To compare health care utilization in people with systemic lupus erythematosus (SLE) in health maintenance organizations (HMOs) and fee-for-service (FFS).
To compare health care utilization in people with systemic lupus erythematosus (SLE) in health maintenance organizations (HMOs) and fee-for-service (FFS).
A structured survey was administered to a cohort of 982 people with SLE who were assembled between 2002 and early 2005. A total of 2,656 person-years of observation were completed by the end of 2005. In each year, respondents reported their health care utilization and whether they had HMO or FFS coverage. We compared health care utilization of those in HMOs and FFS, with and without adjustment for socioeconomic, demographic, and health characteristics using repeated-measures regression techniques.
Compared with people with SLE who were in FFS, those in HMOs were younger (3.3 years), received a diagnosis at an earlier age (3.6 years), had slightly less disease activity (0.4 on a 10-point scale), were more likely to be nonwhite (8.8%), were less likely to be below the poverty line (7.8%), and were less likely to have public insurance (29.7%). The 2 groups did not differ in other characteristics. On an unadjusted basis, subjects with SLE in HMOs had significantly fewer physician visits (3.1; 95% confidence interval [95% CI] 1.7, 4.5) and were less likely to report one or more outpatient surgical visits (6.3%; 95% CI 2.5, 10.0), and hospital admissions (5.5%; 95% CI 1.7, 9.3) than those in FFS. Adjustment reduced the differences in physician visits (2.3; 95% CI 1.1, 3.5), outpatient surgical rates (4.4%; 95% CI 0.6, 8.1), and hospital admission rates (4.0%, 95% CI 0.4, 7.7).
Subjects with SLE in HMOs utilized substantially less ambulatory care and were less likely to have outpatient surgery and hospital admissions than those in FFS; the effects were not completely explained by socioeconomic, demographic, and health characteristics.
The expansion of health maintenance organizations (HMOs) has raised concerns that people with chronic diseases will not receive the health care services needed to monitor and treat their conditions. Most studies examining the impact of HMOs on utilization have concerned relatively healthy, employed populations (1–8). In these studies, HMOs have been found to lower hospital utilization by ∼25%, but have little effect on ambulatory care utilization (8). However, results of studies of people with a range of discrete chronic diseases have not demonstrated the same lower utilization rate of expensive services such as hospital admissions or surgery (9).
Research in this area for the rheumatic diseases has largely focused on subjects with rheumatoid arthritis (RA). In these studies, there was no difference in hospital utilization, and inpatient and outpatient surgery (historically the most expensive services for patients with RA) between those with HMOs and those with fee-for-service (FFS) (10–14). However, 1 recent study did show that the use of biologic response modifiers was much lower in patients with HMOs than those with FFS with and without other forms of utilization control, primarily because rates of initiation of such agents were lower in HMOs (15).
The existing literature examining the impact of HMOs or other forms of managed care on health care utilization in people with systemic lupus erythematosus (SLE) is much sparser. In the only published study found, Stewart and Petri reported that the patients of a single physician with HMO coverage did not differ from the same physician's patients who had FFS in terms of treatment or outcomes of lupus nephritis (16). However, the Stewart and Petri study had a small sample size and involved a tertiary care provider. In contrast, our study was designed to examine differences in health care utilization between those in HMOs and those who are in FFS in a large sample of people with SLE from a greater range of health care environments and disease severity; most were not sampled from providers, let alone tertiary care providers.
The analysis used data from the first 3 waves of the University of California at San Francisco (UCSF) Lupus Outcomes Study (LOS), an ongoing longitudinal survey of individuals with SLE who had previously participated in a study of SLE genetic risk factors. In the genetics study, SLE diagnoses were verified according to the American College of Rheumatology criteria (17, 18) through medical records abstraction. Participants became eligible for the LOS upon completion of their enrollment in the genetics study. The initial LOS enrollment period was September 2002 through April 2005. During that period we attempted to enroll 1,519 individuals, of whom 1,265 (83%) were successfully contacted, and 982 completed at least 1 interview (65% of all eligible and 78% of those successfully contacted). The LOS participants were from 40 states, and represented over 50 HMOs. In 2004, a second annual interview was conducted with 832 (92%) of the 900 LOS participants who had enrolled prior to the end of 2003. In 2005, an additional followup interview was conducted, successfully interviewing 842 participants (92% of those eligible for the interview), of whom 765 were completing their third interview. Of the 140 persons not re-interviewed after either a first or second interview, 20 had died, 22 were medically unable to be interviewed, 63 declined further participation, and 35 were lost to followup. Followup rates did not differ between LOS participants in HMOs and FFS.
The original genetics study recruited participants from several clinical and community-based sources, including UCSF-affiliated clinics, non-UCSF rheumatology offices, lupus support groups and conferences, and newsletters, Web sites, and other forms of publicity. Ultimately, two-thirds of the LOS participants were derived from non-clinical sources.
The principal source of data for the LOS was a structured 1-hour telephone survey conducted by trained interviewers. The survey included well-validated items covering the following domains: demographics and socioeconomic status, SLE status, disability, general health status and social functioning, employment status, psychological and cognitive status, health care utilization, medications, and health insurance coverage.
The health care utilization section of the questionnaire asked participants about their medical care over the previous 12 months. It included an enumeration of all health care practitioner visits by specialty. The section also included information about emergency department use, hospitalizations, outpatient surgery, and diagnostic procedures. The medication section included an extensive list of prescription drugs for SLE and for other conditions. Participants were asked to report on current and prior use of these medications. The health insurance section was derived from the Medical Expenditures Panel Survey (MEPS) (19, 20) and included items regarding the type of coverage (HMO versus FFS) and source of coverage (employment-based, individually purchased plan, or public program), as well as the specific aspects of coverage such as extent of copayments and deductibles, mode of access to specialty services, and nature of drug coverage. LOS participants reported HMO membership reliably across multiple interview waves; this was consistent with a validation study conducted by the Agency for Health Care Research and Quality prior to the initiation of the MEPS (20). The enrollment and data collection protocol was approved by the UCSF Committee on Human Research.
The primary independent variable for this analysis was coverage in an HMO versus an FFS plan, based on the participants' explicit report of being in an HMO. A secondary independent variable was the source of insurance, categorized as employer-based (including the participant's or a family member's current or former employer), government-based (primarily Medicare or Medicaid), or privately purchased (by the participant or a family member).
In addition, we considered 3 other types of covariates: socioeconomic and demographic characteristics, general health status variables, and SLE-specific variables. The socioeconomic and demographic characteristics included age in years at interview, sex, race/ethnicity (Hispanic, African American, Asian, or other, with white non-Hispanic as referent), marital status (never married or widowed/separated/divorced, with married as referent), education (less than high school, high school graduate, some college, or college graduate, with postgraduate degree as referent), and poverty status (above or below 125% of the federal poverty threshold).
General health status variables included a measure of health-related quality of life (12-Item Short Form Health Survey [SF-12] physical and mental component scores) (21), body mass index (BMI), smoking status (current or former smoker, with never smoked as referent), and comorbid conditions (heart disease, cancer, pulmonary disease, or diabetes). Although some of these conditions may actually be SLE manifestations, they were included in the analyses because they were likely to contribute to the overall health status of the individual and to the need for medical care. Comorbidity was classified as 1 condition or >1 condition, with 0 conditions as the referent.
SLE-specific variables included age in years at diagnosis, self-reported SLE flare in the 3 months prior to interview, self-reported SLE activity level in the preceding 3 months (on a 10-point scale, where 0 = no activity and 10 = highest level of activity) (22), and self-reported kidney, lung, or vascular manifestations over the past year or the past 2–5 years. Evidence of kidney manifestations included transplant, kidney biopsy, or initiating dialysis. Evidence of pulmonary manifestations included lung biopsy, bronchoscopy, or hemoptysis. Evidence of vascular manifestations included myocardial infarction, stroke, transient ischemic attack, deep venous thrombosis, pulmonary embolism, or presence of other blood clots. At the first interview, participants reported the year in which each event or procedure occurred; in subsequent interviews, they reported events or procedures that occurred since the previous interview. From their responses, we determined whether there had been any relevant manifestations in the year preceding the interview or in the 2–5 year period preceding the interview. However, because of the ongoing need for followup care, individuals with a kidney transplant were considered to have had a kidney manifestation in the previous year, no matter when the transplant was actually performed.
The dependent variables for the present analysis were measures of ambulatory and inpatient health care services, diagnostic tests, and medications. Total physician visits were calculated as the sum of visits to all generalist and specialist physicians during the 12-month period preceding the interview. Because of the complex nature of SLE and the difficulty distinguishing between complications due to the disease and unrelated comorbidity, we did not separate health care utilization attributed to SLE from all other health care. The total number of visits to all physicians, nonphysician health care providers, rheumatologists, and generalists (family practitioners, internists, and general practitioners) were treated as continuous measures, because a large proportion of the cohort had at least 1 such visit. To address the skewed distribution of these variables, we truncated the values for each variable at the third standard deviation above the mean, resetting the upper limit of the variable, rather than dropping anyone from the analysis. An alternative to this method, a log transformation, was also considered, but because the results did not materially differ from the results using the truncated values, we used the latter because doing so permitted estimation of the confidence intervals for the difference in utilization between those in HMOs and FFS.
For visits to less common, but important specialists involved in the care of SLE patients (including nephrologists, pulmonologists, dermatologists, and physical therapists), we dichotomized the variables as any visits versus no visit. We treated emergency department visits, outpatient surgeries, and hospitalizations in the same manner, and also calculated the average length of stay among individuals with ≥1 night in the hospital. Procedures evaluated included magnetic resonance imaging and computed tomography scans, pulmonary function testing, and bone density scans. Medications were analyzed by category, including nonsteroidal antiinflammatory agents, cytotoxic agents, immunosuppressive agents (with mycophenolate mofetil examined as a special case because of its higher cost), antimalarial agents, oral steroids, pulse steroids, antihypertensive agents, and antidepressive agents. All categories of medications were treated as dichotomous variables, distinguishing between those who were taking the type of medication at the time of the interview and those who were not.
Each time participants were interviewed, they contributed 1 observation to the analysis. Therefore, 765 participants who completed 3 interviews contributed 3 observations each (representing 85% of all those eligible for 3 waves), 144 participants contributed 2 observations each, and 73 participants with no followup interviews contributed 1 observation each, for a total of 2,656 person-years of observation. Fifty-one observations were excluded because participants had no health insurance coverage, 80 observations were excluded because the participant could not identify if their health plan was an HMO, and an additional 122 were excluded because of missing values for other key variables, leaving 2,403 observations (90% of the total) for the analysis.
Re-interview rates were significantly higher in women than in men (93% versus 83%), in whites than in nonwhites (94% versus 88%), in those with higher rates of formal education (88% among those who had a high school degree or less, 93% among those with some college, 92% among those who had completed college, and 94% among those with post-baccalaureate training), but there was no difference in followup rates according to poverty status or age. Participants who reported end-stage renal disease and pulmonary manifestations and those with longer durations of disease were significantly less likely to be re-interviewed. There was no difference in followup rates by the following health measures: report of fair or poor health versus any other status, recent history of weight loss, fatigue, fevers, muscle pain or weakness, joint stiffness or swelling, history of clotting disorders, vision loss, or seizures.
Because of the potential impact of differential rates of followup, we developed attrition weights, using a logistic regression model that estimated the probability that an individual had been re-interviewed, based on characteristics ascertained in the baseline interview. The inverse of this probability was used as an attrition weight so that individuals with a lower probability of completing a second interview had larger weights. The resulting proportional attrition weights ranged from 0.95–1.6, with a mean (necessarily) of 1.0. For all analyses in which the person-year was the unit of analyses, we included this calculated attrition weight for wave 2 and wave 3 observations, and set the attrition weight to 1.0 for wave 1 observations.
The 982 participants in the study derive from 955 distinct families (∼1.03 per family stratum). This small nesting effect reduced the effective sample size to ∼2,375 observations (from 2,403). Because the change was so small, we did not incorporate this design effect into the analyses described below. However, we used SUDAAN software (23) to take into account the correlation among the multiple observations contributed by individuals and the attrition weight described above.
We first described the socioeconomic, demographic, and health characteristics of the sample, and used those characteristics to compare individuals who reported that their principal health insurance coverage was an HMO with those who reported other types of health insurance, using t-tests for continuous variables (e.g., age, BMI, or SF-12 scores) and chi-square tests for categorical or dichotomous variables (e.g., race/ethnicity, smoking status, or sex). We next estimated the mean of the more common utilization measures (such as the total number of physician visits) and the proportion with any of the less common services (such as those with ≥1 hospital admission) for participants in HMOs versus FFS plans, on an unadjusted basis, and adjusted for the socioeconomic, demographic, and health characteristics listed above, with 2 modifications. The first modification occurred after we determined that SLE disease activity level and recent SLE flare were too highly correlated to be included in the same model (r = 0.58); therefore we dropped the latter variable from the models. Second, we combined the 3 types of SLE manifestations (kidney, pulmonary, and vascular) into a single variable with the following 3 levels: any manifestation in the year prior to interview, any manifestation in the 2–5 year period prior to interview, and no manifestations in the past 5 years. We found this combined variable to be equally predictive of utilization as the 3 separate variables.
For continuous measures, we estimated the least squares means from a linear regression model, calculating the difference and 95% confidence interval (95% CI) between those in HMOs and FFS. For dichotomous measures, including ambulatory and nonambulatory care, medications, and diagnostic procedures, we estimated the adjusted rates from logistic regression models. The adjusted rates were calculated from the predicted probability, also known as the predicted marginal, from which we also calculated the difference (and 95% CI) between those with HMO and FFS plans. To estimate the differential impact of sets of independent variables on the relationship between HMO status and health care utilization, we showed a series of models for 2 important markers of utilization—total physician visits and proportion with hospitalizations. The models were built sequentially beginning with the bivariate model (HMO versus FFS only) and adding first health-related variables, then socioeconomic variables, then source of insurance, and finally, an interaction term for source of insurance by HMO status. In addition to estimating the adjusted mean physician visits and adjusted hospitalization rate for each model, we also compared the predictive power of each model with the immediately preceding one, by calculating the difference in the model R2 or model chi-square for the linear and logistic models, respectively. In the final model, we compared the adjusted mean physician visits and adjusted hospitalization rate for HMO versus FFS for each of the 3 categories of source of insurance.
When comparing the sociodemographic and baseline health characteristics, the subjects with SLE that had HMOs were significantly, although marginally, younger and less likely to be non-Hispanic whites than those with FFS (Table 1). The HMO group was also less likely to have a household income of <125% of the federal poverty threshold, and to receive their insurance from Medicare and/or Medicaid. The HMO and FFS groups did not differ in terms of sex, marital status, or education. The people with SLE in HMOs and FFS were remarkably similar in their baseline health characteristics, differing significantly only in the age at diagnosis (those in HMOs were 3.6 years younger at diagnosis) and the reported current level of disease activity (4.1 versus 4.5 on a scale of 0–10) (Table 2). Of note, the 2 groups did not differ significantly in SF-12 physical or mental component scores, BMI, smoking status, comorbidity, disease duration, the proportion reporting a flare in the 3 months prior to survey, or reporting kidney manifestations, pulmonary manifestations, or vascular events in the 1 year or 5 years prior to interview.
|Characteristic||Total cohort (n = 887)||HMO (n = 303)||FFS (n = 584)||P†|
|Age, mean ± SD years||47.2 ± 3.1||45.0 ± 12.6||48.3 ± 13.3||< 0.001|
|Widowed, separated, divorced||17.4||13.2||19.5|
|Less than high school||3.5||3.0||3.8|
|High school graduate||16.9||14.5||18.2|
|Some college or trade school||41.7||44.6||40.2|
|Income <125% of poverty threshold||11.7||6.6||14.4||< 0.001|
|Principal insurance source||< 0.001|
|Employer (self or family member)||61.0||78.2||52.1|
|Medicare and/or Medicaid||31.5||11.9||41.6|
|Characteristic||Total cohort (n = 887)||HMO (n = 303)||FFS (n = 584)||P†|
|Overall health status|
|SF-12 physical component||39.6 ± 6.5||40.1 ± 6.5||39.3 ± 6.5||0.10|
|SF-12 mental component||42.0 ± 6.3||42.5 ± 6.2||41.8 ± 6.4||0.15|
|Body mass index||27.0 ± 6.7||26.7 ± 6.7||27.2 ± 6.8||0.31|
|Smoking status, %||0.19|
|Comorbid conditions, %||0.13|
|SLE status and history|
|Age at diagnosis, years||34.3 ± 13.3||31.9 ± 12.5||35.5 ± 13.6||< 0.001|
|Disease duration, years||12.9 ± 8.6||13.1 ± 8.8||12.8 ± 8.5||0.65|
|Flare in past 3 months, %||47.1||45.5||47.9||0.50|
|Level of SLE activity||4.4 ± 3.1||4.1 ± 3.2||4.5 ± 3.1||0.04|
|Past 2–5 years||5.4||6.3||5.0||0.42|
|Past 2–5 years||7.3||5.6||8.2||0.16|
|Past 2–5 years||8.6||7.9||8.9||0.62|
On an unadjusted basis, compared with people in FFS, those in HMOs had significantly fewer total ambulatory visits to all physicians, generalist physicians, and nonphysician providers (Table 3). The HMO group was also significantly and substantially less likely to report 1 or more visits to pulmonologists (6.8% versus 11.1%), dermatologists (27.3% versus 35.6%), and physical therapists (16.8% versus 24.6%), and to have 1 or more outpatient surgical procedures (17.4% versus 23.7%) or hospital admission (18.4% versus 23.9%). Adjustment reduced the differences in utilization between subjects in HMOs and FFS somewhat, so that the difference in the number of generalist physician visits and in the proportion visiting a pulmonologist was no longer significant. Nevertheless, the overall pattern of lower utilization rates among those in HMOs remained.
|Type of health care service||Unadjusted||Adjusted|
|Model n†||All||HMO||FFS||Difference (95% CI)||HMO||FFS||Difference (95% CI)|
|Mean number of visits|
|Any MD||2,403||17.2||15.2||18.3||3.1 (1.7, 4.5)‡||15.7||18.1||2.3 (1.1, 3.5)‡|
|Rheumatologist||2,401||3.6||3.5||3.7||0.1 (−0.3, 0.5)||3.5||3.7||0.1 (−0.2, 0.5)|
|Generalist||2,386||4.1||3.6||4.4||0.8 (0.3, 1.2)‡||4.0||4.2||0.3 (−0.2, 0.7)|
|Non-MD providers||2,403||8.5||6.0||9.7||3.7 (2.0, 5.4)‡||6.4||9.5||3.1 (1.4, 4.7)‡|
|% with one or more visits to|
|Nephrologist||2,403||17.7||18.4||17.7||−0.7 (−5.1, 3.7)||16.4||18.8||2.4 (−1.2, 6.0)|
|Pulmonologist||2,403||9.7||6.8||11.1||4.3 (1.1, 7.6)‡||7.8||10.5||2.7 (−0.6, 5.9)|
|Dermatologist||2,403||32.8||27.3||35.6||8.3 (3.5, 13.1)‡||27.4||35.5||8.0 (3.3, 12.8)‡|
|Physical therapist||2,403||21.9||16.8||24.6||7.9 (3.9, 11.8)‡||17.9||23.9||6.0 (2.0, 9.9)‡|
|Any emergency department visit, %||2,372||40.6||37.8||42.3||4.5 (−0.4, 9.3)||39.4||41.5||2.0 (−2.6, 6.6)|
|Any outpatient surgical procedures, %||2,403||21.6||17.4||23.7||6.3 (2.5, 10.0)‡||18.6||22.9||4.4 (0.6, 8.1)‡|
|Any, %||2,403||21.9||18.4||23.9||5.5 (1.7, 9.3)‡||19.4||23.4||4.0 (0.4, 7.7)‡|
|Average length of stay among admissions||520||4.9||4.9||4.9||−0.1 (−1.3, 1.2)||4.9||4.9||0.0 (−1.2, 1.2)|
Table 4 presents the extent to which health and socioeconomic characteristics, the source of insurance, and the interaction between source of insurance and HMO status mediate the relationship between HMO status and physician visits and hospital admissions. With respect to physician visits, all sets of variables contribute to explaining the variation in the number of visits. Clearly, health and sociodemographic characteristics and the source of insurance account for some of the difference between those in HMOs and FFS in medical care visits, reducing the difference from 3.1 to 1.5 visits per year. With respect to the proportion having 1 or more hospital admissions, health characteristics and the source of insurance account for some of the difference in admission rates between subjects with SLE in HMOs and FFS, but sociodemographic characteristics do not add significantly to the explanation of differences in admission rates between the 2 groups. All sets of variables combined, however, reduce the difference in admission rates between those in HMOs and FFS from 5.5% to 1.7%. The model including the interaction between HMO status and source of insurance (final set of rows in Table 4) indicates that most of the difference in the number of physician visits and in the proportion with ≥1 hospital admission occurred among those with government-based insurance.
|Model (n = 2,403)||Total medical care visits||Hospital admissions|
|HMO, mean||FFS, mean||Difference (95% CI)||Model R2||HMO, %||FFS, %||Difference (95% CI)||Model χ2|
|Adjusted for HMO only||15.2||18.3||3.1 (1.7, 4.5)†||0.01||18.4||23.9||5.5 (1.7, 9.3)†||466.4|
|Adjusted for HMO and health characteristics||15.7||18.1||2.4 (1.1, 3.6)†||0.16‡||18.9||23.7||4.8 (1.2, 8.4)†||578.0‡|
|Adjusted for HMO, health characteristics, and SES/demographics||15.7||18.1||2.4 (1.1, 3.6)†||0.18‡||19.4||23.3||3.9 (0.2, 7.5)†||596.9|
|Adjusted for HMO, health characteristics, SES/demographics, and source of insurance||16.3||17.8||1.5 (0.2, 2.7)†||0.19‡||20.9||22.6||1.7 (−2.2, 5.6)||604.4‡|
|Adjusted for HMO, medical characteristics, SES/demographics, source of insurance, and interaction between HMO and insurance source§||0.19‡||605.2‡|
|Employer-based insurance||15.2||16.6||1.4 (0.0, 2.8)||18.5||19.7||1.2 (−3.1, 5.6)|
|Government-based insurance||17.2||20.5||3.4 (−0.2, 7.0)||24.8||28.3||3.4 (−5.9, 12.7)|
|Privately purchased insurance||16.2||14.7||−1.5 (−4.9, 1.9)||15.5||16.4||0.9 (−9.1, 10.9)|
Among diagnostic tests and medications used in the year prior to interview (Table 5), on an unadjusted basis, people with SLE in HMOs are less likely to report a computed tomography scan, magnetic resonance imaging, pulmonary function test, bone density scan, and use of an antidepressive agent. Sociodemographic and health characteristics account for some of these differences; after adjustment, the only difference that remains statistically significant is the lower rate of bone density scans reported by people in HMOs. Nevertheless, it is noteworthy that the point estimates indicate that subjects with SLE in HMOs are less likely to receive each of the diagnostic tests. Overall, on an unadjusted basis, persons with SLE in HMOs are 10.4% less likely to report 1 or more of these tests in the year prior to interview; after adjustment the difference is 7.2%. With respect to medications, there is no clear pattern of usage among people with SLE in HMOs and FFS.
|All† (n = 2,403)||HMO||FFS||Difference (95% CI)||HMO||FFS||Difference (95% CI)|
|Diagnostic tests in past year (%)|
|Computed tomography scan||28.7||24.9||30.8||5.9 (1.3, 10.4)‡||26.2||30.0||3.8 (−0.5, 8.1)|
|Magnetic resonance imaging||29.0||24.7||31.3||6.6 (2.3, 10.9)‡||26.5||30.3||3.8 (−0.4, 8.0)|
|Pulmonary function tests||15.0||11.6||16.8||5.2 (1.8, 8.6)‡||12.8||16.0||3.2 (−0.3, 6.7)|
|Bone density scan||37.3||33.4||39.3||5.9 (1.2, 10.5)‡||34.1||38.9||4.8 (0.2, 9.4)‡|
|Any of the above diagnostic tests||62.8||56.0||66.4||10.4 (5.4, 15.3)‡||58.1||65.3||7.2 (2.6, 11.8)‡|
|Current medication usage (%)|
|Nonsteroidal antiinflammatories||56.8||53.1||58.5||5.4 (−0.1, 10.9)||55.0||57.6||2.6 (−2.6, 7.8)|
|Cytotoxic agents||9.6||9.6||9.6||0.0 (−3.3, 3.3)||9.0||9.9||0.9 (−2.3, 4.1)|
|Immunosuppressive agents||18.6||20.1||18.0||−2.1 (−6.6, 2.4)||19.1||18.4||−0.7 (−5.0, 3.6)|
|Mycophenolate mofetil||9.2||9.9||9.0||−0.9 (−4.2, 2.4)||8.6||9.7||1.1 (−1.8, 4.0)|
|Antimalarial agents||49.8||52.3||48.4||−4.0 (−9.9, 2.0)||52.2||48.5||−3.7 (−9.4, 2.0)|
|Oral steroids||42.5||44.4||41.9||−2.5 (−8.3, 3.3)||43.8||42.1||−1.7 (−7.2, 3.8)|
|Pulse steroids||6.3||7.0||6.0||−0.9 (−3.2, 1.4)||7.4||5.9||−1.5 (−3.8, 0.8)|
|Antihypertensive agents||41.3||41.4||41.7||0.2 (−5.5, 6.0)||41.4||41.7||0.3 (−5.0, 5.6)|
|Antidepression agents||39.6||32.8||43.0||10.2 (4.9, 15.6)‡||36.3||41.1||4.9 (0.0, 9.8)|
A substantial body of research has established that HMOs, particularly the prepaid group practice form, utilize hospital care ∼25% less than FFS settings when caring for largely healthy populations (1–8). In contrast to these observations about healthy populations, studies indicate that there are no major differences in health care use among people with discrete chronic conditions in HMOs and FFS (7, 8); included in this literature are studies examining people with RA (with the possible exception that subjects with RA in HMOs may use biologic response modifiers less frequently) (10–15). There is not a substantial body of research examining the impact of HMO versus FFS care for persons with SLE. The present study was designed to address this gap.
We found that people with SLE in HMOs used substantially less health care services than those in FFS, with and without adjustment for sociodemographic and health characteristics. Of note, the difference in the number of physician visits and in the probability of hospital admissions between subjects with HMOs and FFS was greatest for those with government-based insurance, suggesting that reduced utilization in HMOs is concentrated among the aged and disabled beneficiaries of Medicare and the poor and disabled beneficiaries of Medicaid. The largely middle-class employed population may be able to negotiate the rules of HMOs in a manner that minimizes the effects of HMO coverage.
When we evaluated the specific diagnostic tests and medications used for SLE, after adjustment the HMO group was significantly less likely to report a bone density scan, although the point estimates suggest a trend in the same direction for the other diagnostic tests. In contrast, there was no pattern to the results with respect to specific medications.
The strength of the present study derives from the relatively large sample size, the diversity of the clinical and community sources of the people with SLE, and the heterogeneity of their demographic backgrounds. The principal limitation of the study is that it is based on self-report of HMO versus FFS status and health care utilization, a limitation somewhat mitigated by the use of health care utilization measures developed for and validated in the Medical Expenditure Panel Survey and by the fact that we cross-validated the multiple responses of respondents about HMO and FFS status across years, increasing the confidence in the results. Nevertheless, one cannot rule out the possibility, however unlikely, that being in a HMO affected the report of utilization in a biased fashion (i.e., because those in HMOs hear negative publicity about HMOs and this affects their responses).
The lower utilization rate of ambulatory visits, the lower frequency of outpatient surgery and hospital admissions, and the trend toward lower frequency of diagnostic testing among people with SLE in HMOs is not an indication of better or poorer quality of care since there are no validated measures of the quality of care for SLE or published guidelines for SLE care. It is unlikely, however, that the utilization differences are due to differences in the characteristics of people with SLE in HMOs and FFS because we could find no evidence of substantial differences in health characteristics between those with HMOs and FFS (the only measure on which the 2 groups differed significantly was the level of current SLE activity, and on that measure, there was only a 4% difference). Assuming that the utilization differences are not a reflection of unmeasured differences in severity or quality of care, this suggests the hypothesis that outcomes may be obtained in a more cost-effective manner in HMOs, a hypothesis to be tested using longitudinal data from the current project over the next few years.
Dr. Yelin had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study design. Yelin, Trupin, Katz, Criswell.
Acquisition of data. Yelin, Trupin, Criswell.
Analysis and interpretation of data. Yelin, Trupin, Katz, Criswell, Yazdany, Gillis, Panopalis.
Manuscript preparation. Yelin, Trupin, Katz, Criswell, Yazdany, Gillis, Panopalis.
Statistical analysis. Trupin.
The authors gratefully acknowledge the assistance of Janet Stein, Jessica Spry, and Rosemary Prem.