Partner Violence (IPV) is experienced by 2–4 million US women a year, it is unclear how IPV affects health function. Our objectives were to compare currently abused to non-abused young women to determine: 1) health outcome score differences 2) dose response relationship between severity of abuse and health function.

METHODS:  A representative sample of women (ages 18–44) attending a primary care, community health center in Albany NY completed a computer-administered survey. Emotional, physical and sexual abuse was measured with a modified Composite Abuse Scale (CAS), with 0–3 response = no abuse, 4–10 score = moderate abuse and > 11 = severe abuse. Physical and mental health functions were measured with the SF12. We analyzed this cross-sectional data using frequencies and chi square analyses.

RESULTS:  Of the 293 respondents, 53% were 18–29, 65% had <12 years education; 63% were African Americans; 18% had experienced moderate, 17% had severe abuse. Of the 12 SF items, 10 showed significant differences. EMOTIONAL/PHYSICAL/SEXUAL ABUSE SELECTED SF12 ITEMS (*% in each) None Mod High p Self rated health (good/fair/poor) 57.6* 68.3 65.8 0.369 Climbing stairs (limited a little or lot) 36.7 51.3 63.4 0.003 Accomplished: less due to physical health (yes) 21.1 47.5 55.3 0.000 less due to emotional problem (yes) 26.8 43.6 50.0 0.008 Pain interfered with work (little to lot) 42.8 69.1 85.7 0.000 Lots of energy (little to none of the time) 94.9 88.1 88.1 0.118 Physical/Emotional health interfere with social life (little-lot) 51.6 73.8 92.7 0.000

CONCLUSION:  There was a dose-response relationship for IPV and both physical and mental function. Young women with low levels of function should have an assessment for IPV.

IMPROVING ANTIBIOTIC PRESCRIBING FOR AMBULATORY PATIENTS WITH SINUSITIS.S.J. Bernstein1; M. Daoud1; R.V. Harrison1; C. Standiford1. 1University of Michigan, Ann Arbor, MI. (Tracking ID #117330)

BACKGROUND:  Acute sinusitis can range from acute viral rhinitis to acute bacterial sinusitis. Symptoms resolve within 2 weeks without antibiotics in 70% of cases and with antibiotics in 85% of cases. Trimethoprim/sulfamethoxazole (TMP/SMX) and amoxicillin have been shown to be superior to placebo in treating patients with acute bacterial sinusitis. Although other antibiotics have been compared to these agents, none have been shown to be superior. Despite this evidence there is wide-spread use of other broad-spectrum antibiotics for patients with sinusitis. We evaluated the effectiveness of a program to reduce non-recommended antibiotic use for patients with acute sinusitis.

METHODS:  We conducted a quasi-experimental study comparing antibiotic prescribing before (1/2000–6/2001) and after (7/2001–12/2002) distributing physician-specific antibiotic prescribing data. During the latter period a natural experiment occurred when one department permitted academic detailing, conducted by two physicians, and a second did not. Data were collected on 4,153 and 3,820 managed care patients with sinusitis in the two periods, respectively, who were treated by 112 physicians working in 14 Family Medicine and General Medicine out-patient clinics of an academic health care system. Antibiotics were classified as recommended (TMP/SMX, amoxicillin), second-line alternatives (e.g., amoxicillin/clavulanate, cefuroxime) and not recommended (e.g., azithromycin) according to the institution's sinusitis guideline. To control for time-trends, we assessed antibiotic prescribing for upper respiratory infection (URI) for which there were no guidelines. Data were collected on 5,589 and 5,709 managed care patients diagnosed with a URI in the two periods, respectively. Diagnoses were obtained from encounter forms and antibiotics from a managed care organization's pharmacy files.

RESULTS:  Antibiotics were prescribed to 67% of patients with sinusitis before feedback and to 57% of patients after physician-specific feedback (P < .001). Similarly, antibiotic prescribing for URI decreased from 20% to 14% (P < .001). The use of non-recommended antibiotics for sinusitis decreased from 24% to 15% among physicians receiving feedback and academic detailing (P < .001) and from 11% to 8% among physicians who received feedback only (P = .09).

CONCLUSION:  While academic detailing was associated with reduced use of non-recommended antibiotics it had no significant impact on overall prescribing. Careful evaluation of guideline implementation projects is necessary to avoid falsely attributing success to a program.

IMPROVING DEPRESSION CARE: SYSTEMATIC REVIEW OF MULTIFACETED INTERVENTIONS IN PRIMARY CARE SETTINGS.M.S. Gerrity1; J.W. Williams2; S.K. Dobscha1; J. Deveau2; T. Holsinger2; B.N. Gaynes3; B.K. Moser2; A.J. Dietrich4. 1Portland VA Medical Center, Portland, OR; 2Duke University Medical Center, Durham, NC; 3University of North Carolina at Chapel Hill, Chapel Hill, NC; 4Dartmouth Medical School, Hanover, NH. (Tracking ID #116435)

BACKGROUND:  Depression is a prevalent illness with poor outcomes in primary care settings. A variety of multifaceted interventions to improve depression outcomes have been evaluated. We conducted a systematic review of the literature to determine to what extent multifaceted interventions improve depression outcomes in primary care and to define key components, patients likely to benefit, and required resources.

METHODS:  We searched Medline, HealthSTAR, CINAHL, PsycINFO, and a specialized registry of controlled trials from 1966 to December 2002 and bibliographies of pertinent articles. Searches were limited to the English language. We also consulted experts. Included studies met the following criteria: 1) study involved primary care patients; 2) intervention had multiple components enhancing direct patient care; 3) design was a randomized controlled trial; 4) outcomes included depression severity. Pairs of investigators independently abstracted information regarding the 1) setting and subjects, 2) components of the intervention, 3) care management system, and 4) outcomes. Authors were contacted for key missing data.

RESULTS:  Ten of the 12 interventions improved depression symptoms over 3 to 12 months (absolute increase of 16% to 30.6% in patients with 50% improvement in symptoms). The two studies that did not improve this outcome involved VA patients and may have differed in other ways. Interventions had 3 to 4 components (decision support, self-management support, delivery system redesign, information systems) and required additional resources to implement. Care management included 4 functions (education and support, monitoring symptoms and adherence, self-management support, psychological treatments) and 3 processes (duration, number of contacts, mental health supervision). Interventions varied greatly within these functions and processes.

CONCLUSION:  Multifaceted interventions that include care management can improve outcomes of depression. Research is needed to identify essential intervention elements and subgroups of patients who benefit from additional elements (e.g., psychological treatments).

IMPROVING MEDICAL RESIDENT FATIGUE MAY REQUIRE MORE THAN WORK HOUR LIMITS.J. Friedman1; S. Vazirani2; J. Tillisch1; N.S. Wenger1. 1University of California, Los Angeles, Los Angeles, CA; 2Greater Los Angeles VA Healthcare System, Los Angeles, CA. (Tracking ID #117409)

BACKGROUND:  Concern over the effect of fatigue on quality of care led to rules capping the number of permissible resident work hours, with the assumption that fewer work hours will lead to more sleep time. But little is known about the relationship of number of hours of work and sleep and perceived fatigue and its effect on patient care. We investigated resident work and sleep behavior, how this related to perceived fatigue and the relationship to faculty lifestyle and perceptions.

METHODS:  We conducted a survey prior to mandatory implementation of new work hour restrictions with internal medicine residents in one training program and physicians who attend on the medicine wards. The survey asked about the precise hours spent in various activities over the past 7 days, fatigue and the perceived effect of fatigue on activities in and out of the hospital. We compare attending and resident physician time working and sleeping, and perceptions of rest adequacy.

RESULTS:  Fifty-two residents (70% response) and 46 attending physicians (68% response) completed the survey. Residents (mean age 29, 56% men) reported that during the prior week they worked a mean of 57 hours (63 hours if on a ward or ICU rotation) and had 13 hours of non-clinical work and 50 hours of sleep (7.1 hours per night). This left a mean of 48 hours per week for other activities, during which most residents reported at least one hour of exercise (54%), medical reading (76%) and pleasure reading (54%). Compared to residents, attending physicians (mean age 43, 62% men) reported working more hours (70, P < .01) but the same amount of sleep (50 hours) during the prior week. Forty-six percent of residents reported that they did not get enough rest, 48% reported being overworked, and many reported a moderate to high chance of falling asleep with common activities: reading 63%, sitting in a public place 27% and sitting in a medical conference 50%. Most residents felt that they would learn better (84%), be less stressed (92%) and make fewer errors (50%) if they got more sleep. Contrary to resident views, attending physicians felt that residents got enough rest (79%) and were not overworked (72%), although many attending physicians felt that residents would learn better (45%), be less stressed (70%) and make fewer errors (44%) if they got more sleep.

CONCLUSION:  Although residents worked less then current maximum work hours, they described substantial fatigue and perceived important implications from lack of rest. Attending physicians underrated levels of resident fatigue, perhaps because they reported more work hours than their residents. Reducing internal medicine resident fatigue will likely require changes beyond limiting work hours.

IMPROVING OSTEOPOROSIS SCREENING AMONG ELDERLY WOMEN.B.K. Muma1; J. Elston Lafata2; E. Peterson2; D.M. Kolk2. 1Henry Ford Hospital Detroit, Detroit, MI; 2Henry Ford Health System, Detroit, MI. (Tracking ID #116805)

BACKGROUND:  The US Preventive Services Task Force and the National Osteoporosis Foundation recommend that women age 65 years or older be screened routinely for osteoporosis. Current estimates are that only 20% of this population has received a bone density test. Thus, a large number of elderly women are at risk of avoidable late-stage osteoporosis and its consequences. We report the preliminary findings of a randomized prospective study designed to evaluate strategies to enhance the quality of care for osteoporosis in this high risk population.

METHODS:  We identified women aged 65–89 that received their primary care from 15 ambulatory care clinics staffed by a large medical group in southeast Michigan between 4/1/2001 and 3/31/2003. The clinics were stratified by size and Bone Mineral Densitometry (BMD) equipment availability, and randomized to one of 3 types of osteoporosis disease management strategies: (1) usual care; (2) mailed patient letters; and (3) mailed patient letters plus physician prompts. Women that received care from the clinics who had a BMD test, a diagnosis for osteoporosis, or a prescription drug claim for an osteoporosis-specific therapy (alendronate, risedronate, raloxifene, or calcitonin) were excluded. A total of 10,565 eligible women were included in the study: The remaining 3,110 women in strategy (1) clinics; 3,368 in strategy (2) clinics; and 4,087 in strategy (3) clinics. Statistically significant differences in BMD testing rates were evaluated using generalized estimating equation approaches.

RESULTS:  Preliminary analyses indicate a BMD screening rate of 2% in the usual care group, compared to 2% in the mailed patient letter group and 9% in the mailed patient letter plus physician prompt group. Differences between usual care and patient mailed reminders alone were not statistically significant (P = .67). The use of patient mailed reminders plus physician prompts lead to significant improvements in BMD testing compared to both usual care and patient mailed reminders alone (P < .001). Results did not change with adjustments for clinic, race or both.

CONCLUSION:  Preliminary results indicate that the use of mailed patient reminders in combination with physician prompts can lead to significant increases in osteoporosis screening compared to usual care and the use of mailed patient reminders alone.

IMPROVING OUTPATIENT CHART DOCUMENTATION BYMEDICAL RESIDENTS.C.M. Vergara1. 1Hartford Hospital, Department of Medicine, Hartford, CT. (Tracking ID #116389)

BACKGROUND:  Introduction: Very little data exists that describe the quality of chart documentation by medical internal medicine residents during their weekly, outpatient, continuity clinic. The quality of chart documentation may serve as a surrogate measure of the quality of medical care that is provided. Hypothesis: Documentation of patient encounters by medical residents are suboptimal and amenable to simple quality improvement interventions. Clinical Setting: Hospital-based, ambulatory center that delivers primary care to a predominantly underserved population and serves as the one of the major outpateint teaching sites for a university-based internal medicine residency program.

METHODS:  Intervention: A paper flyer in the form of a tabular checklist is placed on the front of the chart for all patients seen by residents who have their continuity clinic practice on Monday and Wednesdays for two consecutive weeks. No paper flyer was provided for those residents who have their clinic sessions on Tuesday, Thursday and Friday. This latter group served as the comparative control group. The checklist in the paper flyer prompted reminders to complete five quality measures of the level of written chart documentation. The five quality variables are as follows: (1) updated medical problem list (2) updated medication list (3) updated health maintenance log (4) legible signature or name of the resident documenting the patient encounter (5) documentation of the supervising attending with whom the case was discussed. A single technician audited all charts seen by medical residents during the two-week period to assess completion of the five aforementioned variables.

RESULTS:  142 of 161 charts were audited (89% and 87%, intervention and control group, respectively). The intervention group demonstrated significant improvements in all of the quality measures. Documented entries for the problem list, medication list and health maintenance log were 59.2%, 60.5% and 26%, respectively, in the intervention group. The comparative percentenges for the control group were 42.6%, 42.6% and 6.6%, respectively. 77% versus 62.2% of intervention and control group, respectively, documented that the patient encounters were discussed with the supervising physician. Similarly, 76.9% vs. 63.9% of the written medical progress note had legible signatures in the intervention and control group, respectively.

CONCLUSION:  A simple and inexpensive paper flyer can improve the level and quality of the medical documentation by medical residents during their regularly scheduled continuity clinic practice. Whether the improved documentation translates to improved medical care and clinical outcomes is not known. Future follow up studies are needed to study the clinical consequneces of improved medical documentation.

IMPROVING PHYSICIAN AND PATIENT COMMUNICATION SKILLS IN OUTPATIENT SETTINGS: LESSONS LEARNED FROM INTERVENTION RESEARCH.J.K. Rao1; L.A. Anderson2. 1Centers for Disease Control and Prevention (CDC), Atlanta, GA; 2Centers for Disease Control and Prevention, Atlanta, GA. (Tracking ID #115969)

BACKGROUND:  Effective physician-patient communication is associated with satisfaction with care and better health outcomes. We conducted a systematic review of interventions designed to enhance physicians’ and patients’ communication skills in outpatient clinical settings.

METHODS:  We searched 13 computerized databases (i.e., Medline, Psychlit, Cochrane Collaboration, etc.) and performed hand searches of the reference lists of eligible articles. Eligible studies: 1) tested an intervention designed to facilitate patient or physician communication skills; 2) targeted the intervention(s) toward patients, physicians, or both; 3) were RCTs; 4) included an objective, quantitative assessment of verbal communicative behavior as the primary study outcome; and 5) were published in English between 1966 and 2001. Verbal communicative behaviors were classified into: information gathering, information providing, information verifying, interpersonal skills, or global evaluation of communication. We assessed the effect of the intervention on verbal communication and selected secondary outcomes.

RESULTS:  Twenty-eight studies met the eligibility criteria. Twelve intervened on physicians only, 13 intervened on patients only, and 3 intervened on physicians and patients. Twenty-one studies were published after 1990 (10 physician, 8 patient, 3 physician and patient). Twenty studies involved primary care encounters whereas 8 involved specialty visits. Twenty-one studies examined information gathering skills as the primary verbal outcome. Of these, the experimental groups demonstrated more question asking behavior than the control group in 12 studies. In 3 of 6 studies, the experimental group showed improvements in information verifying skills compared to controls. In 8 of 10 studies, the experimental groups received higher global ratings of interpersonal communication skills than controls. Of 13 studies assessing satisfaction, 2 studies demonstrated an increase, 1 demonstrated a decrease, and 10 demonstrated no difference in satisfaction between the experimental and control groups. In the 9 studies assessing visit duration, the interventions showed little impact [experimental versus control: increased visit length (n = 2), decreased (n = 1), no effect (n = 6)].

CONCLUSION:  There was considerable variation in the communication behaviors studied. Overall, the interventions demonstrated positive effects on communication skills, particularly in studies involving physicians. These interventions demonstrated little positive impact on patient satisfaction or visit duration. Further theoretical and practical work is necessary to understand how the specific types of communication behaviors may relate to patient outcomes.

IMPROVING THE EFFICIENCY OF BREAST CANCER RISK ASSESSMENT IN PRIMARY CARE.R.C. Burack1; N. Patel2; J. George1; R. Everson1. 1Wayne State University, Detroit, MI; 2Karmanos Cancer Institute, Detroit, MI. (Tracking ID #117366)

BACKGROUND:  Chemoprevention with tamoxifen can reduce the 5-year risk of breast cancer (BCa) by 47% among women whose Gail model risk is 1.7%. However, the efficient identification of these higher-risk women is a challenge in primary care. We assessed the performance of a risk factor-based triage system designed to identify women for whom a complete Gail risk assessment might be unnecessary.

METHODS:  BCa risk factor data was collected for 18,817 women at the time of mammography and a Gail score calculated for each. We included 2 factors in a triage system (family history and biopsy history) based on their strength of association with BCa in the Gail model. Women with either factor were classified “triage +”; those with neither were “triage-”. We calculated the maximum Gail score possible within age-ethnicity strata, assuming that family history and biopsy history were absent (“triage-”) but that all other factors were at their “highest risk” level. Within age-ethnicity strata we compared the maximum expected and observed Gail scores and calculated the triage system's sensitivity and specificity.

RESULTS:  Among 11,214 African-American (AA) women, the Gail score was 1.7% for 5.9% and the triage function + for 38%. Among 7,603 White women, the Gail score was 1.7% for 35% and the triage + for 47%. Among AA women 40–69 and White women 40–52, no woman lacking a family history of breast cancer or a previous breast biopsy (“triage-”) had a Gail score 1.7%.

CONCLUSION:  A relatively simple triage approach can be used to identify women within age-ethnicity subgroups for whom a full Gail risk assessment is unnecessary. In our population this approach would have allowed avoidance of full Gail risk assessment for 62% of AA and 31% of White women without missing any woman with a Gail score 1.7%. Since these results are based on the structure of the Gail model itself (in the absence of our triage factors the maximum possible Gail score cannot exceed 1.7% for the specified ethnicity-age groups) our approach should be generalizable.

inline image

INCIDENCE AND SEVERITY OF DISPENSING ERRORS IN A HOSPITAL PHARMACY.E.G. Poon1; J. Cina1; E. Featherstone1; W. Churchill1; P. Mitton1; M. McCrea1; C.A. Keohane1; D.W. Bates1; T.K. Gandhi1. 1Brigham and Women's Hospital, Boston, MA. (Tracking ID #115800)

BACKGROUND:  Hospital pharmacies dispense very large numbers of medication doses for hospitalized patients every day. Many of these seriously ill patients are particularly susceptible to adverse drug events (ADE). Because of the large number of medications used, even a very low dispensing error rate may potentially lead to many ADEs. We therefore sought to characterize the incidence and severity of medication dispensing errors in a hospital pharmacy.

METHODS:  Using direct observation techniques, we studied the medication dispensing process in the hospital pharmacy at a 720-bed tertiary academic medical center. This pharmacy employed a commonly-used 2-stage dispensing process in which first a pharmacy technician picked medications from pharmacy supplies according to physician orders, and then a registered pharmacist checked the picked medications for errors. As part of this research study, a trained observer visually inspected medications after completion of the routine 2-stage dispensing process in order to look for non-intercepted errors. These errors were defined as discrepancies between what was approved for dispensing by the pharmacist vs. what was ordered by physicians, or other deviations from established pharmacy policies. Each error was further reviewed by a 2-physician panel to determine its potential to cause an ADE if the error was not intercepted by the nurse prior to medication administration.

RESULTS:  We observed 19,338 medication doses that were dispensed over a 2-month period. Overall, 2.9% (557 of 19,338) of the medication doses picked by the pharmacy technician contained an error. While the pharmacist intercepted 69% (n = 393) of the errors committed by the pharmacy technician, 31% (n = 179) of errors were undetected by the routine 2-stage dispensing process. This corresponds to an overall dispensing error rate of 9.2 errors per 1,000 medication doses dispensed. Of the 179 non-intercepted errors, 26% (n = 47), or 2.4 per 1,000 medication doses dispensed, had the potential to cause an ADE. Of these 47 potential ADEs, 25 (53%) were considered significant, 18 (38%) severe, and 2 (4%) life-threatening.

CONCLUSION:  Dispensing errors in hospital pharmacies are common and more than 30% of these errors are not intercepted by the routine checking processes in the hospital pharmacy; this error rate is typical for low-fidelity practices such as inspection. More than a quarter of these non-intercepted errors have the potential to cause ADEs. These data suggest that to make dispensing safer, high-reliability technologies such as electronic bar-coding or robot dispensing should be considered.

INCREASING REFERRALS TO TELEPHONE COUNSELING FOR SMOKING CESSATION.S.E. Sherman1; N. Takahashi1; P. Kalra2; W. Kuschner3; J. Canfield2; E. Gifford2; J. Kelly2; J. Finney2. 1VA Greater Los Angeles Healthcare System, Sepulveda, CA; 2VA Palo Alto Healthcare System, Palo Alto, CA; 3Stanford University, Stanford, CA. (Tracking ID #116290)

BACKGROUND:  Despite widespread availability of free Quitlines, proven Quitline effectiveness, and scarce institutional resources for counseling, few healthcare organizations refer patients to Quitlines for smoking cessation. We tested the effectiveness of a telephone care coordination program (TCCP) for smoking cessation, whose objectives were to increase referrals to the state Quitline and to provide smoking cessation medication management.

METHODS:  We randomly assigned 10/18 sites in 2 California Veterans Health Administration (VA) health care systems to receive the TCCP intervention. This included simple referral (2 mouse clicks) by the primary care provider or staff, after which the telephone care coordinator contacted the patient and assisted them in contacting the state Quitline. The care coordinator arranged providing smoking cessation medications, and monitored their use during follow-up contact at 2, 4, 6, and 8 weeks. At baseline, we conducted a very brief survey of providers at all 18 sites, which included a question on referrals to telephone counseling. Our outcome measures were the number of patients starting treatment and the number who completed the 2-month program and were abstinent from cigarettes.

RESULTS:  At baseline, providers reported essentially never referring patients for telephone counseling (median 0 referrals/month). During the first seven months, TCCP received 2,221 referrals. We were unable to reach 1,106 patients (50%) despite multiple attempts. 241 patients (11%) opted not to start. Of the remaining 874 patients who began the telephone cessation program, 723 (87%) patients completed their counseling from the Quitline and entered into the care management program and 95% of them started smoking cessation medications. At this time, 28% of patients completed the 2 month TCCP, 21% dropped out and 51% were still in progress.

CONCLUSION:  The TCCP dramatically increased use of the state Quitline. At least 28% successfully completed the 2-month program. This program provides an effective method to increase referrals to telephone counseling and coordinate smoking cessation medications. This program now accounts for 5–10% of all counseling calls received by the state Quitline, making it the most referrals from any health care system. It gives medical centers another option, increasing access to smoking cessation treatment without overwhelming primary care providers.

INDIVIDUALIZED TREATMENT PREDICTIONS FOR BRCA POSITIVE WOMEN WITH BREAST CANCER.H. Burke1; J. Weitzel2; S. Narod3. 1George Washington University, Washington, DC; 2City of Hope, Duarte, CA; 3University of Toronto, Toronto, Ontario. (Tracking ID #116882)

BACKGROUND:  A woman newly diagnosed with breast cancer who is BRCA positive needs to receive information that can help her choose her best treatment. We have developed a clinical decision tool that integrates genetic and clinical information and provides the physician and patient with the patient's individual probability of contralateral breast cancer (a second primary tumor) and of disease-specific survival for each treatment option.

METHODS:  The data was a multi-institutional collaboration of 1,139 women with breast cancer who were BRCA positive. Women were included if they were less than 66 years of age, were diagnosed with Stage I or Stage II breast cancer, and were either BRCA1 or BRCA2 positive. 647 women were alive and 492 were deceased. SAS (Cary, NC) was used to construct proportional hazards models for the risk of contralateral breast cancer and of disease-specific survival.

RESULTS:  The model inputs were: age at diagnosis, BRCA1 or BRCA2 positive, ER and PR status, lymph node involvement, tumor size, type of primary surgery, tamoxifen, and chemotherapy. The model's outputs were the five-year and ten-year probability of contralateral breast cancer and of disease-specific survival for prophylactic contralateral mastectomy and prophylactic oophorectomy. Contralateral mastectomy provided a modest but significant (P < .05) improvement in survival. Oophorectomy reduced a woman's risk of death due to breast cancer by over 70% (P < .001).

CONCLUSION:  We have created a powerful clinical decision support system that integrates genetic and clinical information. It is helping women with breast cancer who are BRCA positive make informed treatment decisions.

INDIVIDUAL-LEVEL AND NEIGHBOURHOOD-LEVEL INCOME MEASURES: AGREEMENT AND ASSOCIATION WITH OUTCOMES IN A CARDIAC DISEASE COHORT.D. Southern1; L. McLaren2; P. Hawe1; M.L. Knudtson1; W.A. Ghali1. 1University of Calgary, Calgary, Alberta; 2Unviersity of Calgary, Calgary, Alberta. (Tracking ID #116700)

BACKGROUND:  Census-based measures of income are often used a proxy for individual-level income. Yet, the validity of such area-based measures relative to ‘true’ individual-level income has not been fully characterized. The objectives of this study were 1) to determine whether area-based measures of household income are a valid proxy for self-reported household income, and 2) to assess whether one or both of these measures are associated with outcomes in a cardiac disease cohort.

METHODS:  We used follow-up data (1999 to 2001) from the Alberta Provincial Project for Outcome Assessment in Coronary Heart Disease (APPROACH) to address this question. The APPROACH study's follow-up survey contains an item for self-report of household income, as well as postal codes for deriving census-based measures of income. Census enumeration area (EA)-derived and self-reported measures of household income were thus both available and the extent of agreement between measures was determined. We also assessed the association between each of these income measures and clinical outcomes after cardiac catheterization—survival (from the time of the APPROACH follow-up survey) and quality of life (based on the EuroQoL)

RESULTS:  : 4,372 patients had complete EA-derived and self-reported household incomes. Agreement between the two income measures was generally poor: For any given self-reported household income grouping provided in the follow-up, EA-derived household income estimates were only in perfect agreement for 1.9% of the lowest (5th) income group, 17.2% of the 4th group, 44.1% of the 3rd group, 36.5% of the 2nd group, and 30.5% of the highest (1st) income group. This level of agreement corresponds to a kappa level of only 0.07, and a weighted kappa (accounting for partial agreement across quintiles) of 0.16. Despite this poor agreement, both income measures were associated with progressively improving survival and EuroQoL scores as income increased. An outcome analysis that simultaneously considered self-reported (individual level) income and area-based income revealed that low-income neighbourhoods are associated with poorer survival and lower quality of life scores regardless of individual-level income

CONCLUSION:  The area-based estimates of household income in these data demonstrate poor agreement with self-reported household income at the level of individual patients, particularly for low-income patients. Despite this, both income measures appear to be prognostically relevant, perhaps because individual and neighbourhood income are not a single construct. These findings intriguingly suggest that individual wealth, although important as a determinant of health, may be overshadowed in importance by neighbourhood-level factors.

INFLUENCE OF PERSONAL CHARACTERISTICS AND PRE-MEDICAL EXPERIENCES ON MEDICAL STUDENTS’ CLINICAL PERFORMANCE.D.M. Windish1; E.G. Price1; S.L. Clever1; J.L. Magaziner1; P.A. Thomas1. 1Johns Hopkins University School of Medicine, Baltimore, MD. (Tracking ID #116095)

BACKGROUND:  Studies suggest that prior academic achievement does not predict future clinical performance or patient satisfaction. This study explores the influence of personal characteristics and pre-medical educational experiences on medical students’ communication and clinical reasoning skills in the second year.

METHODS:  119 second-year students were surveyed regarding their age, gender, college major, previous interviewing experience and prior medical training. We assessed their clinical reasoning skills and their ability to recognize communication behaviors at baseline through their assessment of a taped medical encounter after one week of interviewing training. We subsequently evaluated communication and clinical reasoning skills in 93 of the students during a standardized patient (SP) exercise after 9 weeks of training. Communication skills were rated by the SP using a 5-point Likert scale and a history item checklist. We assessed clinical reasoning skills by the accuracy of differential diagnoses as well as the number and breadth (different components listed from the history and physical) of their problem lists. We compared differences among variables using Mann-Whitney U, chi square and t tests.

RESULTS:  At baseline, there were no statistical differences in students’ self-rated proficiency in creating problem lists or in generating differential diagnoses by any characteristic studied. There were also no differences in their ability to recognize different communication behaviors. During the SP exercise, older students were more likely to use vocabulary at the level of patient understanding [median 4 (interquartile range (IQR) 3–4) vs. 4 (IQR 4–5); P = .03], but did not differ in any other area of performance. Women were more likely than men to list personal stress as a potential medical problem (93% vs. 74%; P = .004) but performed similarly in their communication skills. Science majors scored statistically lower on five interpersonal items including encouraging the patient to ask questions [3 (IQR 2–3) vs. 4 (IQR 3–5); P = .008], but did not differ from non-science majors in their clinical reasoning ability. Students with prior interviewing experience obtained more history items from the SP (mean 11.6 vs. 10.4; P = .01) and were more likely to be recommended to a friend than those with no prior experience [5 (IQR 4–5) vs. 4 (IQR 4–5); P = .03]. Students with previous medical training performed at the same level as those without former training in all areas assessed.

CONCLUSION:  Certain personal characteristics and past educational experiences are associated with second-year students’ clinical performance. Medical schools may wish to closely explore these areas and consider tailoring clinical skills training for students that have different backgrounds.

INFLUENCE OF PHYSICIAN ADVICE ON THE USE OF LOW-YIELD CIGARETTES: RESULTS FROM THE 2000 NATIONAL HEALTH INTERVIEW SURVEY.H.A. Tindle1; R.B. Davis2; I. Kawachi3. 1Harvard Medical School, Boston, MA; 2Beth Israel Deaconess Medical Center, Boston, MA; 3Harvard School of Public Health, Boston, MA. (Tracking ID #116838)

BACKGROUND:  Low-nicotine and low-tar cigarettes, referred to collectively as “low yield” cigarettes (LYC) have not been demonstrated to result in lower harm. Yet many smokers are known to switch to LYC in an attempt to reduce their health risk. In a nationally representative survey, we examined the predictors of LYC use among current smokers, including receipt of physician advice to quit smoking.

METHODS:  We studied 32,374 respondents to the 2000 National Health Interview Survey (NHIS). The Adult Sample Core and Supplement on Cancer Control elicited information on sociodemographic factors and chronic health conditions as well as questions on perceived personal and family cancer risk. Current smokers were asked, “Did you ever use or switch to a lower tar and nicotine cigarette to reduce your health risk?” We performed multivariable logistic regression using backward elimination to identify determinants of LYC use. All analyses used SAS-callable SUDAAN and were weighted to reflect national estimates.

RESULTS:  Of the 5,378 smokers who responded to the questions of interest, 2,363 (43.9%) indicated that they had switched to LYC in order to reduce their health risks. Respondents who used LYC were more likely than non-LYC users to be white (87.4% vs. 72.7%) and female (57.6% vs. 49.4%). LYC users were more likely to report a history of vascular disease (including coronary artery disease and stroke) as compared to non-LYC users (8.6% vs. 5.7%). LYC users were also more likely than non-LYC users to have received physician advice to quit smoking (60.2% vs. 45.7%). In the adjusted analysis, whites were more likely to use LYC than Hispanics (AOR: 2.91 [2.08–4.06]) and non-Hispanic blacks (2.60 [2.10–3.23]). Females were more likely than males to use LYC (1.31 [1.12–1.52]), and respondents with a history of vascular disease were more likely than their healthy counterparts to report use of LYC (1.55 [1.15–2.10]). Finally, respondents who had received physician advice to quit smoking were over 50% more likely to use LYC as respondents who were not advised to quit (1.55 [1.34 – 1.79]).

CONCLUSION:  Smokers who received physician advice to quit smoking were over 50% more likely to use LYC as those who did not receive such advice. The weighted sample of LYC users represents over 12 million Americans who smoke. In the absence of convincing evidence of harm reduction resulting from LYC use, there needs to be wider discussion within the medical community, as well as between physicians and their patients, concerning the risks of switching to LYC.

INTEGRATING PERSONALIZED DECISION SUPPORT INTO PRIMARY CARE: IMPACT ON DECISIONAL CONFLICT.N.F. Col1; J.M. Fortin1; L. Ngo2; A.M. O’Connor3; R. Goldberg4. 1Brigham and Women's Hospital, Boston, MA; 2Harvard School of Public Health, Boston, MA; 3University of Ottawa, Ottawa, Ontario; 4University of Massachusetts Medical School (Worcester), Worcester, MA. (Tracking ID #117423)

BACKGROUND:  Good communication and strong patient-provider relationships are associated with better clinical outcomes, yet increased productivity pressures on physicians limit time spent with patients and opportunities for informed decision making. We developed a computer-generated, personalized decision support aid (DSA) to help women make more informed decisions about menopause and to become more active participants in their health care, testing alternative approaches to delivering the intervention in primary care.

METHODS:  We developed decision support software that provides women with personalized information about their menopausal symptoms and chronic disease risks (heart disease, breast cancer, and osteoporosis), the impact of hormone therapy and other treatment options on their symptoms and risks, and strategies for risk reduction. A 3-armed randomized control trial was conducted among postmenopausal women in 4 clinical practices comparing a control group (receiving a brochure of American College of Physicians guidelines; n = 50), to the DSA alone (n = 45), or the DSA with coaching conducted just prior to the clinic visit (n = 50). Healthy menopausal women between the ages of 45 and 65 were recruited through clinic rosters. Participants completed a questionnaire about their lifestyle and medical history; responses were entered into a software program that generates a 35+ page personalized health report (the DSA) for the patient and a summary page for her clinician. Decisional conflict (DC) was measured at baseline and 2 weeks after the clinic visit. Lower DC scores are desirable.

RESULTS:  Of the 145 women included in the trial, 99 had valid responses. The mean age was 52.3, 98% were white, and 58% had completed some college. Total DC declined more in the intervention vs control group (0.6 vs 0.09, P < .001). Improvements were observed in all subscales. No differences between the 2 intervention arms (DSA alone vs DSA + coaching) were observed on either total DC or any subscales.

CONCLUSION:  A DSA that includes personalized risk estimates and treatment options dramatically lowered decisional conflict. The effect size was 2–3 fold larger than that observed among DSAs included in a systematic review of DSAs, suggesting that inclusion of personalized information on risks and symptoms improves effectiveness. No difference between DC scores were observed between the 2 intervention arms, suggesting that computer-generated personalized information can supplant coaching and leverage clinician time.

INTENSITY AND QUANTITY OF PHYSICAL ACTIVITY AND THE RISK OF DIABETES IN WOMEN.A.R. Weinstein1; H.D. Sesso2; I.M. Lee2; N.R. Cook2; J.E. Manson2; J.E. Buring2; J.M. Gaziano1. 1Boston VA Healthcare System, Boston, MA; 2Brigham and Women's Hospital, Boston, MA. (Tracking ID #116381)

BACKGROUND:  Previous studies have demonstrated that physical activity reduces the incidence of type 2 diabetes mellitus (DM); however, data examining the association between the intensity and quantity of physical activity and the risk of DM in women are sparse.

METHODS:  We analyzed 38,180 subjects who provided data on time spent on 8 groups of leisure-time physical activities in the Women's Health Study, an ongoing prospective study of initially healthy women aged ≥45 years who were free of DM. Total energy expenditure (kcal/wk) was calculated. Energy expended by intensity level was calculated and categorized as no activity, light-to-moderate intensity (<6 METs), and vigorous intensity (≥6 METs). Walking pace (mph) and walking time (h/wk) were reported. Multivariate Cox proportional hazards regression models evaluated the relationship between each measure of physical activity and the risk of DM, adjusted for lifestyle, clinical, and dietary risk factors.

RESULTS:  After a median follow-up of 7.2 years, there were 1,376 incident cases of type 2 DM. 3,816 subjects reported no activity, 18,291 reported light-moderate activity only, and 16,068 reported any vigorous activity. The rates of incident DM were 6.3% for inactive subjects, 3.9% for those who performed light-moderate activity only, and 2.7% for those who performed any vigorous activity. In the multivariate-adjusted analysis of overall activity (in kcal/wk), the RRs (95% CIs) were 0.91 (0.79, 1.06), 0.86 (0.74, 1.01), and 0.82 (0.70, 0.97) for the 2nd, 3rd and 4th quartiles compared to the 1st (P trend = 0.01). In the multivariate-adjusted analysis of intensity level, compared to inactive subjects, the RR (95% CI) was 0.91 (0.77, 1.06) for the light-moderate group and 0.83 (0.69, 1.01) for the vigorous group. The difference between the groups was not significant (P trend = 0.06). We further evaluated walking, since it is the most common moderate activity in women. Compared to non-walkers, the multivariate-adjusted RR (95% CI) was 1.00 (0.85, 1.18) for a walking pace of <2 mph, 0.89 (0.77, 1.02) for 2–2.9 mph, and 0.75 (0.62, 0.89) for ≥3 mph (P trend < .01). In the analysis of walking time, compared to non-walkers, the multivariate-adjusted RR (95% CI) was 0.85 (0.62, 1.16) for <1 h/wk, 0.78 (0.57, 1.08) for 1–1.5 h/wk, 0.59 (0.42–0.84) for 2–3 h/wk, and 0.79 (0.57, 1.11) for ≥4 h/wk (P trend = .01).

CONCLUSION:  In this study, greater energy expenditure is associated with lower risk of type 2 DM in women. Even moderate activity, such as walking, is associated with lower risk. For type 2 DM prevention in women, it appears that the amount of physical activity is at least as important as the intensity level of selected exercise each week.

INTEREST IN BRCA TESTING AMONG RELATIVES OF WOMEN WITH BREAST CANCER.M. Gilligan1; M. Schapira1; S. Bourassa1; A. Nattinger1. 1Medical College of Wisconsin, Milwaukee, WI. (Tracking ID #117359)

BACKGROUND:  First-degree relatives of women with breast cancer represent a group of women with a moderately increased personal risk of developing breast cancer, many of whom, nevertheless, would not be appropriate candidates for BRCA testing according to established guidelines. We hypothesized that interest in genetic testing would be associated with perceived risk, knowledge, and coping style, and that an educational intervention regarding breast cancer risk and BRCA testing would decrease interest in testing in this population.

METHODS:  The study questionnaires and educational intervention were developed based on a focus group analysis completed in a previous phase of the study. First-degree relatives were recruited through their relatives with breast cancer listed in the hospital tumor registry, and through general advertisement. A total of 57 women completed the baseline study and the pre- and post-intervention questionnaires. Interest in testing was assessed using a single 5-point Likert item (1-definitely not interested to 5-definitily interested). Knowledge score was the percentage correct out of 25 true-false questions. Perceived risk was assessed using a 0–100% visual analog scale. Numeracy was assessed using a 3-item scale. Monitor-Blunter (M-B) scores were determined using the methods of Miller, et al. Descriptive statistics and univariate testing were performed.

RESULTS:  The study cohort was predominantly white (98%), well-educated (65% college degree or greater), numerate (61%) and had a moderately high lifetime risk of breast cancer as calculated by the Gail model (mean 19%, range 9–44%). The mean M-B score was 5.64, with a range of –11 (High Blunter) to +15 (High Monitor). At baseline, higher interest in BRCA testing was associated with lower knowledge scores (P = .007) and higher self-perceived risk (P = .041) but was not associated with high scores on the M-B scale. The intervention was effective in increasing knowledge with a mean score pre- and post- of 70.86% (SD 9.69) and 83.34% (SD 8.98), respectively (P < .0001). There was no change in perceived lifetime risk (median 50, range 10–90 on both pre- and post-tests). Interest in genetic testing decreased after the intervention, with 65.6% and 54.5% stating that they “probably” or “definitely” wanted testing on pre- and post-tests, respectively (P = .018).

CONCLUSION:  In a well-educated cohort of moderate risk women, interest in genetic testing was high but decreased after an intervention which focused on educating women about breast cancer risk and genetic testing. This study suggests that high perceived risk of cancer and low knowledge may be factors associated with higher interest in genetic testing. No support was found for the hypothesis that coping style is associated with interest in testing.

INTERNISTS AND GYNECOLOGISTS ATTITUDES AND KNOWLEDGE OF CORONARY HEART DISEASE (CHD) PREVENTION IN WOMEN. J.M. Barnhart1; P. Charney2; O. Cohen3; J.L. Houghton4; V. Lewis5; I.R. Merkatz1. 1Albert Einstein College of Medicine, Bronx, NY; 2Norwalk Hospital, Norwalk, CT; 3Ferkauf Graduate School of Psychology, Yeshiva University, Bronx, NY; 4Albany Medical College, Albany, NY; 5University of Rochester School of Medicine & Dentistry, Rochester, NY. (Tracking ID #115849)

BACKGROUND:  The New York State Department of Health funded the “Women and Heart Disease: Physician Education Initiative” developed and implemented by the American College of Obstetricians and Gynecologists District II/NY (ACOG) and American College of Physicians (NYACP). We sought to determine the physicians’ knowledge and attitudes toward CHD prevention in women.

METHODS:  A multi-disciplinary physician panel developed and presented an educational program at 24 sites (Grand Rounds and regional meetings) from October 2003 through April 2004. All attendees were asked to complete a 5-minute questionnaire during the program, which focused on barriers to CHD prevention, knowledge of evidenced-based guidelines for optimal cholesterol and blood pressures levels, and current smoking cessation services available in their practice. Six items assessed barriers to CHD prevention and were rated on a 5-point Likert scale (1 = strongly agree to 5 = strongly disagree). This analysis only includes the responses of attending physicians who were internists or gynecologists (n = 529).

RESULTS:  Surveys were completed by 704 attendees; 75.1% were either internists (n = 378) or gynecologists (n = 151). The mean age was 41 years (S.D. = 12.3), 73.5% (n = 389) practiced in the New York City area, and 42.7% (n = 226) were females. Gynecologists were more likely than internists to report lack of time and training as barriers to CHD prevention in their practice (66% vs. 52%; P = .004 and 53.8 % vs. 29.3%; P < .001, respectively). Moderate knowledge of CHD prevention in women was found (mean score = 9.3 [SD = 1.8]; range 4–13), with higher mean knowledge scores for internists than gynecologists (9.6 vs. 8.6; P < .001). However, 90% of internists and gynecologists did not know that a woman's relative risk of CHD increases 2-fold for a body mass index (BMI) over 22. Gynecologists were less likely than internists to know the ATP III value for optimal LDL cholesterol (65.5% vs. 35.1%; P < .001) or provide nicotine replacement therapy for their patients who wanted to stop smoking (35% vs. 53.7%; P < .001).

CONCLUSION:  Internists and gynecologists perceive time as a barrier to implementing heart disease prevention into practice. Both groups need more educational programs to further enhance their knowledge and practice of CHD prevention. Such programs should reflect specialty differences.

INTERPRETER USE VERSUS LANGUAGE CONCORDANCE IN ASIAN AMERICAN OUTPATIENTS.A.R. Green1; Q. Ngo-Metzger2; A. Legedza1; R.S. Phillips1; L.I. Iezzoni1. 1Beth Israel Deaconess Medical Center, Boston, MA; 2University of California, Irvine, Irvine, CA. (Tracking ID #115567)

BACKGROUND:  Patients with language barriers are less likely to understand their care, are at higher risk for medical errors, and are less satisfied than others. Interpreters can ameliorate these problems, but few studies have compared patients’ experiences with clinicians who speak their own language to those that involve an interpreter.

METHODS:  We conducted a mail survey in four languages of Chinese and Vietnamese patients seen within the previous month at one of 11 community health centers across the U.S. Among demographic and other questions, we asked patients to rate their visit overall and to assess 5 aspects of clinician-patient communication. We studied patients whose clinicians spoke their native language (concordant group) or who used an interpreter, excluding patients who spoke English “very well” and for whom an interpreter was unavailable. We used multivariable logistic regression to determine relationships between interpreter use and the communication and visit rating outcomes, adjusting for age, education, primary language, English ability, time in U.S., clinic site, usual clinician, reason for visit, and health status. All analyses were performed using SUDAAN.

RESULTS:  3,258 patients (74%) returned completed questionnaires. After exclusions, 1,362 remained in the interpreter group and 1,353 in the concordant group. Interpreter group patients: had spent less time in the U.S. (mean 10.5 vs. 12.2 years, P = .0002); were less educated (36.6% vs.49.1% with 10+ years, P = .0001); had lower English proficiency (43.0% vs. 36.1% spoke no English, P = .0003); and were more likely to speak Vietnamese (46.9% vs. 36.8%, P = .0001). In unadjusted chi-squared analyses, patients who used an interpreter were less likely than language concordant patients to report they always understood their care (61.1% vs. 67.8%, P = .02). They were more likely to report having questions about their care (20.9 vs. 30.1, P = .0005) or questions about mental health (18.2% vs. 25.3%, P = .0045) that they did not ask. In adjusted models, these differences persisted. Patients with interpreters were less likely to report they always understood care (adjusted OR 0.73 [95%CI 0.54–0.97]), and more likely to have unasked questions about care [1.68 (1.20–2.34)] or about mental health (1.49 [1.04–2.12]) than language concordant patients. Other communication outcomes and visit rating did not differ significantly between groups: having time to explain reasons for the visit; getting enough information about health/treatment; and rating the visit excellent or very good.

CONCLUSION:  Although visit ratings were not affected, patients who used interpreters were less likely to ask their clinicians questions or to feel that they explained things in an understandable way than language concordant patients. Interpreters can help overcome language differences, but certain communications barriers may remain.

INTERVENTIONS TO CHANGE PHYSICIAN BEHAVIOR: ARE MORE BETTER?D. Haggstrom1; A.B. Bindman1. 1University of California, San Francisco, San Francisco, CA. (Tracking ID #111420)

BACKGROUND:  Trials of interventions to change physician behavior—like reminders—show multifaceted approaches are more efficacious than single ones. We studied whether the number of interventions (1) is associated with behavior change among physicians in practice and (2) varies in effectiveness among different health care organizations.

METHODS:  We surveyed a probability sample of 1,238 California primary care physicians in 2001 with a response rate of 67%. Among physicians who reported receiving practice profile information from any of 4 types of health care organizations (health plan, independent practice association (IPA), medical group, or hospital), we asked whether they had (1) received reminders, guidelines, education, feedback, or financial incentives from the same health care organization and (2) implemented a change in practice as a consequence. Univariate tables and multivariable logistic regression models were used to study the association between the number of interventions (0–5) received from any single type of health care organization and a change in physician practice. Covariates were physician demographics (age, gender, race/ethnicity), specialty, work hours, practice setting (solo, group practice, or staff/group model HMO), and the total number of interventions received from the other 3 types of health care organizations.

RESULTS:  The likelihood of practice change increased as the number of interventions increased. After adjusting for covariates, each additional practice intervention was associated with practice change when received from an IPA (OR = 10.81; 95% CI = 3.49–33.48), medical group (OR = 2.23; 1.23–4.04), or hospital (OR = 5.32; 1.92–14.75) but not from a health plan (OR = 1.47; 0.80–2.68). The total number of interventions received from other health care organizations was not associated with a change in practice.

CONCLUSION:  More interventions are more likely to change physician behavior, and a threshold effect occurs when >3 interventions are received. Interventions from health plans are less likely to change practice than those from other sources. There is a potential for coordinated efforts among health care organizations to change physician behavior, but multiple interventions from a single health care organization appear to be more effective.

inline image

INVESTIGATOR ASSESSMENT OF THE EFFECTIVENESS OF MANDATED INCLUSION OF WOMEN AND MINORITIES IN RESEARCH.G.M. Corbie-Smith1; D. St. George2; C. Blumenthal1; R.W. Durant3; I. Williams1. 1University of North Carolina at Chapel Hill, Chapel Hill, NC; 2Walden University, Minneapolis, MN; 3Harvard University, Boston, MA. (Tracking ID #117322)

BACKGROUND:  In the mid 1990's investigators (PIs) raised concerns about the impact of federally mandated inclusion of minorities and women on the conduct of research. However there has been little subsequent assessment of the effectiveness of the mandate on achieving gender and racial/ethnic diversity in research. We sought to describe PIs’ assessment of the impact of the mandate on inclusion of women and minorities in research; we also evaluated whether PI sociodemographic and professional characteristics, research attributes and attitudes about inclusion are associated with PI assessment of the effectiveness of the mandate.

METHODS:  Between Aug and Dec 2002, we surveyed National Heart Lung and Blood Institute PIs with active human subjects research in 2001. The mailed survey asked about recruitment of minorities and women, opinions on the NIH mandate to include women and minorities, PIs current research and PI demographics. Our dependent variables were respondents’ agreement with 2 statements: “I think the NIH mandate has failed to increase the number of women (or minority) participants recruited into research.”

RESULTS:  440 PIs completed the survey (70%response rate): mean age = 51 years, 75%male, 90% White, and 86% had been funded as a PI for >5 years prior to the survey. Most PIs (69%) felt that the mandate had been successful in increasing gender diversity; only 7% of respondents felt that the mandate had failed to increase female inclusion in research and 24% were unsure. Further, 55% of PIs felt the mandate had been successful in increasing minorities in research; 12% felt it had failed and 32% felt unsure. Race and attitudes about inclusion, but not other PI sociodemographic/professional characteristics or research attributes, were significantly associated with PIs assessment of both mandates. White PIs (P = .05), those who felt more strongly that failure to include women makes it hard to draw conclusions about treatment/prevention strategies (P = .05), and those who ascribed more importance to female inclusion in their own research (P = .03) were more positive about the mandate's success for women. Similarly, White PIs (P = .01), those who felt more strongly that diversity in study samples ensures generalizability (P = .03), and those who ascribed more importance to minority inclusion in their own research (P < .01) were more positive about the mandate's success for minorities.

CONCLUSION:  While the majority of PIs were positive in their assessment of the mandate on increasing diversity in research, many were ambivalent. Non-minority PIs and those who endorsed the importance of inclusion were more likely to believe the mandate had been effective. More research is needed to determine whether perceptions of the mandate reflect increasing diversity in research samples.

INVESTIGATORS’ VIEWS ON THE CONCEPT OF RACE: IN TALK AND TEXT.G.M. Corbie-Smith1; S.E. Estroff1; G.E. Henderson1; C. Blumenthal1; J. Dorrance1. 1University of North Carolina at Chapel Hill, Chapel Hill, NC. (Tracking ID #116632)

BACKGROUND:  While debate on the meaning of race as a variable continues, investigators (PIs) who collect data on race and use race in their analyses must make decisions on how it should be represented in their work. Given the implications of this controversy for research on health disparities by race, we sought to understand how PIs interpret the concept of “race” and use it in their research.

METHODS:  From Nov 2001 to Jan 2003, we conducted semi-structured interviews with PIs from 3 universities. Topics included their experiences with recruitment of minority subjects, use of race as a variable in research/analyses and assessment of the NIH mandate. Transcripts were analyzed using the principles of grounded theory. PIs were enrolled until theoretical saturation (i.e., no new concepts arose during analysis of successive interviews). For each PI, we also abstracted up to 2 publications to assess their use of race in the reports of their research.

RESULTS:  We contacted 43 PIs and conducted 33 interviews (77%response rate). Several PIs endorsed the idea of race as a biologic construct including responses such as the identification of genetic variants, race differences in response to treatments, and race as a risk marker for other diseases. Most PIs who described race as a social construct pointed to differences in risk of disease based on environment, behavior, diet, socioeconomic status, education and/or wealth. Some PIs also described theoretical underpinnings of the operation of race in health (e.g., linking race to oppression, racism or invoking feminist theory). A group of PIs used a mix of social and biologic concepts. Some held primarily a biologic view of race with modification by social factors and others acknowledged possible genetic differences but felt that social factors were the main determinants of health. A third group suggested both social and biologic factors were equally responsible for differences in health and disease. Of the 46 publications examined, 85% (n = 39) mentioned race at any point in the manuscript. Of these 12 manuscripts provided a definition or explanation of why race was included as a variable in the study: 20% as social and 11% as biologic constructs and found 67% had agreement between the how PIs described the concept of race in interviews and how it was described in their manuscripts.

CONCLUSION:  PI concepts of race varied along a spectrum from biological to social constructs. A small number of papers explicitly addressed the meaning of race as variable, however, when explanations were provided, these were similar to conceptualizations of race obtained in PI interviews. PIs views on the concept of race are consistent with the debate on race and should be explicitly documented in their published work.

IS ACETYLCYSTEINE EFFECTIVE IN PREVENTING CONTRAST-RELATED NEPHROPATHY? A META-ANALYSIS.K.G. Shojania1; B.K. Nallamathou2; S.K. Saint2; T.P. Hofer2; H.D. Humes2; M. Moscucci2; E.R. Bates2. 1University of California, San Francisco, San Francisco, CA; 2University of Michigan, Ann Arbor, MI. (Tracking ID #116951)

BACKGROUND:  Acetylcysteine for prevention of contrast nephropathy is quickly becoming standard practice, despite mixed results in clinical trials. A recent meta-analysis reported net benefit for acetylcysteine, but did not include data from 7 additional trials available in abstract form.

METHODS:  We conducted an extensive electronic and manual search for published and unpublished randomized clinical trials evaluating acetylcysteine for prevention of contrast-related nephropathy. At least two authors abstracted data from each trial including assessments of its quality and additional characteristics. The primary outcome of interest was the incidence of nephropathy 48 hours after contrast administration. Data were combined using random-effects models with the performance of standard tests to assess for heterogeneity and publication bias.

RESULTS:  Fourteen trials involving 1455 patients met our inclusion criteria. Trials varied in patient demographics, inclusion criteria, dosing regimens, and few trials reported important elements of study design such as concealment of allocation, or provided any details related to use of placebos or blinding. The summary relative risk ratio for contrast-related nephropathy overall was 0.67 but was of borderline statistical significance (95% CI, 0.44 to 1.0; P = .064). Moreover, the benefit varied significantly across the 14 trials (test of heterogeneity, P = .027), with this heterogeneity remaining unexplained despite numerous subgroup analyses.

CONCLUSION:  Acetylcysteine may reduce the incidence of contrast-related nephropathy, but this finding is inconsistently reported and inclusion of unpublished studies substantially reduced the significance of the summary effect. Bias due to unconcealed treatment allocation in the majority of trials may explain the heterogeneity between available trials and their inconsistent results. Larger, higher quality clinical trials are needed before universally recommending use of acetylcysteine. Although inexpensive and without direct harm, incorporating it into standard practice may indirectly harm patients by delaying necessary imaging tests and angiographic procedures.

IS HIGHER QUALITY PRIMARY CARE ASSOCIATED WITH REDUCED ADDICTION SEVERITY?T.W. Kim1; J.H. Samet1; D.M. Cheng1; M. Winter1; D. Safran2; R. Saitz1. 1Boston University, Boston, MA; 2Tufts University, Boston, MA. (Tracking ID #116393)

BACKGROUND:  Alcohol and drug use disorders are chronic diseases that require ongoing management of physical, psychiatric, and social consequences. While specific addiction-focused interventions in primary care are efficacious, the influence of overall primary care quality on addiction outcomes has not been studied. We examined the relationship between primary care (PC) quality and addiction severity in a prospective cohort study.

METHODS:  Subjects reporting alcohol, cocaine or heroin as drugs of choice were recruited from an inpatient detoxification unit to participate in a randomized intervention trial to link them to PC. For this analysis subjects had initiated PC and completed at least 2 research interviews over 2 years. PC quality was measured using scales from the well-validated Primary Care Assessment Survey (PCAS): financial access; organizational access; visit-based continuity; thoroughness of physical exam; communication; interpersonal treatment; physician knowledge of patient; trust; and preventive counseling. Three addiction outcomes were assessed: 1. Addiction Severity Index; alcohol (ASI-alc); 2. ASI-drug; and 3. past month alcohol use to intoxication or any heroin or cocaine use. Each outcome was measured during the interview subsequent to the one at which the PCAS was assessed. Longitudinal regression models adjusted for age, gender, race, randomization group, insurance, baseline ASI-alc and ASI-drug.

RESULTS:  The 183 subjects were: mean age 37 years; female 32%; black 54%; health insurance 40%. All PCAS scales were significantly related to reduced ASI-alc. In addition, a higher rating on the physician knowledge of patient scale was associated with both lower ASI-alc score (mean decrease 0.06 per standard deviation (SD) in PCAS, P < .001), ASI-drug score (mean decrease 0.02 per SD, P = .02), and lower odds of alcohol intoxication or drug use (OR = 0.66 per SD, 95% CI 0.51–0.85). Higher trust scores were associated with lower ASI-alc (mean decrease 0.05 per SD, P = .01) and lower odds of alcohol intoxication or drug use (OR = 0.74 per SD, 95% CI 0.59–0.95). Other comparisons were not significant.

CONCLUSION:  Higher quality of primary care, across many domains, appears to be associated with reduced alcohol addiction severity. Furthermore, a higher quality primary care relationship, as indicated by physician knowledge of patients and trust, is associated with improvements in alcohol and drug outcomes. These data suggest that efforts directed towards transforming the currently fragmented, episodic medical care often received by patients with addictions into high quality primary care could reduce their alcohol and drug use and addiction severity.

IS SENIORITY BETTER? DIFFERENCES IN PERFORMANCE OF PREVENTIVE HEALTH SERVICES BY YEAR OF RESIDENCY TRAINING.L.L. Willett1; G.R. Heudebert1; S. Massie1; C. Benton1; T. Houston1. 1University of Alabama at Birmingham, Birmingham, AL. (Tracking ID #117137)

BACKGROUND:  An objective practice-based learning project was implemented to assess the performance of preventive health services in the resident continuity clinic. We present results of performance by level of residency training.

METHODS:  Six quality indicators were chosen based on national guidelines and were abstracted from patient charts by a trained abstractor blinded to resident demographics using a standard computer-based tool. A patient had to be seen by the same resident more than once during the year and qualify for the preventive service by national guidelines. We conducted univariate and multivariate analyses using chi square and logistic regression respectively to compare performance between interns and residents.

RESULTS:  Patient charts (N = 761) were abstracted for 71 residents and interns. Compared to patients cared for by residents, those cared for by interns tended to receive more preventive health services including: fecal occult blood testing (FOBT) (42% vs 31%, P = .01) and tobacco use screening (60% vs 52%, P = .02). A non-significant trend was also seen for pneumococcal vaccination (63% vs. 57%), mammography (43% vs. 38%), and counseling smokers to quit (71% vs. 62%). This trend was not seen for lipid screening or invasive colon screening. After adjustment for patient age and race, patient and resident gender, and number of visits during the measurement year, patients who were seen by interns were more likely to receive FOBT [OR 1.6 (CI 1.04–2.23)] and be screened for smoking [OR 1.4 (1.01–1.98)] as compared to those seen by residents. Seven residents met or exceeded the “benchmark” level of performance (top 10% as compared with other residents) in 4 out of the six indicators. Six of these seven top performers were interns.

CONCLUSION:  Patients cared for by interns as compared to residents over a one year time period were more likely to receive several preventive health services recommended by national guidelines. Potential explanations include longer clinic contact time, closer attending supervision, or more attention to detail by interns as compared to residents. Regardless of year of training, performance of these quality indicators needs improvement.

IS THE PLACEBO HARMLESS? A META-ANALYSIS OF CARDIOVASCULAR TRIALS.R. Mallory1; E.B. Bass1; S.N. Goodman1. 1Johns Hopkins University, Baltimore, MD. (Tracking ID #116386)

BACKGROUND:  There is continuing debate over the ethics and utility of using placebos as controls in randomized trials. We sought to determine if patients who receive placebos in randomized controlled trials (RCTs) of cardiovascular interventions are at an increased risk of death or serious adverse events compared to patients who receive interventions.

METHODS:  We used Cochrane Collaboration search strategies to identify all English-language, placebo-controlled RCTs indexed by Medline in 2002 in the field of Cardiovascular Medicine. Cross-over studies, sub-studies and follow-up studies were excluded. We abstracted data on adverse events, serious adverse events, trial withdrawals and all-cause mortality. Risk differences were calculated and then pooled using random-effects meta-analytic methods. Trials were scored on a 1–6 scale ranging from placebo highly preferred to intervention highly preferred, based on the original investigators’ interpretation of the overall trial results.

RESULTS:  We identified 654 RCTs of which 236 met the inclusion criteria. A total of 127,626 patients were enrolled. Intervention arms were preferred or highly preferred in 66% of the studies, placebo arms preferred or highly preferred in 15%, and the two arms considered equal in 19%. 138 trials provided information on adverse events, 101 on serious adverse events and 115 on mortality. The median proportion of patients with adverse events was 21% (IQR: 6, 51). The pooled estimate of the absolute risk difference between placebo and intervention arms was –2.6% (95% CI: −4.1, −1.1) with an increased risk of adverse events for patients in intervention arms. For serious adverse events the pooled estimate of the risk difference was 0.7% (95% CI: 0.1, 1.4) with an increased risk for patients in placebo arms. The median overall trial mortality was 2.7% (IQR: 0.8, 6.9), and the pooled estimate of the absolute risk difference in mortality was 0.28% (95% CI: 0.12, 0.44) indicating 10% greater mortality in the placebo group. In a meta-regression model including trial size, trial length, number of trial sites, trial funding source and journal impact factor only trial location remained associated with a risk difference in mortality The absolute risk difference in mortality for US trials was −0.03 (95% CI: −0.36, 0.29) while the risk difference for non-US trials was 0.49 (95% CI: 0.28, 0.70).

CONCLUSION:  This analysis suggests that patients enrolled in the placebo arms of randomized controlled trials in Cardiovascular Medicine are at a modestly increased risk of death and serious adverse events. Whether this reflects an over-reliance on the use of placebo controls, or is actually a welcome sign of medical progress remains to be determined. Further work focusing on possible sources of difference between US and Non-US trials is also required.

IS THERE A RELATIONSHIP BETWEEN SHARED DECISION MAKING AND SUBJECTIVE COLLABORATION? D. Schillinger1; G.W. Saba1; C.P. Somkin2; A. Fernandez1; C. Wilson1; K. Grumbach1; S.T. Wong3. 1University of California, San Francisco, San Francisco, CA; 2Kaiser Permanente Division of Research, Oakland, CA; 3University of British Columbia, Vancouver, British Columbia. (Tracking ID #117233)

BACKGROUND:  Shared decision-making (SDM) is promoted as an ideal method of physician-patient communication and indicator of collaborative care. Whether SDM is experienced by patients and providers as genuinely collaborative is questionable.

METHODS:  To examine the relationship between the extent of SDM and subjective collaboration, we combined methods of direct observation of patient-physician encounters with patient and physicians’ stimulated recall of these same encounters. We videotaped and transcribed a diverse convenience sample of chronic care visits in a public hospital's primary care clinic. We coded decision moments, critical events in the encounter containing a choice of action, according to modified criteria of Elwyn and Braddock. Decision-making was shared if the dialogue included at least 5 of 10 possible communication criteria. Patient and physician participants separately viewed their encounter with a clinical psychologist, who asked them to discuss whether the decision-making was congruent with their preferences and was experienced as collaborative. Transcripts of the stimulated recall sessions were analyzed for themes at each decision moment and characterized as subjectively collaborative if each partner reported the decision-making congruent with their personal or professional values and/or reflective of genuine partnership.

RESULTS:  We analyzed 5 visits and 10 stimulated recall sessions involving 5 patients and 5 primary care physicians. We identified 29 decision moments, 16 of which (55%) had objective evidence of SDM. Among the SDM decisions, 9/16 were experienced as collaborative; among the non-SDM decisions, 10/13 were experienced as collaborative.Inconsistencies were observed in the association between SDM and subjective collaboration even within patient-physician dyads. Thematic analysis revealed that relationship dynamics involving trust, truth telling, expectations of roles, and unchecked assumptions mediated the extent to which SDM was associated with subjective collaboration.

CONCLUSION:  Because decision-making in primary care appears to occur in a context of a set of assumptions and experiences within the patient-physician relationship, the presence or absence of shared decision-making behavior is not necessarily related to perceived collaborative care. However, how decisions are made does appear to be related to important issues in the patient-physician relationship such as trust and role expectation.

IS TRAINING IN SHARED DECISION MAKING IN UNDERGRADUATE MEDICAL EDUCATION ADEQUATE?S.L. Clever1; E.G. Price1; D.M. Windish1; J.L. Magaziner1; P.A. Thomas1. 1Johns Hopkins University, Baltimore, MD. (Tracking ID #117294)

BACKGROUND:  Shared decision making (SDM) is an important component of patient-centered care, which is key element of quality of care. Unfortunately, SDM is rarely incorporated into routine office visits. Medical school is a critical time for shaping future physicians’ attitudes and practices with patients, and inadequate preparation in SDM skills in medical school may be contributing to this problem. As preparation for a curriculum focusing in part on SDM, we set out to determine the status of SDM education at our institution.

METHODS:  We surveyed a convenience sample of 3rd- and 4th-year medical students while they attended 5 clerkship orientation meetings in January 2003. We asked them to indicate whether they had formal training during their preclinical or clinical years in taking a medical history and performing a complete physical exam, as well as in key SDM skills of: encouraging questions from patients; eliciting patients’ concerns, beliefs and expectations of their illness; eliciting patients’ preferences for decision making; and reaching agreement on treatment plans. We also asked them to rate their level of competence in these skills on a 5-point scale, 0 = No exposure, 1 = Familiar, 2 = Can perform somewhat, 3 = Can perform well, 4 = Can teach to other students. In addition, we surveyed 6 clinical clerkships directors (Ambulatory Medicine, Emergency Medicine, Internal Medicine, Neurology, Pediatrics and Surgery) and asked them to rate the “average” student's performance in these same skills on a 3-point scale, 1 = Less prepared than s/he should be, 2 = At the level s/he should be, 3 = More prepared than expected.

RESULTS:  We received surveys from ≥96% of students who attended the orientation meetings, for a total of 96 (40%) of the 240 3rd- and 4th-year medical students. The table below presents the percent of students who indicated they had not received formal training in the targeted skills. It also presents the percent of students who rated themselves as being able to perform the skills “somewhat” or worse, and the percent of course directors who rated the average student as less prepared than s/he should be.

CONCLUSION:  In our institution, while students are consistently receiving training in history taking and physical exam, a substantial proportion of 3rd- and 4th-year students indicated that they do not have any formal training in key SDM skills. Even if they receive this training, many students and a majority of faculty rate students’ skills as inadequate. If these results hold true at other institutions, it indicates a need for improved training in SDM skills during undergraduate medical education.

Skill% students with no formal training% students rating “Can perform somewhat” or worse% faculty rating average student “Less prepared than s/he should be”
Performing a medical history 014 43
Performing a complete PE 030 17
Encouraging questions1152100
Eliciting patients’ concerns1239 83
Eliciting patients’ preferences3053 67
Reaching agreement on treatment3145 50


BACKGROUND:  Transition of care from pediatric to adult providers for adults born with chronic disease (ABCD) is often an unplanned experience that is occurring more frequently as medical advances increases the lifespan of these patients. Sickle cell (SS) patients are one group who must undergo this transition and in this study we explored their attitudes and perceptions about their transition of care and sought to identify ways to improve this process.

METHODS:  We interviewed SS patients at an academic medical center that serves an urban minority population. Patients were selected sequentially based on their presentation for care and a diagnosis of SS disease. Open-ended questions from a structured survey tool were asked by a same race, similar age, non-medical interviewer. All interviews were anonymous and audiotaped for later transcription. Interviews were continued until recurrent themes emerged. Three researchers independently reviewed and coded the interview transcripts using Atlas.ti coding software to identify overarching themes.

RESULTS:  Eleven patients were interviewed, 6 men and 5 women. The mean age was 29 (range 22–46) and 8 patients had at least a high school education. Patients’ experiences with the transition of care were generally negative. Some were informed upon seeking health care that they had already been transferred to the adult service without their knowledge. “Just like that, I was moved and I had no clue as to where I was going or who was my new doctor.” Patients perceived that their pediatric providers spent more time with them, communicated better and had better attitudes compared with their adult providers. Nursing interaction was as critical as physician interaction in patients’ perceptions of treatment. Family support and self-care were integral components of the overall care. Patients reported significant lack of control over their care while hospitalized but felt in control in the ambulatory environment. Patients desired more professional support, communication, education and encouragement. All participants believed that there was a need for a structured transition program.

CONCLUSION:  These sickle cell patients perceived a significant, negative change in their care during their transition from pediatric to adult health care providers. Some potential results of this negative experience are decreased continuity of care, decreased patient satisfaction and overall increased health care utilization. A formal transition of care program that addresses both medical and psychosocial support is indicated for this patient population and other ABCD patients. This program should involve a team approach including physicians, staff, families and patients.

KNOWLEDGE OF MOST RECENT HEMOGLOBIN A1C VALUES AMONG ADULTS WITH DIABETES: PREVALENCE AND CORRELATES.M. Heisler1; J. Piette1; E.C. Kieffer2; M.S. Spencer2; S. Vijan1. 1VA Center for Practice Management and Outcomes Research, Ann Arbor, MI; 2University of Michigan, Ann Arbor, MI. (Tracking ID #116254)

BACKGROUND:  In chronic diseases such as diabetes, patient knowledge of their actual and target disease outcomes (such as Hemoglobin A1c values) is hypothesized to be an important prerequisite for effective patient involvement in managing their conditions. Accordingly, among a sample of adults with diabetes, we assessed: 1) the frequency and clinical, socio-demographic and health care correlates of knowing one's most recent A1c value; and 2) whether knowing one's most recent A1c was associated with better self-reported diabetes care understanding and self-efficacy.

METHODS:  Cross-sectional survey of a sample of 840 U.S. adults with type 2 diabetes receiving care in a suburban Veterans Affairs facility, academic medical center, or inner-city urban health care facility (response rate of 68%). The primary outcome measure was accurately reporting (within a ±0.5 range) one's last HbA1c value (comparing respondents’ survey report with the most recent A1c value prior to the survey from medical records). Independent variables were patients’ income, education, race/ethnicity, age, gender, diabetes duration, diabetes medications regimen, health care site, and several variables for respondents’ evaluation of their health care providers.

RESULTS:  Twenty-five percent of the respondents accurately reported their last A1c level. In the multivariable logistic regression analyses, when only respondents’ socio-demographic and clinical characteristics were included in the model, Latino respondents had significantly lower odds than white respondents of knowing their most recent A1c value (AOR: 0.12, 95% CI: 0.02–0.74), as did respondents with less than a high school education (AOR: 0.26, 95% CI:0.13–0.51). When the health care variables were added to the model, only education level continued to be an independent predictor of knowing A1c. Respondents who strongly agreed that their health care provider thoroughly answered their questions had higher odds of accurately knowing their last A1c value (AOR: 1.64, 95% CI: 1.05–2.56). After adjusting for all the above variables, respondents who knew their last A1c value had significantly higher reported understanding of their diabetes care (P < .001). A1c knowledge, however, was not associated with respondents’ reported self-efficacy in managing their diabetes.

CONCLUSION:  Few respondents in this sample of adult patients with diabetes accurately knew their last A1c value. While Latino patients had more than 80% lower odds of knowing their A1c value than white patients, these ethnic differences were no longer statistically significant once the health care variables were added to the models. A1c knowledge was indeed associated with better reported diabetes care understanding, but not with increased confidence about managing diabetes.

LABOR INPUTS AND COSTS OF A SUCCESSFUL PRIMARY CARE-BASED, DIABETES DISEASE MANAGEMENT PROGRAM.R. Rothman1; S. So1; R.M. Malone2; B. Bryant2; D.A. DeWalt2; M. Pignone2; R.S. Dittus1. 1Vanderbilt University, Nashville, TN; 2University of North Carolina at Chapel Hill, Chapel Hill, NC. (Tracking ID #116447)

BACKGROUND:  Government and private sector organizations have recently invested in disease management programs to improve quality of care and reduce costs for patients with chronic disease. However, little has been published on the labor inputs and costs necessary to implement a disease management program. We examined the labor inputs and costs of a successful primary care-based, diabetes disease management program.

METHODS:  We performed a randomized controlled trial of a diabetes disease management program for 217 patients in an academic primary care practice. The intervention group received intensive management from clinical pharmacists and a diabetes care coordinator (DCC) who: provided diabetes education; applied algorithms for treating glucose and decreasing cardiovascular risk, and addressed barriers to care. Control patients received a one-time management session from pharmacists followed by usual care from their primary provider. Clinical outcomes included glycated hemoglobin (A1C), blood pressure, total cholesterol, and aspirin use at 12 months. Process outcomes included number of actions/contacts with patient, time spent with patient, number of drug additions/titrations. Labor costs were calculated based on Bureau of Labor Statistics national wage data with a sensitivity analysis.

RESULTS:  For the 194 (89%) patients with 12-month data, the intervention group had significantly greater improvement than the control group for: systolic blood pressure (−6.9 mmHg vs. +2.3 mmHg, P = .004); A1C (−2.7% points vs. –1.5% points, P = .01); and aspirin use (+47% vs. +6%, P < .0001). Change in total cholesterol was not significant (−19 mg/dl vs. –14 mg/dl, P = .35). There were no significant differences in use of clinical services or adverse events. The disease management team made a median of 44 contacts or care-related activities for a median 450 minutes for intervention patients and 12 contacts for 125 minutes for control patients (P < .0001). For intervention patients, 68% of contact time was conducted by clinical pharmacists, 19% from the DCC, and 13% from both. This time was allocated as follows: 46% in-person patient contact, 42% phone management, 12% chart-review/data management. Of note, 20% of time was spent on incomplete actions (ex. incomplete phone calls). The most common pharmacist actions included: medication titration/addition (34%), patient assessment (21%), and prescription renewal (19%). Intervention patients had a median of 7 drug titrations or additions during the study. In sensitivity analysis, the incremental labor costs for the intervention were $16.24 per patient per month (sensitivity analysis: $13.27−$22.58).

CONCLUSION:  A diabetes disease management program that improved patient outcomes can be successfully integrated into a primary care clinic at a reasonable cost.

LACK OF ASSOCIATION BETWEEN LITERACY AND DEPRESSIVE SYMPTOMS OR MENTAL HEALTH FUNCTION OF ADULTS WITH ADDICTION.M. Paasche-Orlow1; A.K. Lincoln1; D.M. Cheng1; M. Winter1; R. Saitz1; J.H. Samet1. 1Boston University, Boston, MA. (Tracking ID #117307)

BACKGROUND:  While multiple surveys have reported an association between low literacy and worse health status in various chronic diseases, longitudinal data have been sparse. We hypothesized that low literacy would be associated with more depressive symptoms and worse mental health functioning in patients with addictions.

METHODS:  In a prospective cohort study, we evaluated literacy using the Rapid Estimate of Adult Literacy in Medicine (REALM) (≤6th grade, 7th–8th grade, ≥9th grade), depressive symptoms using the Center for Epidemiologic Studies-Depression (CES-D) scale, mental health functioning using the Mental Component Summary (MCS) scale of the Short Form-36 health survey, and drug and alcohol addiction severity using the Addiction Severity Index (ASI-Drug and ASI-Alc). Eligible subjects were inpatients at a detoxification unit, had no primary medical care, had a Mini-Mental State Exam (MMSE) of >21, completed the REALM, and were in a randomized controlled trial of a clinical intervention to link them with primary care from 1997–1999. Follow-up was at 6-month intervals for 2 years. Using unadjusted and adjusted linear regression models at study entry and then unadjusted and adjusted longitudinal generalized linear models we investigated the relationship between 1) literacy and CES-D score and 2) literacy and MCS score. Adjustment variables included age, race, gender, income, education, primary language, MMSE, and randomization group.

RESULTS:  Literacy levels among the participants were: 14%≤6th grade, 33% 7th–8th grade, 53%≥9th grade. At cohort entry, there were significant unadjusted correlations between low literacy and less depressive symptoms (P = .01) and better mental heath functioning (P = .05), but these correlations were not significant when adjusted for potential confounders. In longitudinal analyses no association was found between literacy and CES-D or MCS in unadjusted or adjusted models.

CONCLUSION:  Literacy was not associated with depressive symptoms or mental health function in either cross-sectional or longitudinal analyses in patients with addictions. The impact of literacy on health outcomes is not uniform; insight into the mechanisms by which literacy effects are elaborated are important next steps.

LANGUAGE BARRIERS: EFFECTS ON HEALTH CARE COMPREHENSION. E. Wilson1; A.H. Chen2; K. Grumbach1; F. Wang3; A. Fernandez1. 1University of California, San Francisco, San Francisco, CA; 2Asian & Pacific Islander American Health Forum, San Francisco, CA; 3University of California, San Francisco, Orinda, CA. (Tracking ID #117257)

BACKGROUND:  Language barriers can negatively impact the quality of care and may place patients at risk for adverse outcomes. Language concordant physicians may reduce these risks. We studied a linguistically diverse sample of Californians to determine a) the effect of limited English proficiency on comprehension in the healthcare setting and b) whether access to language concordant physicians can reduce language barriers.

METHODS:  Telephone survey of 1200 randomly selected Californian immigrants conducted in 11 languages, excluding English. The survey asked about spoken English fluency on a 4-point scale, usual sources of care, demographics, and included four questions on comprehension in a healthcare setting: problems understanding a medical situation, confusion about how to use medication, trouble understanding labels on medication, and bad reactions to medications due to a problem understanding instructions. Respondents were defined as limited English proficient (LEP) if they indicated they spoke English “not well” or “not at all.” We used multivariate logistic regression to determine the relationship between LEP status and medical comprehension, adjusting for age, sex, education, income, health insurance, language/ethnicity, number of years in U.S., and usual source of care. In order to determine the effect of physician language concordance on comprehension, we performed a stratified analysis for respondents with language concordant and discordant physicians.

RESULTS:  592 (49%) of the 1200 respondents were defined as LEP. 49% of LEP and 21% of non-LEP respondents had problems understanding a medical situation (AOR 3.2/CI 2.1,4.8); 35% of LEP and 24% of non-LEP respondents were confused about how to use medication (AOR 1.4/CI 0.9,2.1); 42% of LEP and 25% of non-LEP respondents had trouble understanding labels on medications (AOR 1.5/CI 1.0,2.3); and 16% of LEP and 8% of non-LEP respondents reported a bad reaction to medications (AOR 2.3/CI 1.3,4.4). In the stratified analysis, among respondents with language concordant physicians, LEP respondents were more likely to have problems understanding a medical situation than non-LEP respondents (AOR 2.2/CI 1.2,3.9); there were no differences between the two groups in the other three measures. Among respondents with language discordant physicians, LEP respondents were more likely to have problems understanding a medical situation (AOR 9.4/CI 3.7–23.8), more likely to have trouble understanding medication labels (AOR 4.2/CI 1.7–10.3), and more likely to report a bad reaction to medications (AOR 4.1/CI 1.2–14.7) than non-LEP respondents.

CONCLUSION:  Limited English proficiency among Californians is a barrier to understanding medical situations and instructions, and increases the risk for adverse medication reactions. Access to language concordant physicians greatly mitigates but does not eliminate these barriers.

LANGUAGE CONCORDANCE AND GLYCEMIC CONTROL IN AN URBAN HISPANIC COHORT.A. Montero1; J. Jones1; R. Gross1; O. Carrasquillo1. 1Columbia University, New York, NY. (Tracking ID #117367)

BACKGROUND:  Hispanics diabetics have poorer glycemic control than non-Hispanic Whites. We carried out this study to evaluate whether language discordance between patient and provider may present a unique barrier to achieving glycemic control among Spanish speaking Latino patients.

METHODS:  Subjects were drawn from a larger study of 1,157 patients randomly selected from an urban general medicine group practice. Our cross sectional analysis linked responses from that study's questionnaire with data from our practice's clinical information system. Our sample consisted of the 234 patients in the parent study who were Spanish-speaking (preferred the Spanish questionnaire), diabetic (based on ICD-9 code or laboratory data diagnostic for diabetes), and had at least one hemoglobin A1C(HgA1C) measurement in 2001 or 2002. There were 89 distinct primary care providers; our unit of analysis was the patient. Our dependent variable was language concordance categorized as a) language concordant (provider known to be fluent in Spanish using implicit criteria; N = 99) b) providers who routinely used interpreters based on interpreter logs; (N = 86) or c) non-concordant (providers who were neither Spanish fluent nor routinely used interpreters; N = 49). Our outcome variable was poor glycemic control (last HgA1C > 9.5). We used logistic regression to adjust for those covariates which in bi-variate analysis were correlated with poor glycemic control.

RESULTS:  We found poor glycemic control in 9% of patients in the language concordant group as compared to 26% of patients in the interpreter group (P = .015) and 20% of patients in the non-concordant group (P = 14). Other correlates (P < .05) of poor glycemic control were age, provider type (resident vs. attending), Medicaid insurance, and <3 clinic visits per year with primary provider. In multivariate models, the interpreter and non-concordant group were not more likely to have poor glycemic control than the concordant group (OR = 1.96 {95% CI .73–5.29} and OR 1.49 {95% CI .45–4.89}, respectively). In this model the only age remained independently associated with poor glycemic control with younger patients having poorer control.

CONCLUSION:  In our sample of urban Latino patients we did not find an independent relationship between language concordance and poor glycemic control. Caveats of this negative study include limited statistical power, high rates of language concordance or interpreter use, and inability to more precisely classify Spanish proficiency among providers who were not Spanish fluent.


BACKGROUND:  Several multivariate indices exist to aid the clinician in determining cardiac risk and include evaluation of the ECG for arrhythmia or prior myocardial infarction. Although left ventricular hypertrophy (LVH) on pre-operative echocardiography has been associated with cardiac events, this has not been demonstrated using ECG criteria. The purpose of this study was to determine whether the presence of LVH on the pre-operative ECG is independently associated with perioperative cardiac complications.

METHODS:  This was a retrospective-prospective case-control study. Cases and controls were obtained from a previous prospective study designed to compare cardiac risk indices and were matched on type of surgery and Goldman cardiac risk group. Cases were identified if they had any of: unstable angina, acute pulmonary edema, non-fatal myocardial infarction or cardiac death. Up to three matched controls were selected randomly per case. Preoperative ECGs were obtained by searching hospital charts and ECG archives. Patients were excluded if the ECG was not obtainable or if a left bundle branch block (LBBB) was found, as these were otherwise uninterpretable. LVH was diagnosed blindly, using the method of Romhilt & Estes.

RESULTS:  1465 patients were enrolled in the original study, 74 of whom had a major perioperative cardiac complication. 207 matched controls were obtained. Preoperative ECGs were not available or uninterpretable in 15 (20%) of the cases and 33 (16%) of the controls. Of the 59 cases with interpretable ECGs, 18 (31%) had at least “probable” LVH, whereas 26 (15%) of the 174 control ECGs had this pattern (OR 2.5, 95% CI: 1.25–5.00, P < .008). When the criteria for “definite” LVH were applied, 17% of the cases and 10% of the controls received this diagnosis (OR 1.88, 95% CI: 0.81–4.38, P = .137).

CONCLUSION:  These results suggest that LVH on the preoperative ECG is an independent predictor of perioperative cardiac complications. Cases and controls were well-matched, underscoring the independent nature of this predictor. The frequency of missing ECGs between the cases and controls was similar, and ECGs were interpreted blind to case-control status. Thus there are few potential sources of bias in this study. Notwithstanding, a prospective study demonstrating that LVH is a predictor of perioperative cardiac complications would represent stronger evidence and would be a useful future project. In the interim, it is reasonable to consider LVH an additional risk factor to be evaluated preoperatively.


BACKGROUND:  The majority of Asian Americans are immigrants whose risk of heart disease, diabetes, and obesity increases as duration of residence in the United States (U.S.) increases. Regular physical activity reduces the risk of these diseases, yet little is known about physical activity in Asian Americans and how it changes after immigration. Prior economic literature on physical activity suggests there may be different effects on leisure time and non-leisure time physical activity following immigration.

METHODS:  Data from the 2001 California Health Interview Survey, which oversampled Asian Americans, were analyzed to investigate the effects of ethnicity, foreign birth, and years in the U.S. on leisure time physical activity and non-leisure time physical activity. The sample included 962 Chinese, 766 South Asians, 667 Filipinos, 658 Vietnamese, 623 Koreans, 550 Japanese, and 27,853 U.S.-born non-Asians (ages 18–59).

RESULTS:  Asian Americans were much less likely to participate in leisure time physical activity than U.S.-born non-Asians (odd ratio (OR), men = 0.47, 95% confidence interval (CI): 0.39, 0.56, OR, women = 0.47, 95% CI: 0.41, 0.56). Foreign-born Asians were least likely to participate in leisure time physical activity; leisure time physical activity increased as years in the U.S. increased. Asian Americans were not more likely to participate in non-leisure time physical activity than U.S-born non-Asians. Ethnicity was significantly associated with being sedentary among Asian subgroups. Chinese Americans were most sedentary.

CONCLUSION:  Asian Americans, especially immigrants, are at risk for low levels of leisure time physical activity. Low levels of leisure time physical activity were not offset by increased non-leisure time physical activity. Research is needed to understand the socioeconomic and cultural factors that promote or limit physical activity in Asian Americans. Increasing regular physical activity is key to protecting the health of this rapidly growing population.


BACKGROUND:  Advance directives are infrequently useful in emergent situations, unavailable, or misunderstood by physicians. Nevertheless, considerable emphasis has been placed on the completion of advance directives by the public. Serious unintended consequences may be the result.

METHODS:  Three audiences were recruited: one consisting entirely of laypersons recruited by advertisement on National Public Radio (NPR) (n = 32), one of health care professionals (university internal medicine faculty, residents, and students [n = 78]), and one mixed audience, a hospital ethics committee (70% laypersons [n = 14]). Information was gathered about knowledge of advance directives, personal experiences with them, and answers to four hypothetical cases. An Audience Response System (ARS) was used, which allowed 100% participation, anonymous answers, electronic tallies of the results, and immediate feedback so that interest in the poll was sustained. All results were preserved by the ARS and subsequently analyzed.

RESULTS:  Healthcare professionals considered themselves very knowledgable (30%) or somewhat knowledgable (53%) about the proper use of advance directives. Nevertheless, the information was used at the point of care in <25% of proposed situations; instead, “DNR” status or “talking to the nearest relative” were preferred. Only 28% of physicians were aware their patients had completed advance directives. Further, physicians frequently misapplied the supplied information. In one scenario, a lethal error was made by >55% of the physicians which would have resulted in the unintended death of a patient; in another scenario, 30%. In contrast, lay audiences used the information contained in advance directives correctly the majority of the time (76%). When presented with this information, lay audiences indicated mistrust of physicians, and requests for physician-assisted suicide.

CONCLUSION:  In emergent situations, advance directives are seldom present or useful at the point of care. Some physicians do not understand their limited applicability, and errors resulting in unintended patient death may occur. The general public does not appreciate these consequences, and feels betrayed and mistrustful of physicians when informed.

LIMITED HEALTH LITERACY IS ASSOCIATED WITH INADEQUATE UNDERSTANDING OF ANTICOAGULATION RISKS, BENEFITS, AND MANAGEMENT.M.C. Fang1; E. Machtinger1; D. Schillinger1. 1University of California, San Francisco, San Francisco, CA. (Tracking ID #116665)

BACKGROUND:  Because of the risk of bleeding complications and the need for ongoing monitoring, oral anticoagulation requires a significant degree of patient involvement and education about warfarin's risks and benefits. Little is known about the extent to which patients comprehend basic concepts inherent in the management of anticoagulation as well as what factors are associated with poor comprehension.

METHODS:  Bilingual research assistants surveyed outpatients followed in a hospital-affiliated anticoagulation clinic and assessed patient demographics, cognitive status (using the s-CASI), and identified patients with limited health literacy (score of <23 on the Test of Functional Health Literacy in Adults [TOFHLA] in English or Spanish). Using expert opinion and adapting questions from prior studies on anticoagulation knowledge, we developed a 6 item Fundamentals of AntiCoagulation Test (FACT), composed of 5 multiple-choice questions about the mechanism, monitoring, and complications of warfarin, as well 1 question about risk-reduction, all phrased in lay terms. Inadequate understanding was defined as scoring in the lowest FACT quartile. We then used multivariable logistic regression to find predictors of inadequate understanding.

RESULTS:  We interviewed 179 anticoagulated patients whose mean age was 60 yrs and who had been on chronic warfarin for an average of 4 yrs. The population was racially and ethnically diverse, with 34% being Hispanic, 25% Asian-American/Pacific Islanders, 20% White, 17% African-American, and 4% of other race. Most patients (85%) were anticoagulated for atrial fibrillation or prosthetic valves. The median FACT score was 4; 14% of patients correctly answered all 6 questions, while 19% scored in the lowest quartile (≤2). 46% of patients with limited health literacy scored in the lowest quartile, compared to only 6% of those with adequate health literacy. After adjusting for age, sex, language, education and cognitive status, patients with limited health literacy were substantially more likely than those with adequate health literacy to have inadequate anticoagulation knowledge (odds ratio 7.2 [95% CI 2.1–24]).

CONCLUSION:  Limited health literacy was strongly associated with inadequate understanding of concepts essential to collaborative anticoagulation management. Further studies should examine ways to more consistently communicate these concepts so that all patients can participate in informed decision-making.

LITERACY AND ACCURACY FOR SELF-ADJUSTED DIURETIC DOSING IN PATIENTS WITH HEART FAILURE.N. Rogers1; D.A. DeWalt1; M. Pignone1; R.M. Malone1; B. Bryant1; K. Felix1; C. Rawls1; K. Corr1; M.C. Kosnar1; R. Rothman2; B.F. Angel1. 1University of North Carolina at Chapel Hill, Chapel Hill, NC; 2Center for Health Services Research, Nashville, TN. (Tracking ID #116117)

BACKGROUND:  Patients with low literacy have less disease specific knowledge and may be prone to more self-care errors. Daily weight measurement and diuretic dose adjustment are associated with improved outcomes for patients with heart failure. We sought to determine if low literacy is associated with decreased accuracy in weight based diuretic dose adjustment in a structured heart failure disease management program.

METHODS:  We performed a chart review of patients enrolled in the interventional arm of a HF disease management study. Patients received HF educational materials designed for patients with low literacy and a one-hour education session on self-care. We provided a digital scale and asked patients to measure and record weights daily in a specially designed logbook and to adjust their diuretic dose based on weight, using an individualized care plan. Patients also received phone calls to reinforce program adherence and answer questions. We measured literacy using the Rapid Estimate of Adult Literacy in Medicine (REALM) and dichotomized the measure at the 6th grade level. We reviewed 2 weeks of patient logs from two separate periods (3–7 weeks and 18–22 weeks) after enrollment. For each two-week period 28 errors were possible, including no daily weight recorded and improper diuretic dose for weight.

RESULTS:  Our final sample of 56 patients was taken from the 64 intervention patients with the following exclusions: deceased prior to 6 months (2), withdrawn/lost to follow-up (5), 6 month assessment not completed (1). Of the 56 patients reviewed, 11 did not use the logs. Low literacy patients (25 total) were more likely to use the logs than higher literacy patients (31 total) (92% vs. 71%, P = .05). During weeks 3–7, the mean number of errors was higher in the low literacy group (6.7 vs. 3.6, P = .14). During weeks 18–22, the mean number of errors was similar between the groups (3.6 vs. 4.2, P = .76) and lower than during the early time period. There was no relationship between literacy and number of encounters with the disease management team.

CONCLUSION:  Low literacy patients are more likely to use the self-monitoring materials, but they may be more likely to make early errors in diuretic dosage adjustment Errors become less common over time. Patients with low literacy skills can learn to perform complex self-care for HF, but may benefit from more intensive early support.

LITERACY AND PAIN, FUNCTIONAL STATUS, AND DEPRESSION SCORES IN PRIMARY CARE PATIENTS WITH CHRONIC PAIN.J.S. Perhac1; T.J. Ives1; S. Prakken1; K. Felix1; R.M. Malone1; B. Bryant1; J. Ripton1; D.A. DeWalt1; M. Pignone1; P.R. Chelminski1. 1University of North Carolina, Chapel Hill, NC. (Tracking ID #115921)

BACKGROUND:  Low literacy is associated with several adverse health outcomes. We investigated whether literacy is associated with patient report of pain, depression, functional status, and disability and whether literacy affects the degree of change in these measures over time for patients enrolled in a primary care based general medicine disease management program for chronic non-malignant pain.

METHODS:  Eligible patients had chronic pain (>3 months duration), were taking or being considered for opioid medication, and were referred by their primary care providers. We measured literacy using the Rapid Estimate of Adult Literacy in Medicine (REALM). We assessed pain with the Brief Pain Inventory (BPI), depression with the Center for Epidemiological Studies scale (CES-D), and functional status with the Pain Disability Index (PDI). We dichotomized literacy at 9th grade reading level and evaluated baseline measures and change over 3 months according to literacy.

RESULTS:  We enrolled 114 patients with chronic pain. Mean age was 52.7 years, 56% were female, 79% Caucasian, 37% did not graduate HS, 85% reported an annual household income less than $20,000, 63% reported receiving disability benefits, and 52% were unemployed at enrollment. We observed 57% (65) of patients read below 9th grade level. According to CES-D, major depression was more prevalent in patients with literacy at or above 9th grade level (69%) verses below 9th grade level (46%, P = .017). Receipt of disability benefits was more common among patients reading below the 9th grade level (64% vs. 44%, P = .039). We found no difference at enrollment in BPI or PDI according to literacy. Literacy was not related to changes in pain, depression, or functional status at 3 month follow-up.

CONCLUSION:  Higher literacy was associated with higher levels of depression among patients in a primary care based general medicine disease management program. Low literacy patients were more likely to receive disability benefits. Literacy is not related to patient reported pain, functional status, or to change in these parameters over three months.

LONG TERM OUTCOMES OF INFECTIVE ENDOCARDITIS IN AN URBAN HIV COHORT.M. Burkey1; L. Wilson1; R.D. Moore1; K. Gebo1. 1Johns Hopkins University, Baltimore, MD. (Tracking ID #115971)

BACKGROUND:  Infective Endocarditis (IE) has been shown to occur frequently in HIV infected injection drug users (IDUs) . Few data exist on the long term outcomes of HIV infected patients with IE.

METHODS:  We evaluated 64 first cases of IE between 1996–2002 in a large Baltimore clinical cohort in which comprehensive demographic, clinical, therapeutic and outcomes data are longitudinally collected. Primary outcomes were one year mortality or recurrence of IE. Multivariate logistic regression were used in the analysis, with 95% CI reported.

RESULTS:  Of 3,995 patients followed after January 1, 1996, there were 64 first IE episodes between January 1, 1996 and December 31, 2002 for an incidence of 5.2 IE events per 1000 PY. Of the 64 IE subjects, the majority (91% African-American, 64% male) were IDUs (86%), with a mean age of 41 years (range 28–57). The median CD4 at first IE was 68 cells/mm3 (range 1–1224), with a median HIV-1 RNA 78,288 copies/ml (range 0–226,195). At the time of IE , 31% were on HAART, 88% on PCP prophylaxis, 67% on MAC prophylaxis, and 78% were HCV+. The most common etiologic organisms of IE were Staphylococcus aureus (69%), coagulase negative Staphylococci (5%), Streptococcus viridans (5%), and Enterococcus faecalis (5%). Within one year of IE diagnosis, 17% had an IE recurrence and 48% of all patients died; 58% of all patients either died or had IE recurrence. Adjusting for gender, IDU, and HAART usage, age >40 years (AOR 4.06, 95% CI 1.17, 14.2) was associated with one year mortality, but not with one year recurrence (AOR 1.51, 5% CI 0.31, 7.51). All recurrences occurred in African-American patients. Race was not associated with one year mortality. History of IDU, HCV infection, PCP prophylaxis, MAC prophylaxis, and HAART usage were not significant predictors of mortality or recurrence.

CONCLUSION:  In our cohort, IE is a common diagnosis, and the rates of recurrence and one year mortality are high. Our data suggest the need for close follow-up focusing on preventing recurrence of IE—especially in those over 40 years. Future studies investigating the utility of IE prophylaxis in HIV patients with a history of IE may be warranted.

LONGITUDINAL CHANGES IN A1C TO ASSESS DIABETES QUALITY OF CARE. W. Thompson1; H. Wang1; M. Xie1; L. Pogach2; M.M. Safford3. 1Rutgers, The State University of New Jersey, Piscataway, NJ; 2VA New Jersey Health Care System, East Orange, NJ; 3University of Alabama at Birmingham, Birmingham, AL. (Tracking ID #115738)

BACKGROUND:  We explored longitudinal analysis as a potential tool for diabetes quality measurement.

METHODS:  Using administrative data from 125 VA facilities, we identified 284,895 veteran users with diabetes and at least one National Glycosylated Hemoglobin Standardization Project certified A1c lab test in fiscal year (FY) 1999 or FY 2000. Individual A1c was the dependent variable in a multi-level linear regression analysis incorporating facility-level fixed effects, individual initial A1c, time between A1c tests, and baseline patient characteristics. To evaluate variation, we examined coefficients for facility-level fixed effects, which reflected the average change in A1c for patients cared for at that facility.

RESULTS:  Mean age of the population was 64.1 ± 11.3 years, 97.5% were male, 31.0% were on insulin at baseline, and 76.0% had at least two A1c tests. Mean A1c for FY 1999–2000 was 7.6 ± 1.8%. Overall, for patients with at least 2 tests, A1c declined over FY 1999–2000, (mean −0.31 A1c%) but varied substantially by facility (SD ± 0.45, range −1.90 to +1.03 A1c%). Sixty-eight facilities demonstrated significant (P < .05) declines and 5 demonstrated significant increases. Among the bottom 10 ranked facilities, only four were within the bottom decile with 90% confidence.

CONCLUSION:  Individual glycemic control in patients in most VA facilities improved over two years, suggesting improvements in the effectiveness and outcomes of care. Average facility-level change in A1c levels varied substantially, suggesting process of care variability and room for further improvement.

LONG-TERM FOLLOW UP OF A LONGITUDINAL FACULTY DEVELOPMENT PROGRAM IN TEACHING SKILLS.A.M. Knight1; K.A. Cole1; D.E. Kern1; L.R. Barker1; S.M. Wright1. 1Johns Hopkins University, Baltimore, MD. (Tracking ID #116869)

BACKGROUND:  In this country, millions of dollars and thousands of hours of faculty time continue to be invested each year on faculty development programs. Little is known about the impact of these programs. This study describes follow-up data for participants in a longitudinal faculty development program in teaching skills and attempts to understand the impact that it had on the participants.

METHODS:  A ten-page, 160-item, survey was developed for this study. All those who participated in the Johns Hopkins Faculty Development Program in Teaching Skills from 1987 through 2000 were surveyed, as well as members of a comparison group selected by participants from 1988 through 1996. Surveys were mailed to past participants and their matches in July 2002. Comparisons between the two groups were made using Chi-square tests.

RESULTS:  Surveys were received from 200 participants (83% response rate) and from 104 matches (82% response rate). Participants were more likely than matches to have taught or mentored learners in the past year (89% vs. 80%; P < .05) and more likely to currently have a medical school faculty appointment (77% vs. 66%; P < .05). Participants had more commonly written down or reviewed a list of their professional work goals in the last year (70% vs. 49%; P < .0001). A larger number of participants than matches rated their proficiency ’very good’ or ’excellent’ in “giving feedback to those you teach” (88% vs. 75%; P < .01), “eliciting feedback on your own performance” (70% vs. 56%; P < .05), and “working in groups” (96% vs. 89%; P < .05). Greater numbers of participants than matches said that they ’frequently’ or ’always’“ask learners what they would like to get out of interactions” (70% vs. 50%; P < .01), “spend time building supportive relationships with learners” (85% vs. 71%; P = .01), “assess and focus on the learner's needs rather than their own agenda when precepting” (77% vs. 57%; P < .01), “start feedback sessions by asking learners to assess their performance” (71% vs. 42%; P < .001), and “let learners know their limitations as a teacher” (60% vs. 43%; P < .01).

CONCLUSION:  Past participants in the Johns Hopkins Faculty Development Program have a higher self-assessment of their teaching proficiencies and self-report performing desirable teaching methods more frequently than do matched faculty who never experienced the longitudinal curriculum.

LOOKING FROM THE OTHER SIDE: A MEASURE OF PHYSICIANS’ PERCEPTIONS OF PATIENTS’ COMMUNICATION.H.S. Gordon1; K.J. O’Malley1; B.F. Sharf2; R.L. Street2; P. Haidet1. 1Houston Center for Quality of Care and Utilization Studies, Veterans Affairs Medical Center, Houston, TX; 2Texas A&M University, College Station, TX. (Tracking ID #115518)

BACKGROUND:  Patients’ communication skills are associated with improved health outcomes. Nevertheless, most studies of physician-patient communication have focused on physicians’ communication. Methods to assess patients’ communication, particularly from the physician's perspective, have received comparatively little attention.

METHODS:  We report the development and psychometric testing of a new instrument to measure physicians’ perceptions of patients’ communication behaviors. Based on prior research indicating that patients who provide, seek, and verify information have better outcomes, we created an initial set of 7 items to assess a doctor's perception of the extent to which a patient performs these behaviors. Items assess physicians’ perceptions of patient's discussion of symptoms, discussion of concerns, question asking, and overall communication. Items were tested in a developmental sample to gather evidence for unidimensionality, reliability, and validity. In the development sample, 17 physicians rated the communication of 98 patients undergoing consultation for pulmonary nodules or lung cancer. In the validation sample, 31 physicians rated the communication of 265 patients in primary care settings.

RESULTS:  In the developmental sample, patients were mostly male (94%), white (79%), had mean age of 65.7 years (range 41–85), were seen for a first visit (76%) and were consulting a thoracic surgeon (58%) or an oncologist (42%). Physicians reported mean visit lengths of 20–30 minutes. In the validation sample, patients had mean age of 56.2 years (range 19–85), were mostly male (61%), were approximately equally representative of African American, Hispanic, and white ethnicities, and physicians were mostly general internists (58%). In the developmental sample, the exploratory factor analysis resulted in a single factor of five items that accounted for 51% of the variance, reliability was fairly high (Cronbach's Alpha of 0.74), and convergent/discriminant validity evidence was strong. In the validation sample, the 5-item factor accounted for 62% of the total variance, the reliability was high (0.84), and validity evidence was strong. Total scores on the measure ranged from 12 to 50 in the developmental sample and 13 to 50 in the validation sample, with higher scores indicating “better” communication. Mean (SD) scores on the measure were 37.6 (8.4) and 37.4 (8.7) in the developmental and validation samples, respectively.

CONCLUSION:  Our measure provides a method for evaluating physicians’ perceptions of patients’ communication behaviors. Future studies should evaluate whether communication skills training improves performance on the measure and whether improved performance is associated with behavioral and health outcomes.

LOW HEALTH LITERACY IS COMMON AND UNHEALTHY: DO WE RECOGNIZE IT IN OUR OWN PATIENTS?C.R. Horowitz1; S. Monteith1; M. McLaughlin1; J.E. Sisk1; S. Chatterjee1. 1Mount Sinai School of Medicine, New York, NY. (Tracking ID #116390)

BACKGROUND:  Functional health literacy is a patient's ability to read and interpret common health information such as prescription labels and patient instructional handouts. Low health literacy has been associated with poorer health outcomes and higher health care costs. We surveyed patients to assess their literacy and factors associated with low literacy, and surveyed clinicians to determine if they could accurately identify which of their patients have low health literacy.

METHODS:  At the onset of a randomized controlled trial aimed at improving the health of patients with congestive heart failure (CHF), each patient completed the the Short Test of Functional Health Literacy in Adults (STOFHLA) in English or Spanish. At the end of the trial, we identified the clinicians of enrolled patients seen at one of the participating urban academic medical centers. We then developed a brief survey asking clinicians to rate their patients’ functional health literacy and provide some demographic data. We distributed the surveys to the clinicians at the medical center who most frequently saw each enrolled patient and asked the clinicians to rate their patient's health literacy in English or Spanish using the STOFHLA categories of inadequate, marginal and adequate health literacy.

RESULTS:  The 264 enrolled patients were 20% non-Hispanic White, 42% non-Hispanic Black, 34% Latino, 4% Asian, and 48% female, with a mean age of 58 (20–96). Overall, 48% had low literacy, or LL (which encompasses inadequate and marginal literacy). Non-whites were more likely to have LL (56% vs.17%; P < .0001), as were persons aged 60 and over (66% vs. 34%; P < .0001). LL individuals had more severe heart failure (NY Heart Association Class IV = 49% vs. 38%, P = .01) and more self-reported hospitalizations in the past 3 months (23% vs. 13%; P = .03). The 82 clinicians who returned surveys for 193 (73%) patients were 20% non-white; 48% were female, 30% cardiologists, 31% primary care providers, and 39% residents. Clinicians overestimated the literacy of 81% of all LL patients. Their estimates did not differ significantly by their patients’ race, age or gender. Accuracy did not differ by clinicians’ own race or gender, but 76% of internists vs. 91% of cardiologists overestimated patients’ literacy (P = .08).

CONCLUSION:  Nearly half the urban adults with CHF enrolled in this randomized trial are low literate, especially those who are non-white, older, more severely ill and more frequently hospitalized. Their clinicians overestimated their health literacy in over 80% of cases. Future research should examine whether it is beneficial for clinicians and researchers managing complex chronic conditions, such as CHF, to ascertain whether patients are literate and address low literacy as part of efforts to improve patients’ health.

LOW INITIATION OF FRACTURE REDUCING THERAPY IN OLDER WOMEN WITH REDUCED BONE DENSITY.K.M. Ryder1; F.A. Tylavsky2; A.J. Bush2; D.C. Bauer3; E. Simonsick4; E. Strotmeyer5; T. Harris6; R.I. Shorr2. 1University of Tennessee Health Sciences Center, Memphis, Memphis, TN; 2University of Tennessee, Memphis, TN; 3University of California, San Francisco, San Francisco, CA; 4NIA, Bethesda, MA; 5University of Pittsburgh, Pittsburgh, PA; 6National Institute on Aging, National Institutes of Health, Bethesda, MD. (Tracking ID #116739)

BACKGROUND:  The use of antiresorptive therapy is low following osteoporotic fracture, but little is known about medication initiation following a diagnosis of low bone density.

METHODS:  Two year cohort study of well-functioning community-dwelling older women ages 70–79 residing in Memphis, TN or Pittsburgh, PA who were participants in the Health, Aging, and Body Composition (Health ABC) study between 1997 and 1999. At baseline, a hip DXA scan (Hologic) was performed and an alert letter was sent to participants and their physicians who had osteoporosis of the total hip. All participants and their designated physicians received the baseline bone mineral density information. Prescribed medications and OTC vitamin use were transcribed from drug containers annually. The primary endpoint was the use of fracture reducing therapy, categorized as antiresorptive therapy (ART: bisphosphonates, estrogen, calcitonin and raloxifene) or calcium and/or vitamin D (Ca-D).

RESULTS:  Of 1,557 women enrolled in Health ABC, 507 had evidence of low bone mineral density and/or fracture risk (T-score ≤2.0, or a T-score ≤1.5 with an additional major risk factor for fracture). Of these, the 371 not using fracture-reducing therapy at baseline were included in the study. Initiation rates within 2 years of baseline DXA were 12.7% for ART and 24.3% for Ca-D. Initiation of any fracture reducing therapy was 29.9%. In a multivariable model, predictors for initiating fracture-reducing therapy included: white race (relative risk = 3.5; 95% confidence interval 2.1–5.7), receiving a flu shot (2.1; 1.3–3.4), and medication insurance (1.6; 1.0–2.6). An alert letter (OR 2.3, 1–5.2) was a significant independent predictor of initiation, even after adjusting for absolute T-score at the total hip. Neither difficulty standing, mental status exam score, study site, reporting any difficulty with ADL's over 2 years, self-efficacy, receiving a mammogram, educational attainment, nor history of fractures were predictors.

CONCLUSION:  About 70% of high-functioning older women at risk for fracture have not initiated on fracture reducing therapy within 2 years of screening. Alert letters may increase initiation of therapy in appropriate candidates. Further efforts are needed to make older women and their physicians aware of guidelines for management of osteoporosis. Acknowledgement: NIA contract numbers N01-AG-6-2101; N01-AG-6-2103; N01-AG-6-2106

LOW LITERACY AND POOR HEALTH IN THE ELDERLY: THE HEALTH ABC STUDY.R. Sudore1; K. Mehta1; T. Harris2; A. Newman3; E. Simonsick4; S. Satterfield5; C. Rosano3; R. Rooks6; S. Rubin1; H. Ayonayon1; K. Yaffe1. 1UCSF, San Francisco, CA; 2NIA, Chevy Chase, MD; 3U of Pitt, Pittsburgh, PA; 4NIA, Bethesda, MA; 5U of TN, Knoxville, TN; 6KSU, Kent, OH. (Tracking ID #117197)

BACKGROUND:  Little is known about the correlates or health outcomes of low functional health literacy (LFHL). Thus, we determined the prevalence and correlates of LFHL and its association with poor health among well-functioning elders.

METHODS:  We studied 2,512 subjects in the Health, Aging, and Body Composition Study who completed a literacy evaluation with the Rapid Estimate of Adult Literacy in Medicine (REALM) in 1999–2000. REALM scores were categorized into LFHL (0–8th grade) or adequate literacy (≥9th grade). We assessed correlates of LFHL including demographics, socioeconomic status (SES), co-morbidities, cognition (3MS score), and health care access. The multivariate logistic regression model included all variables that were correlated with LFHL (P < .05). Adverse health variables included any of 3 ADL difficulties, IADL difficulty (0–5 scale), poor/fair self-reported health measured cross-sectionally, and 18-month mortality. We performed logistic regression to determine association between LFHL and adverse health and the multivariate model included demographics and SES.

RESULTS:  Participants had a mean age of 75.6 years, 48% were male, 38% were Black, and 24% had LFHL. In the multivariate analyses, correlates of LFHL included being male, Black, and having lower education, income, and cognition (Table). After adjusting for age, race, gender, education, and income, elders with LFHL had greater IADL difficulty (OR = 1.3, 95% CI 1.18–1.50), were more likely to have poor/fair health (OR 3.4, 95% CI 2.68–4.28), and were more likely to have died (OR 2.3, 95% CI 1.63–3.37).

CONCLUSION:  LFHL is prevalent among well-functioning elders and is associated with being male, black, having lower education, income, and cognitive scores. LFHL was also associated with worse health including functional disability, poor/fair health status, and mortality.

Table 18. Multivariate Analysis of the Correlates of LFHL
CharacteristicsUnadjusted OR (95% CI)Adjusted OR (95% CI)
Male1.6 (1.33–1.93)2.9 (2.11–3.95)
Black6.8 (5.53–8.34)2.8 (2.07–3.87)
Education (per/yr less)1.6 (1.49–1.64)1.3 (1.23–1.39)
Income <$10,0002.7 (2.34–3.05)2.7 (1.39–5.29)
3MS score (per/pt less)1.2 (1.18–1.20)1.1 (1.09–1.14)

LOW-LITERACY INTERVENTIONS TO PROMOTE DISCUSSION OF PROSTATE CANCER SCREENING: A RANDOMIZED CONTROLLED TRIAL.E. Justice1; J. Sharma1; J. Justice1; S. Kripalani1; C.A. Spiker2; T. Jacobson1; L.E. Laufman2; A.D. Weinberg2. 1Emory University, Atlanta, GA; 2Baylor College of Medicine, Houston, TX. (Tracking ID #115707)

BACKGROUND:  The efficacy of prostate cancer screening is uncertain, and professional organizations recommend that physicians discuss the potential risks and benefits with patients to promote informed decision making. However, few studies have tested strategies to encourage such discussions, particularly among high-risk populations. We examined the effect of two low-literacy, culturally-appropriate handouts on prostate cancer discussion and screening rates in an inner-city primary care clinic.

METHODS:  Randomized, blinded, controlled trial with concealed allocation. Subjects were men, age 45–70, with no history of prostate cancer, who presented for a routine appointment. They received one of the three handouts while waiting to see their physician—1) a two-sided patient education handout on prostate cancer screening (PtEd), 2) a one-sided flyer simply instructing patients to talk to their doctor about prostate cancer (Cue), or 3) a control handout. The PtEd and Cue handouts did not advocate for or against screening. After the physician appointment, patients completed a brief interview and the Rapid Estimate of Adult Literacy in Medicine (REALM). We abstracted additional data from clinic charts.

RESULTS:  Of the 281 eligible patients randomized, we obtained data on 250 (89% follow-up). Most subjects (91%) were African-American. While 87.6% reported having at least a 9th grade education, only 21.2% read at a 9th grade level or higher by the REALM. Overall, 48.4% of patients reported discussing prostate cancer during the appointment. Patients who received either intervention were more likely to have this discussion (50.0% PtEd and 58.0% Cue, vs. 37.4% control, P < .05 for each comparison to control), and when it took place, they more commonly initiated it (47.6% PtEd and 40.0% Cue, vs. 9.7% control, P < .01 for each comparison). Rates of prostate specific antigen (PSA) testing were higher in the intervention groups (14.1% PtEd and 12.8% Cue, vs. 2.4% control, P < .05 for each comparison). Rates of documented digital rectal examination (DRE) did not differ significantly (4.7% PtEd, 5.1% Cue, and 6.0% control).

CONCLUSION:  Two low-literacy interventions significantly increased patient-physician discussion of prostate cancer screening and PSA testing, but did not affect DRE rates. A detailed patient education handout and a short motivational flyer had similar efficacy in empowering low-literate patients to take a more active role in their health care by initiating the screening discussion.

MAKE THE BEST OF YOUR ORAL PRESENTATION: WHAT DO I FOCUS ON?C. Estrada1; S.R. Patel1; G.M. Talente1; M.S. Kraemer1. 1East Carolina University, Greenville, NC. (Tracking ID #115095)

BACKGROUND:  Presenters at professional meetings need to clearly summarize their work. Such presentation skills can be learned. We sought to identify features that make oral presentations successful by determining which aspects of a presentation experienced reviewers focused on.

METHODS:  Observational study over a 3-year period. Three reviewers observed presenters during oral sessions at a regional meeting. Reviewers identified the best features (“B”) and made suggestions for improvement (“I”) using an open-ended form. The authors reviewed selected comments and developed a coding system based on three domains; content, slides, and presentation style. Two raters blinded to the presenter then coded the comments. Disagreement was resolved by concurrent review.

RESULTS:  Twenty-nine presenters participated. The interrater agreement (kappa) was >0.75 in 19 of 32 coding domains (range 0.35–1). The content areas most frequently mentioned were: Key Concepts (B = 41.4%, I = 13.8%), Relevance (B = 41.4%, I = 3.5%). Comments about slides focused on: Text/Font (B = 31%, I = 41.4%), Graphics (B = 27.6%, I = 20.7%), and Clarity (B = 44.8%, I = 0%). Comments regarding presentation style focused on: Audience Engagement (B = 48.3%, I = 3.5%), Clarity (B = 51.7%, I = 0%), and Pace (B = 58.6%, I = 6.9%). Non-verbal communication, voice and eye contact, was noted in 83% (24/29) of the reviews as either a best aspect or an area for improvement.

CONCLUSION:  Features perceived as important during oral presentations relate to specific areas of content, clear slides, and a presentation style that was well paced, engaging, and clear. Non-verbal communication, specifically voice and eye contact, is important in oral presentations. Presenters should be mentored to focus on these areas to ensure successful presentations.

MANAGING ADULT SORE THROATS: REANALYZING A PUBLISHED COST-EFFECTIVENESS (C-E) ANALYSIS.R.M. Centor1; R. Matthew1. 1University of Alabama at Birmingham, Birmingham, AL. (Tracking ID #115688)

BACKGROUND:  A recently published C-E analysis of the management of adult sore throats concluded that empiric treatment was neither the most effective nor the least costly strategy. That analysis imputed an acute pharyngitis utility (0.95) from published utilities for equivalent diagnoses (dyspepsia and diarrhea). It also assumed symptomatic benefit only for group A strep (while Zwart also found benefit for non-group A strep in those with severe symptoms). We postulated that patients with varying severity of sore throat would have different utilities. We also hypothesized that including elicited utilities and a clinical benefit for treating non-group A pharyngitis would significantly alter the analysis.

METHODS:  We administered a time-tradeoff utility survey (using clinical scenarios) to 340 undergraduate and graduate students. We calculated both mean and median results. We then entered the utilities into a C-E analysis of the management of adult sore throats. The analysis also included a clinical benefit for treating non-group A pharyngitis.

RESULTS:  202 women and 138 men answered the survey. Almost 80% of the respondents had previous medical care for sore throat. Using a utility scale anchored by 1 (perfect health) and 0 (death), mild pharyngitis had a median utility of 0.975 (mean of 0.95). Severe pharyngitis had a median utility of 0.885 (mean of 0.861). Peritonsillar abscess had a median of 0.819 (mean of 0.864). Using these new utilities in a C-E model yielded different results than the recently published C-E analysis. We found the treat all strategy to have the greatest effectiveness in patients with severe strep symptoms. While culture was slightly less expensive, the treat empirically strategy had an incremental C-E per quality adjusted life day of less than $9 and per quality adjusted life year of less than $4,000. Sensitivity analysis showed that pharyngitis utility was the most important factor in making these decisions. Patients with mild strep symptoms had watchful waiting recommended.

CONCLUSION:  Directly assessing utilities for mild and severe pharyngitis changed the recommendations of a C-E analysis of adult sore throats. Our analysis shows that clinical pharyngitis severity (i.e., the degree of discomfort) should influence our decision to consider empiric treatment.

MARRIAGE AND LATINO ETHNICITY PREDICT PROSTATECTOMY IN HEALTHY MEN WITH LOCALIZED PROSTATE CANCER.T. Denberg1; B. Beaty1; F. Kim1; E.J. Perez-Stable2; J. Steiner1. 1University of Colorado Health Sciences Center, Denver, CO; 2University of California, San Francisco, San Francisco, CA. (Tracking ID #115713)

BACKGROUND:  Otherwise healthy men with early-stage prostate cancer have several treatment options of comparable effectiveness, including prostatectomy, radiation, and watchful waiting. Because uncertainty and controversy continue to surround treatment, we studied predictors of prostatectomy, the most aggressive form of therapy, through the latest years available in SEER-Medicare. Previously unexamined, we also included Latino ethnicity and marital status in the analysis.

METHODS:  A cohort of 27,802 white, black, and Latino men without comorbidities from the national linked, retrospective SEER-Medicare dataset (years 1995–1999) were used in an observational design. Bivariate and multinomial logistic regression were used to predict receipt of prostatectomy based on age, tumor grade, race, marital status (married vs. other), census-tract inferred educational attainment, and several interaction terms.

RESULTS:  In bivariate analysis, age, race/ethnicity, marriage, high school education, and tumor grade were predictive of prostatectomy (P < .0001). In multivariate analysis, age was the strongest predictor of treatment choice with men 70 significantly less likely to receive prostatectomy than men < 70 (adjusted OR = 0.17(0.16–0.18)). While tumor grade was also a strong predictor of prostatectomy, marriage was the second strongest predictor among all ages and without variation by race (adjusted overall OR = 1.95(1.83–2.09)). Finally, blacks were significantly less likely (adjusted OR = 0.62 (0.56–0.69)) while Latinos were significantly more likely than whites to receive prostatectomy (adjusted OR = 1.45(1.28–1.65)).

CONCLUSION:  In all racial/ethnic groups, marriage is strongly associated with receipt of prostatectomy in healthy men with localized prostate cancer. Hypothetically, this might reflect the influence of wives encouraging husbands to have more “definitive” therapy or perhaps clinicians recommending more aggressive treatment to married men. Second, Latinos appear to receive more prostatectomy than whites. This may be due to unknown cultural factors or, possibly, less access to radiation therapy. Future studies of treatment decision-making in early-stage prostate cancer should include Latinos and assess the role of spouses.

MEASURING PATIENTS’ EXPERIENCES WITH INDIVIDUAL PHYSICIANS.D. Safran1; M. Karp2; K. Coltin2; J. Ogren1; A. Li1; H. Chang1; W. Rogers1. 1TUFTS-New England Medical Center, Boston, MA; 2Massachusetts Health Quality Partners, Waltham, MA. (Tracking ID #116070)

BACKGROUND:  While standardized national measures of health plan performance were instituted in the 1990s, there are no standard performance measures for individual physicians or practices. Several large-scale initiatives to develop such measures are underway. The Massachusetts Ambulatory Care Experiences Survey Project involved a collaborative of 6 payers (5 commercial plans and Medicaid), 5 physician network organizations, and the Massachusetts Medical Society working through Massachusetts Health Quality Partners. This statewide demonstration project tested the feasibility and value of measuring patients’ experiences with individual primary care physicians.

METHODS:  The sample included adults drawn from the panels of 215 generalist physicians at 67 sites statewide. Sample frames were provided by the six payers. The Ambulatory Care Experiences Survey was administered by mail and telephone (May–Aug 2002). The survey produces 11 summary measures covering two dimensions of care: physician-patient interaction quality (communication, interpersonal treatment, whole-person orientation, health promotion, trust, relationship duration) and organizational features of care (access, continuity, integration, clinical team, office staff). Physician-level reliability was computed for all measures using intra-physician correlation and the Spearman-Brown Prophesy Formula. Variance components analysis was used to determine the influence of each level of the system (physician, site, network organization, plan) and system interaction effects on each measure, controlling for patient characteristics.

RESULTS:  Physician-specific samples were successfully constructed by linking payer-provided files. With samples of 45 patients per physician, all measures except two (clinical team, office staff) achieved physician-level reliability exceeding 0.70, and several exceeded 0.80. In variance components analysis, physicians and sites accounted for the majority of system-related variance, with physicians accounting for the majority on all “interaction quality” measures (range: 52.9% to 74.5%) and sites accounting for the largest share on “organizational” measures (range: 48.2% to 78.2%). Network organizations accounted for little variance, and health plans accounted for less then 3% on all measures. There were no significant plan-physician interactions.

CONCLUSION:  With national attention now focused on providing patient-centered care, this project demonstrates the feasibility of obtaining highly reliable measures of patients’ experiences with individual physicians and practices. Variance components results underscore the validity and importance of looking beyond health plans to individual physicians and sites as we seek to improve health care quality.

MEASURING PROFESSIONALISM ATTITUDES AND BELIELFS.A.L. Kalet1; H. Steven1. 1New York University, New York, NY. (Tracking ID #116092)

BACKGROUND:  Little is know about how learners’ feelings and beliefs about medical professionalim develop over time, predict receptivity to learning, and impact behavior. We report on the development of a Medical Professionalism Attitudes and Beliefs Questionnaire (MPABS) which is the first step in an IRB approved controlled pre/post evauation of a new professionalism curriculum and assessment process.

METHODS:  Questionnaire items were generated from; the written responses of 3 entering classes of medical students (n = 480) to the question “What is medical professionalism?”, student focus groups, a literature review. We adminstered the piloted and refined 44 item MPABS to 149, 1st year and 86, 2nd, 3rd, and 4th year students along with the Defining Issues Test (DIT-copyright 1979) a widely used and validated measure of moral reasoning which produces an individual overall and component scores (e.g. Anti-Social Measure, Religious Orthodoxy, Political Liberalism).

RESULTS:  Factor analysis (n = 234) revealed the 4 contructs which explained 52% of the variance with high Crohnbach's alpha coefficients (a); view of professional behavior in medical school (Behavior, 13 items, a = .808), knowledge of the definition of professionalism (Knowledge, 5 items, a = .706), idealism (Ideal, 11 items, a = .803), confidence in one's professionalism (confidence, 16 items, a = .849), and the importance of duty and social justice (Duty, 8 items, a = .736). While none of the contructs related to the the overall DIT score the component variable Anti-Social Measure was inversely associated with the variables knowledge, confidence, duty, and importance (ANOVA, P < .05). Idealism scores increased the greater the Religious Orthodoxy, and duty scores increased with increasing Political Liberalism (ANOVA, P < .05).

CONCLUSION:  While more will be done to further establish the validity of a briefer MPBAS to predict professionalism competency this instrument has impressive reliablity. We are studying both it's change over time and sensitivity to educational interventions.


BACKGROUND:  Learner feedback is the primary method for evaluating medical educators, yet there are few standards for measuring teaching proficiency. Our objective was to review the published literature on instruments for assessing clinical teaching and to summarize themes that will aid in developing universally appealing tools.

METHODS:  Electronic databases including MEDLINE, EMBASE, PsycINFO, ERIC, and Social Science Citation/Science Citation indexes were searched using the terms validity, medical faculty, medical education, evaluation studies, instrument, and the text word reliability. Over 330 studies were identified and additional references were extracted from these papers. Excluded were review articles, editorials, and qualitative studies. Reviewing abstracts from the original search revealed 21 papers describing instruments designed for evaluating clinical faculty by learners. Two investigators studied these papers and tabulated characteristics of the learning environments and validation methods. Salient themes amongst the evaluation studies were also determined.

RESULTS:  Most studies combined evaluations from both outpatient and inpatient settings. Wide ranges in numbers of subjects (10–711), evaluators (3–131), evaluations (30–7845), and instrument items (1–43) were observed. The most frequently encountered statistical methods were factor analysis and determining internal consistency reliability with Cronbach's alpha. The least common methods were the use of test-retest reliability and convergent validity between validated instruments. Seventeen domains of teaching were identified. The most frequently studied domains were interpersonal and clinical-teaching skills.

CONCLUSION:  Characteristics of teacher evaluations vary between inpatient and outpatient settings and between different learner levels, indicating that future investigations should utilize more narrowly defined study populations. Establishing an instrument's temporal stability and convergent validity should also be considered. Finally, existing data supports the validation of instruments simply comprised of interpersonal and clinical-teaching domains.

MEDICAL DEBT AND AGGRESSIVE DEBT RESTITUTION: PREDATORY BILLING AMONG THE URBAN POOR.T.P. O’Toole1; J. Arbelaez1; R.S. Lawrence1; L. Knickmeyer2. 1Johns Hopkins University, Baltimore, MD; 2The Baltimore Community Health Consortium, Baltimore, MD. (Tracking ID #116704)

BACKGROUND:  Health care providers are increasingly relying on collection agencies to recoup charges associated with medical care with medical debt now accounting for 40% of all personal bankruptcies Little is known about the prevalence of this practice in low-income communities and what effect it has on subsequent health seeking behavior.

METHODS:  Cross sectional survey at ten “safety net” provider sites in Baltimore, Maryland. Specific queries were made to underlying medical and mental health conditions, whether they had a current medical debt, actions taken against that debt, and what effect this has had on health seeking behavior. Separate multiple logistic regression models were developed for three depedent variables: (1) having an active medicla debt; (2) being referred to a collection agency; and (3) reporting a change in health seeking behavior as a result of that referral.

RESULTS:  Overall, 274 adults were interviewed. The average age was 43.9 years, 77.3% were African American, 54.6% male, 47.2% were homeless and 34.4% had less than a 12th grade education. 47.2% reported they currently had a medical debt (ave: $3,409) and 65.0% of individuals with a medical debt reported being referred to a collection agency. Overall, 46.0% of individuals referred to a collection agency reported it had affected their seeking subsequent care: 16.7% no longer went to that site for care; 12.7% delayed seeking care when needed; and 7.1% reported only going to an emergency departments now. In the multiple logistic regression model, having less than a 12th grade education (OR 2.6; 95% CI: 1.1–5.9) and being homeless (OR 3.7; 95% CI: 1.3–10.8) were independently associated with reporting an adverse health seeking consequence to a collection agency referral.

CONCLUSION:  Aggressive debt retrieval for medical care appears to be indiscriminately applied with a negative effect on subsequent health seeking behavior among those least capable of navigating the health system.

MEDICAL PRESCRIPTION WRITING BEHAVIORS OF PHYSICIANS IN TRAINING.L. Khrizman1; D.J. Shulkin1. 1Drexel University School Of Medicine, Internal Medicine, Philadelphia, PA. (Tracking ID #102547)

BACKGROUND:  Numerous efforts have been made to reduce medical errors, particularly in the area of medication management. Since medication errors account for a considerable part of all medical errors, in this study we examined both the actual practices of residents-in-training of prescription order writing and the attitudes of this group towards strategies to reduce errors.

METHODS:  A survey, comprising of 15 questions (N = 68 residents in the Internal Medicine Department) was distributed which collected information on factors including the year of residency, use of a hand-held device to look up medications and resident awareness of hospital guidelines of order writing. Residents were also asked to write-out a series of orders in the same way that they would on an actual hospital order form. Lastly, residents were asked their attitudes toward order-writing guidelines, as well as what implementations they felt would reduce medication order-writing errors.

RESULTS:  Sixty-four surveys were collected (response rate 94%). Analysis of the resident orders of medications demonstrated significant deficiencies in medication order writing practices. The most common error in medication order writing was the inappropriate use of abbreviations of the QD/QOD/HS type (58% error rate), followed by incorrect medication names and dosage usage (48% error rate), use of a “U” symbol for units (41% error rate), improper route of administration (34% error rate), using a “d” for days (33% error rate), medication orders that were not dated or timed (36% error rate), and inappropriate abbreviations for “micrograms” (23% error rate). The medication writing error that was made with the least frequency was using a zero after the decimal point in medication dosages (2% error rate). Analyzing resident attitudes towards medication order writing and strategies to reduce errors we found that 47% of residents used a hand-held device to look up medications, with 95% believing they would benefit from a hand-held device. Seventy-eight percent said that they would derive benefit from an order form that prompted standardized order writing. Sixty-nine percent stated that they would benefit from a teaching session on order writing. Nearly 97% said that a call from a pharmacy regarding order-writing errors would help in reducing mistakes.

CONCLUSION:  This study showed that there is a significant error rate ranging from 2% to 58%, depending on the error type, in prescription writing among residents. The most common error type (58%) turned out to be an improper use of abbreviations. Interventions such as teaching sessions, hand-held devise usage, a better order form and more error reporting from the pharmacy would all be well-accepted and bring potential benefit into reducing medication errors according to the residents.

MEDICAL STUDENT RESOURCE USE AND KNOWLEDGE ACQUISITION IN THE MEDICINE CLERKSHIP.K. DeZee1; G. Denton1; S. Durning1. 1Uniformed Services University of the Health Sciences, Bethesda, MD. (Tracking ID #116551)

BACKGROUND:  The third year internal medicine clerkship at the Uniformed Services University (USU) is a required 12-week rotation. While knowledge acquisition through focused reading about assigned patients and general topics in internal medicine is a principal goal of the rotation, there are no assigned textbooks or reading materials for the course. Students are given a print copy of Harrison's Textbook of Medicine and unlimited electronic access to several web based resources, including MD Consult, UpToDate, and many full-text journals. Our specific questions were: 1) What resources do 3rd year IM clerkship students use? and 2) Do these choices affect knowledge acquisition, as reflected by the National Board of Medical Examiners (NBME) shelf exam?

METHODS:  This was a prospective cohort study including three consecutive groups of students rotating through the clerkship at this medical school. On the first day of the clerkship, students took a faculty-developed 100-question multiple-choice pretest, which reflects knowledge base and is well correlated with USMLE Step One. During the last week of the clerkship, students took the NBME shelf exam in internal medicine. Immediately prior to the NBME exam, students completed a one-page survey instrument, ranking the resources they used during the clerkship and listing print or electronic for each. Student NBME scores were standardized and were used as an assessment of knowledge acquisition. Analyses of univariate comparisons were done using Chi square and ANOVA, multivariate analyses, controlling for potential confounding variables, were done using linear regression.

RESULTS:  One hundred and eight (108) students participated (82% of cohort); 61% reported using an electronic reference for their primary (most important reference listed) resource and 39% a print reference. At baseline, there were no differences in age, sex, race, or pre-rotation knowledge between the two groups. At the end of the rotation, there was no difference in mean NBME scores between the two groups (Electronic: 71.71, sd: 7.17; Print: 71.67, sd: 7.29, P = .97). In multivariate modeling, baseline knowledge and student age explained 32% of the variance in NBME scores. The study had 80% power to detect a difference of four points (5.6%) in the NBME exam score, by type of primary reference used.

CONCLUSION:  The majority of students rely on electronic resources for their most important reference for their 3rd year internal medicine clerkship, with UpToDate dominating. The choice of an electronic or print resource for the primary reference material was not associated with differential acquisition of knowledge. Based on our results, students should be allowed to choose their primary reference based on their own preferences.

MEDICARE FIRST DOLLAR COVERAGE OF ACE-INHIBITORS FOR BENEFICIARIES WITH DIABETES SAVES MONEY AND LIVES.A.B. Rosen1; M.B. Hamel1; M.C. Weinstein2; D. Cutler2; S. Vijan3. 1Beth Israel Deaconess Medical Center, Boston, MA; 2Harvard School of Public Health, Boston, MA; 3University of Michigan, Ann Arbor, MI. (Tracking ID #116162)

BACKGROUND:  Diabetic nephropathy is the leading cause of end stage renal disease (ESRD) in the United States and is associated with marked morbidity, mortality, and costs. ACE-Inhibitor (ACE) use in diabetics slows progression of renal disease and also reduces cardiac morbidity and mortality. Six percent of the Medicare budget is spent annually on the care of the 0.6% of the Medicare population with ESRD. The objective of this study was to assess the health outcomes and budgetary impact to Medicare of first dollar (i.e. no cost-sharing) coverage of ACE for elderly beneficiaries with diabetes.

METHODS:  Clinical events, survival, and Medicare costs for 65 yo diabetic patients, with and without ACE, were assessed using a Markov cohort model. We assumed that ‘no coverage’ resulted in 40% ACE use (NHANES 4) at no cost to Medicare, and ‘Medicare coverage’ increased ACE use by 20% (to 60% overall) based on price elasticity data from the literature. Modeled outcomes included progression of renal disease, cardiovascular events, life expectancy, quality-adjusted life years (QALYs), lifetime costs (2003 US $), and incremental cost-effectiveness ratios. Costs and benefits were discounted at 3%; all analyses took a Medicare perspective. One-way and multi-way sensitivity analyses were performed on uncertain model parameters. National aggregate estimates were based on the age distribution of the current 6.2 million Medicare beneficiaries with diabetes.

RESULTS:  Standard Medicare coverage results in a discounted lifetime cost of $80,129 and quality-adjusted life expectancy of 10.44 QALYs. The addition of first dollar ACE coverage decreased lifetime costs to $79,005 and increased benefits to 10.78 QALYs. Thus, Medicare ACE coverage saves both lives and money. Results were robust to a wide range of renal and cardiac risk reductions, costs, utilities, and discount rates. Results were most sensitive to the cost of ACE (threshold at which no longer cost-saving, drug price >1.8 times the annual average wholesale price) and the impact of coverage on utilization rates (threshold at which no longer cost-saving, increase in ACE use of <7% ). Applying our model to the 6.2 million current Medicare beneficiaries with diabetes, first dollar coverage of ACE would result in a total of 1.3 million QALYs gained and 4.88 billion dollars saved over this cohort's life.

CONCLUSION:  Medicare first dollar coverage of ACE-Inhibitors extends life and reduces Medicare program costs. A reduction in program costs from one cost-saving intervention will result in more money to spend on other health care needs of the elderly.

MEDICATION COSTS: THE ROLE PHYSICIANS PLAY WITH THEIR SENIOR PATIENTS.M. Beran1; M. Laouri2; M.J. Suttorp3; R.H. Brook3. 1Park Nicollet Health Services, Minneapolis, MN; 2Genentech, South San Francisco, CA; 3RAND, Santa Monica, CA. (Tracking ID #116494)

BACKGROUND:  Many patients over 65 experience a financial burden from out-of-pocket (OOP) medication costs due to the number of chronic conditions, number of medications, and the lack of adequate prescription drug coverage. The inability to afford necessary medications may lead to cost reducing strategies, such as skipping or stopping medication, and subsequent adverse outcomes. Physicians who care for seniors have the opportunity to address the OOP costs of medications. This study examined how often physicians discuss OOP medication costs with seniors, and the cost reducing strategies they employ when these discussions occur.

METHODS:  A cross sectional, random sample of 1,200 internal medicine and family practice physicians in CA, selected from the AMA Masterfile, were mailed self-administered questionnaires. We asked how often physicians discuss OOP medication costs with seniors, what factors influence these discussions, and what cost reducing strategies physicians recommended to patients who express difficulty affording their medications. We used logistic regression to relate physician and system characteristics to discussions of medication costs.

RESULTS:  We obtained completed surveys from 678 of 1,098 (62%) eligible physicians. 68% of physicians reported medication cost as “somewhat” or “very” important when prescribing to seniors, and 43% reported discussing medication costs with more than half of their senior patients in the last 30 days. 40% reported that there was at least one time in the last 30 days when they did not discuss cost but wish they had. The most common reason given was “I ran out of time” (14%). When cost was discussed, 65% of physicians reported that patients initiated the discussion. Predictors of discussing medication cost with half or more of senior patients were group practice setting (vs.other practice settings) and physician rating of cost as of high importance when prescribing a medication for a senior patient (OR 1.5, CI 1.07–2.24). Physician gender, race, specialty, years in practice, patient volume, and percentage of senior patients were not associated with frequency of discussing medication costs. The most common cost reducing strategy used by physicians was generic substitution (33%) followed by offering samples (25%).

CONCLUSION:  More than two-thirds of physicians believe that OOP medication costs are important, and a surprisingly large percentage of physicians are spending time discussing cost with their senior patients. These results are encouraging, however, due to time constraints in practice, improvement of these numbers may be dependent upon development of a tool to identify patients for whom medication costs represent a substantial financial burden.

MEDICATION KNOWLEDGE AMONG UNDERSERVED PATIENTS IN TWO COMMUNITY CLINICS.N. Kim1; J. Talwalkar2; E.S. Holmboe1. 1Yale University, New Haven, CT; 2Yale University School of Medicine, New Haven, CT. (Tracking ID #117200)

BACKGROUND:  The effectiveness of drug therapy for chronic diseases depends on patients’ knowledge and adherence to medications. Medication mismanagement results in 10% of hospital admissions, 25% of malpractice suits, 50% of therapeutic failures and 2.5 million emergencies. Little is known about under-served patients’ knowledge of their medications, a group disproportionately affected by chronic disease. The purpose of this study was to assess underserved patients’ knowledge of their medications.

METHODS:  This prospective observational study evaluated patients = 18 years receiving ongoing care at two continuity clinics of the Yale Primary Care Program in Waterbury, CT. These clinics serve largely Medicaid or self-pay patients. To assess medication knowledge, we conducted a 10-minute semi-structured interview prior to the patient-physician encounter. Patients were asked to describe the names, doses, frequencies, indications, side effects, and safety of their current medications. We reviewed charts to obtain a reference standard for patients’ current medication regimens.

RESULTS:  One hundred twenty-five interviews were completed. The median age was 46 years, 60% were female, 68% were non-white, 60% had less than a high school education, and 70% had an annual income =$10,000. Nearly half suffered from at least one chronic condition. Patients were on a median of 5 drugs, most commonly anti-hypertensives, glucose-lowering agents, and anti-depressants. Although 89% patients report being satisfied with their personal medication knowledge; they could cite only 50% of their medication names, 25% of medication doses, 50% of medication frequencies, 50% of medication indications, and none (0%) of their medications’ side effects. Paradoxically, when asked what information they felt they needed in order to take their medications safely, 47% responded that side effects were essential information. Forty-three percent of patients reported taking at least one medication that was not documented anywhere in the medical record. In total there were 84 discordant medications of which 57% were prescription drugs.

CONCLUSION:  Under-served patients have poor knowledge of their medications, yet report being satisfied with their medication understanding. Although poor knowledge does not necessarily portend poor adherence to medications, without certain pieces of knowledge, such as medication frequency, it is highly unlikely that patients are taking their medications correctly. This coupled with the large discrepancy between medications taken by the patient and those documented in the medical record as well as patients’ complete lack of knowledge of medication side effects greatly threaten not only medication adherence and achievement of treatment goals, but also patient safety.


BACKGROUND:  Public health officials and clinical physicians were once integral partners. Recent events such as the SARS outbreak and the ongoing threat of bioterrorism have exposed a lack of coordination between the two that might be addressed by education. While a prior survey of medicine residents demonstrated interest in improving education on public health topics, the views of practicing physicians are not well studied. To address this deficit, we surveyed general internists on their attitudes, knowledge and clinical practices regarding public health issues.

METHODS:  In 2003, a city-wide sample of general internists (n = 260) received 3 mailings of a survey. Surveys included 15 opinion items, 3 knowledge items, 7 items on current clinical practices and 2 items on preferences for a future educational seminar. Participants had to be practicing in the city of Philadelphia, board certified in Internal Medicine and identify themselves as a primary care physician or general practitioner.

RESULTS:  The overall response rate was 63%. The respondents were 40% female, averaged 11 years in practice and 8 half-days of clinical practice per week. Most (69%) worked in a group of 4 or more physicians. Attitudes were supportive of public health. 75% saw themselves as a component of the public health system and 90% thought they should be able to recognize new patterns of disease. Most (80%) felt public health topics were insufficiently taught during residency. Nearly half believed it unimportant for them to participate in community-related health activities. Only a third had ever driven or walked around the neighborhood served by their practice, noting characteristics influencing the lifestyle or health of patients. Knowedge was limited: 42% did not know they had to personally submit a report to the city to report certain diseases and only 16% knew that there were over 40 reportable diseases. The public health tools most commonly utilized in practice were methods to identify patients requiring certain preventive services and a system to report diseases to public health officials. The most common reasons for not utilizing public health tools in practice were lack of time and money. Almost all (85%) respondents were interested in a CME-eligible seminar on public health issues and skills. A third were willing to pay for this seminar, with most preferring it take place at the physician's office during lunch or via the internet.

CONCLUSION:  Practicing general internists believe themselves to be a component of the public health infrastructure yet feel inadequately trained to perform their duties. They lack a basic understanding of their role as agents on the front lines of both public health and clinical medicine. As a result, they fail to perform some basic tasks required of them. A brief and targeted intervention could be a first step to correct these deficiencies.

MEETING THE MANDATE TO CARE FOR LOW-INCOME VETERANS. J.A. Long1; J.P. Metlay2. 1Philadelphia VA Center for Health Equity Research and Promotion, Philadelphia, PA; 2University of Pennsylvania, Philadelphia, PA. (Tracking ID #116921)

BACKGROUND:  Part of the Veterans Health Administration (VHA) mandate is to care for low-income veterans. Previously we showed that in 1992 all low-income veterans, compared to high-income veterans, are less likely to receive outpatient and preventive care and more likely to have unmet medical needs. The objectives of this study were to determine current patterns of care for low-income veterans.

METHODS:  The data for this study comes from a telephone survey of 20,048 veterans—the 2001 National Survey of Veterans. The dependent variables were having received the following services in the last 12 months: an out-patient visit; an overnight stay in the hospital; and a prescription medication. Independent variables included site in which the veteran received care (any VHA use, no VHA use, no care), annual income, age, sex, race, marital status, education completed, having additional health insurance, number of chronic conditions, having any difficulties with activities of daily living, and having a service related disability. Results were weighted to reflect the probability of sampling.

RESULTS:  In 2001, 21.3% of all low-income veterans reported receiving no care from a VHA or non-VHA site, which was a decrease from 47.2% in 1992. Overall, compared to veterans who did not use the VHA, VHA users were more likely to get out-patient care, to be hospitalized, and to receive prescription medications. After adjusting for all independent variables, among veterans who used VHA services, there was no association between income level and receiving out-patient care or being hospitalized. However, among veterans who did not use the VHA, compared to veterans whose annual income was over $50,000, veterans whose income was below $20,000 were less likely to receive out-patient care (OR = 0.68, 95% CI 0.58–0.81) and more likely to be hospitalized (OR = 1.24 95% CI 1.04–1.49). Finally, among VHA users, lower income veterans were more likely to receive a prescription medication while among non-VHA users lower income veterans were less likely to receive a prescription medication. For example, compared to veterans whose income was over $50,000, veterans whose income was $20,000 or less and used the VHA had an odds of receiving medications of 1.99 (95% CI 1.22–3.24) and for veterans who did not use the VHA the odds were 0.58 (95% CI 0.48–0.70).

CONCLUSION:  In 2001 utilization of the VHA diminished the number of low-income veterans getting no medical care as well as unwanted income disparities in out-patient care and hospitalizations while increasing access to expensive prescription medications. In a time when Americans are having greater difficulty accessing care, the VHA has made great strides in addressing the mandate to care for low-income veterans.

MENTAL ILLNESS AND PATTERNS OF CHOLESTEROL TESTING IN A POPULATION OF U.S. VETERANS.R.A. Kaplowitz1; R. Scranton1; D.R. Gagnon1; J.W. Levenson1; C. Cantillon1; L. Fiore1; J.M. Gaziano1. 1Massachusetts Veterans Epidemiology Research and Information Center, Boston, MA. (Tracking ID #116574)

BACKGROUND:  Although current data suggest that people with mental illness have worse outcomes from cardiovascular disease, disparities in cholesterol testing associated with mental illness have not been well studied.

METHODS:  Using a retrospective cohort study design, we analyzed the ICD-9 diagnosis codes, pharmacy profiles, and laboratory data of subjects who used Veterans Affairs (VA) New England Healthcare System outpatient services during at least two months of the 42-month study period. We defined mental illness as the concurrence of an ICD-9 code for an Axis I psychiatric disease and a prescription for a psychiatric medication; the control group had neither a psychiatric diagnosis nor a prescribed psychiatric medication. The outcome variable was one serum cholesterol test that included a high-density lipoprotein value. We used logistic regression to model the odds of cholesterol testing. We stratified by quartile of outpatient visit frequency and adjusted for number of months of outpatient service use within each stratum, age, race, gender, baseline hypercholesterolemia, cardiovascular disease, tobacco use, diabetes mellitus, hypertension, alcohol abuse, and substance abuse.

RESULTS:  Of the 64,966 subjects included in the study, 22.6% had a mental illness and 74.3% received a cholesterol test. Subjects with mental illness were less likely than subjects without mental illness to receive a cholesterol test in the first and second quartiles of outpatient visit frequency (Table). The odds of cholesterol testing peaked in the second quartile for subjects without mental illness, and reached a plateau for subjects with mental illness between the second and fourth quartiles.

CONCLUSION:  Mental illness was associated with decreased odds of cholesterol testing in the population that used less than the median frequency of VA outpatient services. Complex relationships between visit frequency and odds of testing were found for patients both with and without mental illness.

Table 19. Odds of Cholesterol Testing by Mental Illness Status and Outpatient Service Use
Quartile (Months of Outpatient Service Use)No Mental Illness OR (95% C.I.)Mental Illness OR (95% C.I.)
1 (2–10)1.00 (ref)0.47 (0.41–0.54)
2 (11–18)1.52 (1.42–1.64)0.74 (0.66–0.82)
3 (19–30)1.13 (1.01–1.27)1.03 (0.90–1.18)
4 (31–42)0.53 (0.44–0.63)0.77 (0.63–0.94)

MENTORING IN INTERNAL MEDICINE RESIDENCY: THE HOUSESTAFF PERSPECTIVE.A. Castiglioni1; G. Heudebert1. 1University of Alabama at Birmingham, Birmingham, AL. (Tracking ID #117350)

BACKGROUND:  Mentoring in academic medicine has been linked to greater career satisfaction and academic success. Despite its recognized importance, there is paucity of literature examining internal medicine residents’ needs and perceptions on mentoring. The objectives were to 1) assess internal medicine residents’ attitudes towards mentoring in residency, 2) identify perceived benefits of successful mentoring.

METHODS:  A two-page questionnaire was sent to all internal medicine and medicine-pediatrics residents. The instrument included questions on demographics, future career plans, attitudes towards mentoring, mentor/mentee demographics, and perceived benefits of mentoring.

RESULTS:  Among the 99 respondents (75% response rate), 30% were female, 32% were PGY1, and 66% indicated interest in fellowship training. On 6 attitude statements, greater than 90% of residents agreed that mentoring is important, that program directors (PD) should promote mentoring (80%) and that PD should encourage faculty participation (87%). When asked to design a structured mentoring program for residents, respondents favored individual faculty mentoring with regular scheduled meetings. Respondents also favored assignment of a mentor during the PGY1 year (51%) and utilizing the advise of the PD to identify such mentor (48%). Among those residents who currently had a mentor (31%), 90% could identify up to 2 mentors (mean 1.5). Most mentors were the same gender (78%) and same ethnic group (64%) as the mentee; a minority of mentees had changed mentors (15%). On a five-point scale (with 5 indicating strong agreement) respondents perceived mentors as most beneficial in career planning (mean 4.1) and growth and development (mean 4.0). Mentoring was felt least beneficial in coping with stress (mean 3.4) and development of clinical skills (mean 3.2). Of those without a mentor, 52% would have liked a mentor assigned. We found no relationship between having a mentor in residency and residents’ gender, future career plans or having had a mentor in medical school.

CONCLUSION:  A minority of our residents have established a mentor. Overall, residents favor mentoring and perceive the program director to have mostly a facilitator role. Respondents favor early identification of a mentor and a flexible structure for mentoring. Residents perceive mentoring as beneficial for personal growth and career choices.

META-ANALYSIS OF WARFARIN PLUS ASPIRIN AFTER MYOCARDIAL INFARCTION. M. Rothberg1; C. Celestin2; J. Cook2. 1Tufts University, Springfield, MA; 2Baystate Medical Center, Springfield, MA. (Tracking ID #115840)

BACKGROUND:  Patients with history of myocardial infarction (MI) are at increased risk for reinfarction. Although adding warfarin to aspirin decreases the incidence of MI and stroke, the inconvenience of therapy and the increased risk of bleeding have limited its use. Because not all patients are likely to benefit equally from warfarin therapy, we conducted a meta-analysis to quantify the benefit, then applied it to individuals at varying degrees of cardiovascular and bleeding risks.

METHODS:  We searched MEDLINE from 1990 for randomized controlled trials comparing high dose warfarin (INR > 2) plus aspirin with aspirin alone in patients with history of MI or acute coronary syndrome. We evaluated the endpoints of recurrent MI, thrombogenic stroke, death, major bleeding, and minor bleeding. Hypothetical patients were then divided into tertiles of cardiovascular and bleeding risk based on published data from 2,677 post-MI patients and the Outpatient Bleeding Risk Index.

RESULTS:  Ten articles (7,689 patients) met the study criteria. Nine found a risk reduction for MI in the warfarin plus aspirin group, but only one reached statistical significance. Combining all trials, adding warfarin to aspirin resulted in a relative risk of 0.67 (95% CI 0.56 to 0.81) for myocardial infarction, 0.40 (95% CI 0.24 to 0.65) for ischemic stoke, and 2.3 (95% CI 1.6 to 3.3) for major bleeding. There was no difference in mortality. Applying the results to risk tertiles, patients with low (1%) or medium(8%) bleeding risk based on personal risk factors stand to benefit most from warfarin therapy, especially if they are at medium (8%) to high (21%) risk of reinfarction.

CONCLUSION:  For patients suffering a prior MI, warfarin plus aspirin decreases morbidity from recurrent MI and stroke. Patients at low risk of bleeding benefit most.

Table 20. Expected change in number of MIs/strokes/major bleeds by adding warfarin to aspirin per 1,000 patient-years, by risk group
 Low (1%)Med (8%)High (30%)
AnnualLow (3%−10/–5/6−10/–5/167
MIMed (8%)−26/–12/6−26/–12/167
RiskHigh (21%)−69/–30/6−69/–30/167

MISCONCEPTIONS ABOUT SCREENING AND PREVENTION: IMPLICATIONS FOR INFORMED DECISION-MAKING.T. Denberg1; S. Wong2; A. Beattie2; E.J. Perez-Stable2. 1University of Colorado Health Sciences Center, Denver, CO; 2University of California, San Francisco, San Francisco, CA. (Tracking ID #115709)

BACKGROUND:  Informed decision-making about cancer screening requires that patients have a correct understanding of a test's purpose, benefits, and risks. Understanding patients’ actual beliefs in these areas may enhance informed decision-making.

METHODS:  Semistructured interviews were transcribed and thematically coded using a convenience sample of 24 socioeconomically diverse white, African American, Latino and Chinese American women recruited from community settings. Interviews focused on participants’ ideas related to cancer prevention and screening.

RESULTS:  A majority of women doubted that the emergence of cancer can be prevented, even while asserting that, if caught early enough, death can be avoided. Lack of knowledge of pre-malignancy was widespread, reinforcing the equation of prevention with cancer survival. Screening was often described in terms that invoked the magical notion of bodily protection rather than prevention: it is therapeutic by itself without regard to cancer or for gathering information to prompt medical follow-up. The most extreme manifestation of this idea was that a failure to get screening can actually cause cancer. A subset of respondents believed that screening is indicated only with symptoms. These beliefs cut across ethnic and social class boundaries.

CONCLUSION:  Women expressed cancer-related beliefs characterized by inaccuracies, distortions, and over-simplifications. Many of these beliefs may go unrecognized in clinical settings yet have a profound influence on risk communication and, therefore, informed decision-making. Effective communication depends, first, on clinicians and patients sharing an accurate understanding of background concepts such as “prevention,”“screening,”“cancer,” and “risk.” Furthermore, screening decisions may be based on beliefs unrelated to risk.

MISSED OPPORTUNITIES: REFUSERS OF ROUTINE HIV TESTING IN AN URGENT CARE SETTING.R.V. Liddicoat1; E. Losina2; M. Kang1; K.A. Freedberg1; R.P. Walensky1. 1Massachusetts General Hospital, Boston, MA; 2Boston University School of Public Health, Boston, MA. (Tracking ID #115690)

BACKGROUND:  Efforts to increase routine HIV testing and case identification in the US are hindered by high test refusal rates. Though little is known about people who refuse voluntary HIV testing, data suggest that refusers may have higher rates of HIV infection than acceptors. Our objective was to identify those likely to refuse HIV testing in order to improve future testing efforts.

METHODS:  We developed a new routine, voluntary HIV testing program at 4 Massachusetts urban, hospital-associated urgent care centers located in areas of high HIV prevalence. In a confidential setting, all patients registering were asked if they were interested in routine HIV testing. Those who refused were asked for demographics and reasons for test refusal. We performed univariate and multivariate logistic regression to identify demographic factors and reasons associated with test refusal.

RESULTS:  Of 10,354 patients offered HIV testing from 1/02–12/02, 7,073 (68%) refused. Refusers had a median age of 35 years, 48% were male, 69% English speaking, 35% white, 33% African American, and 23% Hispanic. Refusers were older, more likely to be female, English speaking and white compared to acceptors (P < .001). In multivariate analysis adjusted for site, refusal was associated with being white and English speaking compared to non-white, non-English speakers [OR 1.24 (95% CI: 1.09–1.41)], female gender[OR 1.23 (1.20–1.35)], older age (P < .001) and having a higher educational level (P < .001). Three of the sites had refusal rates of 77–80%, the fourth site's refusal rate was 49% (P < .001). Most common reasons for test refusal were “felt not at risk” (42%), “previously tested” (37%), and “felt too sick” (14%). Two percent refused because they were already known to be HIV-infected, while only 1% felt the information was “too personal.” Among test acceptors, overall HIV prevalence was 1.8%; prevalence among repeat testers was 2.1%. Demographic groups that refused more often had high prevalence rates of HIV among those who accepted; 1.4% among women, 1.3% among those with greater than a high school education, and 0.9% among patients >40 years.

CONCLUSION:  Two thirds of patients offered routine HIV testing in an urgent care program refused. Refusers were more likely to be female, older, white, and more educated. HIV prevalence among similar demographic groups who accepted testing generally exceeded 1%, highlighting the need to test groups that are more likely to refuse. Efforts should be made to educate patients about the yield of testing as well as the value of repeated testing in the setting of continued risk behavior.

MONITORING DEPRESSION WITH A BRIEF SELF-REPORT SCALE (PHQ-9).K. Kroenke1; J. Unutzer2; C.M. Callahan1; A.J. Perkins1; B. Löwe1. 1Indiana University and Regenstrief Institute, Indianapolis, IN; 2University of California, Los Angeles, CA. (Tracking ID #115147)

BACKGROUND:  Although efficacous treatments are widely available, only 22% of Americans with major depression are treated adaequately. Self-adminstered screening instruments can efficiently help to diagnose and monitor depression in patient care and research. The 9-item depression module from the Patient Health Questionnaire (PHQ-9) is increasingly regarded as a standard instrument for assessing depression in primary care, but its sensitivity to change has not yet been established. In this study, we aimed to investigate sensitivity to change of the PHQ-9 in comparison to validated self-report scales, as well as to interview-based criterion standards for depression diagnosis.

METHODS:  Data from the multisite IMPACT clinical depression trial was used to compare sensitivity to change of the PHQ-9, the SCL-20 depression scale, and the SF-12 Mental Component Summary (MCS-12). Depression diagnoses from two sources—the Structured Clinical Interview for DSM-IV (SCID), and the Depression Clinical Specialist (DCS)—were used as companion criterion standards for persistent major depression, partial remission, and full remission at 3 and 6 months. We analyzed data on 434 patients (63% female, mean age 71 years) from the intervention group of the IMPACT trial.

RESULTS:  The PHQ-9 sensitivity to change as measured by effect size was equal to or greater than either the SCL-20 or MCS-12 at both 3 months (−1.3 vs. −0.9. vs. 0.9) and 6 months (−1.3 vs. −1.1 vs. 0.8). With respect to the DCS's depression diagnosis at 6 months, effect sizes of PHQ-9 change were −2.6 for the patient group with full remission, −1.8 for the patient group with partial remission, and −0.9 for the unchanged patient group. Using the SCID as a more conservative criterion standard, PHQ-9 change scores still discriminated well between the 3 groups.

CONCLUSION:  Half the length of the SCL-20 with the added advantage of establishing depression diagnoses, the PHQ-9 is also comparable as an outcome measure in terms of its sensitivity to change. Moreover, the PHQ-9 discriminates well between patients with full, partial, and no remission of depressive symptoms.

MORE TIME, MORE WORK: TEAM STRUCTURE AND WORKLOAD EFFECTS ON PATIENT OUTCOMES IN AN ACADEMIC GENERAL MEDICINE INPATIENT SERVICE.M.K. Ong1; A. Bostrom2; A. Vidyarthi2; C.E. McCulloch2; A.D. Auerbach2. 1Stanford University, Stanford, CA; 2University of California, San Francisco, San Francisco, CA. (Tracking ID #117243)

BACKGROUND:  Housestaff work hour reduction mandates have forced residency programs to change team structures to limit resident and intern workloads. However, there is little research on how resident workloads influence patient outcomes.

METHODS:  A retrospective cohort of 5,743 patients admitted to the University of California, San Francisco Medical Center general internal medicine wards between July 1998 and June 2001 was analyzed. We examined the effect of housestaff team structure variation (total number of teams, the number of teams admitting per day, and the number of interns on a team) on a combined outcome of 30-day readmission and death. We also examined the effect of daily workload (i.e. number of admissions or discharges, team patient loads, and overall service census), patient demographics, and weekend admission. Significant predictors were analyzed for associations with length of stay (LOS).

RESULTS:  During the study period, 5,743 patients were admitted; the average age was 62 years (SD = 20) and 48% were male. More patients were admitted to two-intern teams (83%) than one-intern teams (17%). Teams averaged 6 admits per admit day (SD = 3) and a daily team census of 10 patients (SD = 4). Significant (P < .05) predictors of 30-day readmission and death are reported below with odds ratios (OR) and 95% confidence intervals (CI). Total team admits during the month and on the admit day significantly (P < .05) increased LOS by 0.6% and 4% per additional admit, respectively.

CONCLUSION:  Reduction in readmission and death associated with team workload variables is likely due to increased lengths of stay caused by high patient loads. High overall service loads also lead to more readmissions and deaths. The worse outcomes associated with more team interns may reflect less time spent by residents working with individual interns. The beneficial effect of weekend admissions may be due to teams having more time to spend working up their patients than on weekdays.

PredictorOR (95% CI)
Number of team interns1.38 (1.02–1.87)
Total service admits on admit day1.06 (1.01–1.11)
Total team admits during admit month0.99 (0.97–1.00)
Total team admits on admit day0.93 (0.87–0.99)
Weekend Admission0.74 (0.55–0.98)

MORTALITY FOLLOWING SUBSTANCE DETOXIFICATION.J. McCary1; J.H. Samet2; D.M. Cheng2; M. Larson3; R. Saitz2. 1Boston Health Care for the Homeless Program, Boston, MA; 2Boston University, Boston, MA; 3New England Research Institutes, Watertown, MA. (Tracking ID #116943)

BACKGROUND:  Alcohol and other drug dependent adults are often not engaged in primary medical care yet they are at risk for serious health consequences. Mortality in this population has not been well described. We hypothesized that despite relatively young age, mortality in adults after detoxification would be substantial, causes of death would be related to addiction, and patient characteristics at the time of detoxification could identify those at higher risk for death.

METHODS:  Between 1997 and 1999, a prospective cohort of adults without primary medical care in an inpatient urban detoxification unit for alcohol, heroin, or cocaine underwent standardized research assessments and was followed for 2 years. Deaths were identified at the time of attempted follow-up contacts and by matching study subjects with 1997–2001 National Death Index data. Underlying causes of death were identified from death certificates. Unadjusted and adjusted (for age and gender) Cox proportional hazards models were fit with death as the outcome.

RESULTS:  Subjects were: 76% male; 46% white, 37% black, 11% latino; 60% uninsured; 47% homeless; 86% smokers; and 47% with chronic medical illnesses. Forty percent reported alcohol, 27% heroin, and 33% cocaine as their substance of choice. The median age was 35. Of the 470 subjects, 27 (6%) died during the 4-year follow up. National Death Index data were available for 88% (24/27) of the deaths. The median age at death was 39. Causes were: poisoning by any substance (38% of deaths); trauma (13%); cardiovascular disease (13%); exposure to cold (8%); alcohol abuse (8%); unknown (8%); diabetes (4%); lung neoplasm (4%); and intracerebral hemorrhage (4%). In unadjusted analyses, age, sex, smoking, homelessness, and addiction severity (Addiction Severity Index) were not associated with mortality. Significant predictors in unadjusted analyses remained so in adjusted analyses: drug of choice (heroin: adjusted Hazard Ratio [HR] 7.9 [95% CI 1.8–35]; alcohol: HR 5.0 [95% CI 1.1–23] as compared to cocaine); past suicide attempt (HR 3.2 [95% CI 1.5–7.0]); previous drug or alcohol overdose (HR 2.4 [95% CI 1.1–5.1]); history of any chronic medical illness (HR 2.3 [95% CI 1.0–5.2]); and Hispanic ethnicity (HR 5.2 [95% CI 1.7–16] as compared to Black).

CONCLUSION:  Mortality was substantial after substance detoxification in these relatively young adults without primary medical care. At the time of detoxification, heroin and alcohol as drugs of choice, overdose, past suicide attempt, medical comorbidity, and perhaps ethnicity, may help identify those at highest risk. These risk factors may help efforts to design and target interventions to prevent premature death in patients with addictions.

MOVING BEYOND ESTABLISHING COMPETENCE: WHY PROGRAMS SHOULD IMPLEMENT OBJECTIVE STRUCTURED CLINICAL EXAMINATIONS.S. Zabar1; A.L. Kalet1; K. Hanley1; D.L. Stevens1; E.K. Kachur2; M.D. Schwartz1; E. Pearlman1; M. Lipkin1. 1New York University, New York, NY; 2Medical Education Development, New York, NY. (Tracking ID #116577)

BACKGROUND:  All programs are struggling with how to incorporate performance based assessment of medical residents that are valid and worth the investment of time and resources. We evaluated the validity and educational benefit of our 3rd annual comprehensive, performance based assessment which has become an important source of individual and programatic formative feedback.

METHODS:  We developed a 10-station, faculty observed, objective structured clinical exam (OSCE) with eight standardized patients (SP) and 2 standardized student (SS) cases. Assessed competencies included: management of chronic and acute diseases, patient education, phone consultation, and teaching. In each station, trained faculty completed a 22 item behavioral checklist evaluating 5 domains: data gathering, rapport building, patient education, knowledge and synthesis and recorded global assessments (1–9 scale) of overall performance, communication skills, and fund of knowledge. Residents self-assessed, SPs and SSs evaluated residents’ overall performance, and faculty feedback was provided at each station. Convergent and construct validity were assessed using nationally normed in-service knowledge test exam (In-Service) which they took 6 months prior to the OSCE. Faculty and residents evaluated the OSCE's educational impact by a post-program survey.

RESULTS:  76 residents participated (8 PGY1, 60 PGY2, 8 PGY3). In-service scores correlated with faculty ratings of knowledge (r = .522, P = .01) but not with ratings of communication skills. SP ratings correlated with faculty ratings of communication skills (r = .747, P = .01) and overall performance (r = .762 P = .01). Resident self-assessment did not correlate with any faculty ratings. Fund of knowledge scores increased with each year of training (P = .001). 82% of faculty reported the OSCE helped them recognize residents’ strengths and weaknesses. 92% felt the OSCE provided valuable new information on resident performance. 60% of residents stated that the OSCE definitely stimulated them to learn more, and 62% felt it taught them something new. 70% of residents reported it provided them with valuable feedback and helped them identify personal strengths and weaknesses. Residents performed poorly on smoking cessation and telephone skills but felt they were appropriate cases of high educational value.

CONCLUSION:  A comprehensive performance based assessment reliably measured clinical skills, correlates with national inservice exam scores, helped residents identify strengths and weakness they are unlikely to be aware of, and led to important new curriculum and new information on residents performance.


BACKGROUND:  Academic medical centers must contend with seemingly competing expectations in their oversight of inpatient services. Hospital administrators demand decreased inpatient length of stay (LOS), while CMS and JCAHO expect documented improvements in quality of care. ACGME mandates instruction of residents in system-based care alongside traditional teaching topics. We sought to determine whether resident-focused multidisciplinary rounds (MDR) measurably addresses these issues. The specific goals were to achieve, without additional personnel, “best demonstrated practice” in JCAHO Core Measures metrics for adult medicine diagnoses, educate and involve residents in Core Measures performance as well as evidence and systems based practice, and to reduce LOS for CHF, community acquired pneumonia, and acute MI.

METHODS:  Patterned after the MDR pioneered at Berkshire Medical Center (Resp Care 2001; 46:1), we implemented thrice-weekly, hour long resident-centered and nurse-driven rounds in July 2003. MDR is led by the medicine faculty and attended also by dieticians, pharmacists, case managers, psychiatry, and physical therapy. Fifty to 70 cases are presented per session on a sequential team by team basis. Core Measures performance is tracked monthly along with LOS for CHF, pneumonia, and acute MI. An anonymous survey to ascertain attitudes and knowledge was administered to all residents participating in MDR.

RESULTS:  Five months after the implementation of MDR, average LOS decreased 0.88 days (P < .05) for pneumonia, 1.55 days (P = .06) for CHF, and 2.31 days (P < .05) for acute MI compared to 9 months prior. After three months, Core Measures performance improved for almost all measures, although statistical significance was reached only for pneumococcal vaccinations (52% vs. 10%, P = .0002). The resident questionairre revealed significant improvement in understanding of evidence-based care delivery as well as in overall quality of care, efficiency, and relationships with all disciplines. Knowledge of core measures increased in 100% of residents, while 67% improved from lowest scores of “none” or “some” knowledge to the top scores of “very good” or “excellent”. Residents felt that the benefits of MDR outweighed the costs of interruption of their work rounds, and they felt comfortable with the nurse-driven MDR presentation process.

CONCLUSION:  MDR led by physicians, driven by nurses, and focused on residents is associated with improvement in core measure performance, significant decreases in length of stay, and increased resident knowledge and competency in systems-based practice and evidence-based care.

N-ACETYLCYSTEINE FOR THE PREVENTION OF CONTRAST-INDUCED NEPHROPATHY: A SYSTEMATIC REVIEW AND META-ANALYSIS.R. Liu1; D. Nair1; J. Ix1; D.H. Moore1; S.W. Bent1. 1University of California, San Francisco, San Francisco, CA. (Tracking ID #115599)

BACKGROUND:  Contrast-induced nephropathy is a common cause of acute renal failure in hospitalized patients. Although patients are often given N-acetylcysteine to prevent renal injury from contrast agents, there are no clear guidelines supporting its use. We conducted a systematic review to determine if administering N-acetylcysteine around the time of contrast administration reduces the risk of contrast-induced nephropathy.

METHODS:  We searched MEDLINE, EMBASE, the Cochrane Collaboration Database, bibliographies of retrieved articles, abstracts of conference proceedings, and consulted with experts to identify relevant studies. Randomized controlled trials of N-acetylcysteine in hospitalized patients receiving contrast were included. Studies were excluded if they did not report change in creatinine or incidence of contrast-induced nephropathy at 48 hours.

RESULTS:  Nine randomized controlled trials satisfied all inclusion criteria and were included in the analysis. The difference in mean change in creatinine between the N-acetylcysteine treated group and controls was −0.27 mg/dl (95% CI, −0.43 to −0.11). The relative risk of developing contrast-induced nephropathy was 0.43 (95% CI, 0.24 to 0.75) in subjects randomized to N-acetylcysteine. Significant heterogeneity existed among studies, suggesting differences in patient populations or study methodology not identified by sensitivity analyses. The incidence of dialysis was rare (0.2%).

CONCLUSION:  Our findings suggest that N-acetylcysteine helps prevent declining renal function and contrast-induced nephropathy. While N-acetylcysteine is inexpensive and non-toxic, use of this agent may delay diagnostic and therapeutic procedures. Future studies are needed to address the long-term effects and cost-effectiveness of this agent.


BACKGROUND:  While antidepressant prescribing in primary care has increased significantly, it has been reported that acceptance of antidepressant medications may be poor among COPD patients. We sought to compare trends in antidepressant prescribing in office-based primary care visits where the patient had a diagnosis of COPD to visits where COPD was not a listed diagnosis.

METHODS:  We used National Ambulatory Medical Care Survey data from 1993 to 2001 to examine antidepresssant prescribing at office-based primary care visits. We defined the presence of COPD in visits by an ICD-9-CM code for COPD (490–496). Age, sex, race/ethnicity, and insurance status were also available for analysis. We used SUDAAN software to perform chi-squared tests and multivariable logistic regressions to compare visits with COPD versus no COPD listed.

RESULTS:  There were ~2.7 billion primary care visits between 1993 and 2001; 7.4% listed COPD as a diagnosis. Overall, those with COPD were more commonly over 65 years old (33.7% versus 26.3%), male (43.4% versus 41.4%), and white (81.7% versus 78.9%), less commonly privately insured (57.4% versus 62.3%). Antidepressants were prescribed less commonly (see figure). The overall adjusted odds of antidepressant prescribing increased by 10% per year (OR 1.10, 95% CI 1.07–1.13). However, the adjusted odds of antidepressant prescribing for visits with COPD were 0.71 (95% CI 0.59–0.84); there was no interaction between COPD and year.

CONCLUSION:  While antidepressant prescribing in primary care has increased overall and among visits with COPD, there were persistently lower odds of antidepressant prescribing to those with COPD. Future research is needed to examine the cause of this disparity.

inline image

NIGHT FLOAT—AN INTERN's EXPERIENCE.J.M. Mercado1; A. Tulsky1. 1University of Pittsburgh, Pittsburgh, PA. (Tracking ID #117143)

BACKGROUND:  As a consequence of the ACGME mandated work hour regulations, many residency training programs are instituting a Night Float (NF) service to ensure compliance with the number of hours a medical trainee spends on call. Having little previous experience with NF, we performed a descriptive study of the cross coverage problems presented to the NF intern during a 20-day period to 1) identify the spectrum and volume of cross coverage experienced on a NF service, and 2) identify which cross coverage problems can be anticipated in advance by the primary team, and thus improve patient care and resource utilization.

METHODS:  All cross-coverage calls received by the intern during a 20-day period were tallied and categorized as urgent (must be seen right away) or non-urgent (handled by telephone); and as anticipated (noted in the resident's sign-out) versus unanticipated (not mentioned in the resident's sign-out) by the primary team. Calls were further categorized by the following system-related problems and complaints: cardiovascular, respiratory, gastrointestinal, endocrine, metabolic, pain, psychiatric, falls and miscellaneous.

RESULTS:  The total number of patients cross-covered during this period was 887 with an average of 44.35 patients per night (range 31 to 55). Two hundred and forty five calls were received (average 12.2 per night, range 6 to 24). Seventy eight percent of the calls were urgent and 67% were unanticipated. The most common calls were for the management of: 1) pain, 36/245 calls (14.7%); 2) blood pressure (BP)instability, defined as systolic BP >170mmHg or <90mmHg, 32/245 calls (13.06%); and 3) insulin regimen, 30/245 (12.24%).

CONCLUSION:  During this 20-day study of NF intern cross coverage, 78% of calls were due to urgent complaints, and 67% due to unanticipated problems. Pain management was the most common reason of all calls. Many of these could have been anticipated with an adequate pain management protocol. The second most common call was related to blood pressure instability, a situation which often cannot be anticipated by the day team. However, a structured teaching experience that reviews each case may identify opportunities to prevent future events, and fulfills one of the new ABIM competencies on practice based learning. The third most frequent problem encountered involved blood glucose management. These occurrences could be reduced with the institution of a standard protocol for Insulin administration. A repeat survey of cross-coverage calls would be useful to generalize these results beyond the study period. Nevertheless, we felt sufficiently confident to present these findings to the Hospital's Quality Management Committee. Consequently, a pain and a diabetes management protocol are currently in development.

NO RELATIONSHIP BETWEEN LITERACY STATUS AND MEDICAL CHARGES.R.M. Malone1; D.A. DeWalt1; M.P. Pignone1; R. Rothman2. 1University of North Carolina at Chapel Hill, Chapel Hill, NC; 2Center for Health Services Research, Nashville, TN. (Tracking ID #115923)

BACKGROUND:  Low literacy has been associated with worse health outcomes, and advocates often assume that low literacy is also associated with higher health care costs. However, few studies have examined the relationship between literacy status and healthcare charges. We performed a retrospective cohort study to examine the role of literacy on health care charges.

METHODS:  We identified patients who participated in diabetes, anticoagulation, heart failure and pain management programs at a university general internal medicine clinic, who previously had literacy measured. Literacy was assessed with the Rapid Estimate of Adult Literacy in Medicine (REALM). All inpatient and outpatient charges, including physician, emergency department, laboratory and inpatient pharmacy charges were obtained for each patient from July 2002 through June 2003. Outpatient pharmacy, hospice, and home health charges were not included. Total charges were compared using multivariate linear regression with cost as a log transformed dependent variable. We also stratified by disease management program.

RESULTS:  Literacy status was measured in 593 out of 1,640 patients enrolled in our disease management programs. Of those, 544 were actively followed during the year of analysis. Mean age was 59 (range 19–97), 54% were female, and 49% were African American. There was a high prevalence of low literacy with 20%≤3rd grade, 14% 4th–6th grade, 26% 7th–8th grade, and 40%≥9th grade. Median charges were: ≤3rd: $3,515, 4th–6th: $3,041, 7th–8th: $3,013, and ≥9th: $3,268. After controlling for age, sex, race and disease management program, we found no difference in charges when stratified by literacy status.

CONCLUSION:  Among patients in chronic disease management programs, literacy status was not associated with higher medical charges. Further research is needed to explore the role of literacy on health care costs.


BACKGROUND:  Active control equivalency trials (ACETs) are used increasingly to meet FDA requirements for evidence of new drug effectiveness. Unlike conventional superiority trials, intention-to-treat (ITT) analyses of ACETs are non-conservative in the presence of non-adherence therefore ‘as-treated’ analyses are often recommended. As-treated analyses, however, are subject to confounding unless adherence is independent of the outcome. We sought to quantify the degree of bias introduced in ACETs by non-adherence with and without modest confounding between adherence and outcome.

METHODS:  We performed analyses based on results of a recently reported ACET that randomized 4,909 patients to valsartan and 4,909 to captopril following myocardial infarction with the primary outcome of all-cause mortality (NEJM 2003;349:1893–906). Risk ratios (RR) were calculated in ITT and as-treated analyses with various proportions of adherence, which was assumed to be equal in both treatment groups and constant throughout the ACET. RR were recalculated after the introduction of modest confounding of the association of adherence and mortality. Two-sided 95% CI were calculated to correspond to the one-sided alpha of 0.025 specified in the ACET.

RESULTS:  The reported results of the ACET strongly rejected the null hypothesis of inferiority of valsartan compared with captopril in ITT analyses (P = .004). The corresponding upper bound of the CI (RR 1.02; 95% CI: 0.94–1.10) was well within the pre-specified non-inferiority margin of 1.13. The reported result, however, assumed 100% adherence, which was not measured in the ACET. To achieve the reported result with 50% adherence in both groups, the true (unbiased) effect would have had to be RR 1.05 (95% CI: 0.97–1.13), consistent with inferiority of valsartan. In the absence of confounding, as-treated analyses provided a valid estimate of this effect and a conservative CI (RR 1.05; 95% CI: 0.92–1.20 for the same degree of adherence). In the presence of modest confounding (10%), however, as-treated analyses yielded a biased effect estimate and a biased CI that falsely suggested equivalence (RR 0.99; 95% CI: 0.88–1.11 for the same degree of adherence).

CONCLUSION:  If adherence is imperfect, ITT analyses of ACETs yield effect estimates and CI that are usually biased toward equivalence. As-treated analyses yield valid effect estimates and conservative CI only if confounding between adherence and outcome is absent. If such confounding is present, as-treated analyses may also falsely suggest equivalence. ACETs reliably provide valid results only in settings in which adherence is guaranteed. In the presence of non-adherence, ACETs become analogous to observational designs and require an assumption of no confounding between adherence and outcome.

NON-ADHERENCE TO EVIDENCE-BASED GUIDELINES IN THE INITIAL TREATMENT OF HYPERTENSION.M.A. Fischer1; R. Levin1; J. Avorn1. 1Brigham and Women's Hospital, Boston, MA. (Tracking ID #116478)

BACKGROUND:  Evidence-based guidelines for the management of hypertension have emphasized the importance of older medications, such as thiazide diuretics, as first-line therapy. Nevertheless, a large percentage of prescribing for hypertension falls outside of published guidelines. We examined trends in new prescriptions for hypertension in a large population of typical elderly patients, and the association between patient factors and the choice of antihypertensive medication.

METHODS:  Using filled prescription data from a large state drug assistance program for the elderly, we identified all patients who began treatment for hypertension between 1997 and 2000. We used Medicare claims data for these patients to identify other medical conditions that might influence choice of antihypertensive medication. Each prescription was classified by drug class and by whether it adhered to evidence-based guidelines as published at the time (JNC-VI guidelines). We then used multivariate regression models to examine the effect of patient factors on choice of medication.

RESULTS:  During the 4 year study period a total of 14,463 patients began antihypertensive medications. The most commonly prescribed drug class was angiotensin-converting enzyme inhibitors, accounting for 25% of new starts. Over the four years studied, the rate of new starts of calcium channel blockers declined while new starts of beta blockers increased. Among patients without major comorbidities, 22% were prescribed thiazide diuretics as their initial antihypertensive medication; this rate did not change over time. Use of beta blockers for patients with a history of myocardial infarction increased from 26% in 1997 to just 39% in 2000. Overall, only 50% of patients received a prescription that appeared to conform to JNC-VI guidelines; there was minimal increase in evidence-based prescribing over time (50% to 51%). Multivariate modeling showed that female patients and non-white patients were more likely to receive medications compatible with JNC-VI guidelines. Patients with cardiac disease, such as history of myocardial infarction or congestive heart failure, were more likely to receive a prescription that conformed to the guidelines, while patients with diabetes were less likely to do so.

CONCLUSION:  Choice of initial medication for hypertension therapy remains heterogeneous and differs from published recommendations in half of patients. Guideline adherence did not change substantially over time, although evidence-based prescribing did increase in patients with cardiac disease.

NON-HDL VERSUS LDL CHOLESTEROL AS A RISK FACTOR FOR FIRST MYOCARDIAL INFARCTION.W.R. Farwell1; H.D. Sesso2; J.E. Buring2; J.M. Gaziano1. 1Harvard University, Boston, MA; 2Brigham and Women's Hospital, Boston, MA. (Tracking ID #116863)

BACKGROUND:  Low-density lipoprotein (LDL) is the primary cholesterol parameter targeted to prevent myocardial infarction (MI). Non-high-density lipoprotein (non-HDL) includes LDL as well as other atherogenic particles and does not require a fasting sample. We examined whether non-HDL was a better risk factor for first MI compared with LDL.

METHODS:  We analyzed data from the Boston Area Health Study, a case-control study from the 1980s investigating risk factors for first MI. Cases were identified as white men or women, less than 76 years old, living in the Boston area, with no previous history of MI or angina pectoris, in whom symptoms of MI began during the 24 hours before admission. The diagnosis of MI was based upon clinical history and elevated creatinine kinase. For each case, a control with no previous history of MI or angina was selected and matched for age, sex, hospital, and home neighborhood. Fasting blood samples were collected and analyzed for lipids, and a comprehensive array of dietary and lifestyle variables were collected. We performed unmatched multivariable logistic regression to compare the independent association of non-HDL and LDL to first MI. Data from 303 cases and 297 controls who had complete fasting lipid profiles were analyzed.

RESULTS:  Cases had a greater prevalence of hypertension, diabetes mellitus, family history of MI and history of smoking than controls. Quartiles of non-HDL were defined among the controls as ≤143, 144–172, 173–200, and ≥201 mg/dL; LDL quartiles were defined as ≤111, 112–137, 138–159, and ≥160 mg/dL. After multivariate adjustment for other cardiovascular risk factors in unmatched analyses, the odds of an MI were greater in the second, third and fourth quartiles of non-HDL than LDL compared to the first quartile of each. For non-HDL cholesterol, the ORs in the second through fourth quartiles were 1.83 (1.07–3.14), 2.07 (1.23–3.49), and 2.33 (1.39–3.90) (p trend = 0.0027); for LDL cholesterol the ORs were 1.10 (0.67–1.81), 0.87 (0.52–1.46), and 1.45 (0.90–2.35) (p trend = 0.15).

CONCLUSION:  Given that non-HDL accounts for LDL plus other atherogenic particles and does not require a fasting sample, our study suggests that non-HDL cholesterol is as useful as LDL to initially screen patients for risk of first MI.

NOTICES OF PRIVACY PRACTICES: A SURVEY OF THE HIPAA DOCUMENTS PRESENTED TO PATIENTS AT U.S. HOSPITALS.M.K. Paasche-Orlow1; J.N. Powell2; D.M. Jacob3. 1Boston University, Boston, MA; 2Harvard University, Cambridge, MA; 3Healthcare Analytics, LLC, New York, NY. (Tracking ID #116993)

BACKGROUND:  Under the Health Insurance Portability and Accountability Act (HIPAA) privacy rule, hospitals must present patients with a Notice of Privacy Practices (NPP) so people can control their health information. Regulations stipulate the content of NPPs, mandate that they are written in language that is readable by patients, and posted on institutional Web sites. While no specific grade-level has been federally endorsed, the average reading level of American adults is 9th grade. We hypothesized that NPP's would meet content requirements, but would not be readable by patients and that readability would be influenced by local literacy rates, the level of research activity at the hospital, and hospital size. Similarly, we hypothesized that hospitals with >5% local rate of low English proficiency (LEP) would be more likely to present easy-to-read NPP's and NPP texts in alternative languages.

METHODS:  To test these hypotheses, we conducted a cross-sectional study linking data from several public-use sources. A total of 115 Web sites of Hospitals selected from the U.S. News and World Report list of top Hospitals in America 2002 were surveyed for NPP texts. All English NPP texts were evaluated for 19 content items and readability using the Flesch-Kincaid scale, which assigns the minimal grade-level required to read a text (range 0–16). Data on local literacy rates, local rates of LEP, the level of research activity, and hospital size were obtained from organizational Web sites.

RESULTS:  We were able to identify NPP's for 91% (105/115) of the selected hospitals. Spanish NPP text was available at 25 of hospital Web sites; seven of these hospitals presented NPP texts in additional languages. All surveyed content items were found in 75% (79/105) of hospital's NPPs. The remaining 25% (26/105) of hospital's NPPs had 1–4 missing items. The most frequent item absent was a statement regarding fundraising. The average grade-level readability of NPP text was 12.4 (95% CI 12.0 to 12.7). Readability was not associated with the rate of local literacy (P = .07), the level of research funding (P = .82), or hospital size (P = .49). Hospitals with >5% local LEP had NPPs that were more difficult to read (P = .005) and were not more likely to have foreign language NPP texts available on their Web sites (P = .30).

CONCLUSION:  While NPPs presented to patients in U.S. hospitals typically cover the content stipulated by regulation, they are written beyond the reading capacity of the majority of American adults. The goals of the HIPAA Privacy Rule cannot be met with NPPs patients cannot decipher.


BACKGROUND:  Number needed to treat (NNT) has been described as an essential paradigm in understanding the clinical significance of a new therapy. However the ability of patients and physicians not trained in EBM, to understand NNT remain unclear from most studies. The purpose of the study was to determine how often patients and physicians understand the concept of NNT. A secondary aim was to determine the limitations of NNT.

METHODS:  Relevant articles were identified by searching various database including MedLine (1980–2003), Embase (1988–2003), PsychInfo(1984–2003), Web of Science (1993–2003), educational websites, and bibliography of relevant articles. Study design, quality of the study, and limitations of the study were abstracted by two independent reviewers. Review articles on therapy, and studies on efficacy of workshops on EBM were excluded.

RESULTS:  Ten articles met the inclusion criteria. Four articles involving patients (1) and physicians (3) revealed that both the groups preferred results when presented as RRR as compared to NNT. Two studies on the physician's ability to understand and explain NNT to others revealed that only 16% in Australian physicians to 35% in a study on UK physicians respectively understood and felt confident in explaining NNT to others. Another study involving medical students revealed impaired understanding of NNT as compared to RRR (25% vs. 75%). A patient only study revealed that 80% of patients consented for therapy regardless of the size of NNT, i.e., 10 or 400. Three studies explored the limitations of NNT; which included difficulty in interpretation when comparing the effects of interventions over different periods of time, or applied in different population, using NNT for preventive interventions, equating NNT with winning a lottery, age of the physician as a determinant of using NNT to guide therapy (younger > older to be affected by NNT), consideration of wasted effort related with any treatment (i.e., [NNT-1] patients who have no benefit).

CONCLUSION:  Despite numerous studies revealing the efficacy of teaching EBM in workshops the overall understanding of NNT among patients and medical personnel is limited. Ongoing effort is indicated in educating both the patients and physicians about terms commonly used in EBM. Limitations of NNT should be stressed to physicians. Drawbacks of our study included; 1) limited number of studies on NNT, 2) variable study design and quality of educational research.