SEARCH

SEARCH BY CITATION

Keywords:

  • Comorbidity;
  • Questionnaire;
  • Health status;
  • Health care utilization

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Objective

To develop the Self-Administered Comorbidity Questionnaire (SCQ) and assess its psychometric properties, including the predictive validity of the instrument, as reflected by its association with health status and health care utilization after 1 year.

Methods

A cross-sectional comparison of the SCQ with a standard, chart abstraction-based measure (Charlson Index) was conducted on 170 inpatients from medical and surgical care units. The association of the SCQ with the chart-based comorbidity instrument and health status (short form 36) was evaluated cross sectionally. The association between these measures and health status and resource utilization was assessed after 1 year.

Results

The Spearman correlation coefficient for the association between the SCQ and the Charlson Index was 0.32. After restricting each measure to include only comparable items, the correlation between measures was stronger (Spearman r = 0.55). The SCQ had modest associations with measures of resource utilization during the index admission, and with health status and resource utilization after 1 year.

Conclusion

The SCQ has modest correlations with a widely used medical record-based comorbidity instrument, and with subsequent health status and utilization. This new measure represents an efficient method to assess comorbid conditions in clinical and health services research. It will be particularly useful in settings where medical records are unavailable.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Studies of the effectiveness of treatments or the performance of hospitals, health care organizations, or individual physicians must take into account patient characteristics that independently affect outcomes (1). These characteristics typically include age, sex, the severity of the disease under study, physical and mental health status, socioeconomic status, patient preferences, and the extent and severity of comorbid conditions. Comorbidity is a particularly important prognostic factor, with well-documented effects upon mortality (2, 3), surgical outcome (4), complications (5, 6), functional status (7), hospital length of stay (8, 9), and discharge status (10).

Measures of comorbidity typically use information from the medical record or administrative data (11). These approaches impose limitations, such as the availability of medical records and the quality of documentation. Moreover, reliance on trained chart abstractors is expensive and time consuming. It may also be unnecessary in some settings. Research has shown that patients can accurately assess their current (12) and past medical conditions (13, 14), including comorbidities (15, 16).

This article details the development and validation of a self-administered measure of comorbidity for clinical and health services research settings. This instrument is short, easily understood, and can be completed by individuals without any medical background. It also allows the subject to note the severity of each comorbid condition and their perception of its impact on their function. We hypothesized that the Self-Administered Comorbidity Questionnaire (SCQ) would have moderately strong agreement with a chart review-based measure of comorbidity, and that it would predict subsequent health status and resource utilization.

SUBJECTS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Development of the questionnaire.

A panel of 5 physicians selected 12 medical conditions according to their frequency in general practice and their inclusion in published and commonly used comorbidity instruments, including the Charlson Index (17), the Cumulative Illness Rating Scale (CIRS) (18), and the Index of Co-existent Disease (ICED) (19). The medical conditions were simplified to language that could be understood without any prior medical knowledge. The instrument is shown in Figure 1. The question “Do you have any of the following problems?” was asked in relation to heart disease, high blood pressure, lung disease, diabetes, ulcer or stomach disease, kidney disease, liver disease, anemia or other blood disease, cancer, depression, arthritis, and back pain. Additionally, subjects have the option of adding 3 additional conditions in an open-ended fashion. Rheumatoid arthritis and osteoarthritis are assessed separately but were considered together in the analysis because of our concern that patients might not distinguish these disorders accurately.

thumbnail image

Figure 1. The Self-Administered Comorbidity Questionnaire

Download figure to PowerPoint

For each problem, the subject is asked “Do you receive treatment for it?” as a proxy for disease severity. Certain comorbid conditions (e.g., hypertension) typically do not limit a subject's overall function, whereas others (e.g., back pain) might impose severe functional limitations. To capture this burden to the subject, we added the question, “Does it limit your activities?” for each medical condition. The questionnaire was pilot tested on a sample of 20 patients to ensure clarity and comprehension.

Scoring of the questionnaire.

An individual can receive a maximum of 3 points for each medical condition: 1 point for the presence of the problem, another point if he/she receives treatment for it, and an additional point if the problem causes a limitation in functioning. Because there are 12 defined medical problems and 3 optional conditions, the maximum score totals 45 points if the open-ended items are used and 36 points if only the close-ended items are used.

Subjects.

The study subjects consisted of 170 consecutive admissions to 3 general medical and 3 general surgical care units at the Brigham and Women's Hospital in Boston, Massachusetts. To be eligible, patients had to be over 50 years of age (to increase the likelihood subjects would have multiple medical/surgical problems) and be able to complete the questionnaire in English. Patients undergoing surgery completed the questionnaire more than 48 hours after surgery. The Human Investigations Committee at the Brigham and Women's Hospital approved the study protocol and all subjects signed informed consent.

Study design and data sources.

All subjects completed the comorbidity questionnaire along with questions about demographic data (age, sex, race, insurance status, educational attainment), prescription and nonprescription medications at admission, and hospitalizations during the previous 12 months. Twenty-six subjects completed the same questionnaire at least 24 hours apart to assess test-retest reliability. A research nurse trained in the use of the Charlson Index abstracted the medical records and assigned a Charlson Index score using a standardized chart abstraction form for each patient. The nurse was blinded to the questionnaire data.

Data on resource utilization during the hospital stay were derived from the Brigham and Women's Hospital Management Information System, which documents charges for all procedures and resources billed by the hospital.

All subjects were contacted by mail 1 year after completion of the original questionnaire. This followup survey elicited information about visits to physicians and use of prescription and nonprescription medications. Health status was assessed with the Medical Outcomes Study Short Form 36 (SF-36), a 36-item inventory that assesses physical functioning, mental functioning, role functioning due to physical and mental limitations, health perceptions, social functioning, bodily pain, and vitality (20). Subjects who did not respond received 2 reminder letters by mail.

Statistical analysis.

The Charlson Index was scored according to the published algorithm (17). Test-retest reliability of the SCQ and the Charlson Index were assessed by the Spearman correlation coefficient and the intraclass correlation coefficient (21). Kappa statistics were calculated for the test-retest reliability of individual items (22).

The association between the medical record-derived Charlson Index and the questionnaire-based SCQ was assessed with the Spearman correlation coefficient. For these comparisons, we computed correlation coefficients for the original scales, including all items, and for reduced scales, restricted to comparable items between both instruments. For individual diseases addressed by both instruments, we calculated the kappa statistic, which adjusts for chance agreement. We also calculated overall agreement, defined as the number of cases in which both the patient and reviewer said “yes” plus the number in which both the patient and reviewer said “no,” divided by the total number of cases.

We assessed the correlation between the Charlson Index and the SCQ within the following strata: patient age (younger than 70 years versus 70 years or older), sex, educational attainment (no college versus at least some college), race (white versus other), insurance status (private insurance versus Medicare, Medicaid, or no insurance), and hospital service (surgical versus medical).

For each instrument, we assessed the correlation between the comorbidity summary score and the self-reported number of prescription medications, number of over-the-counter (OTC) medications, and number of hospitalizations in the previous 12 months. We calculated Spearman correlation coefficients between each instrument and hospital length of stay, total hospital charges, medication charges, and laboratory charges. We tested whether the correlation coefficients for the two different comorbidity measures were significantly different using the method described by Kleinbaum et al (23). The critical P value for this analysis was set at 0.007 to protect against false-positive results due to multiple comparisons.

We examined the correlation between the SCQ and measures of resource utilization. Separate correlations were calculated using 1) a simple count of the comorbidities; 2) the count of comorbidities plus the data on treatment; and 3) the full score, including count of comorbidities, treatment for the comorbid problem, and patient-reported functional limitations due to the comorbidity. We hypothesized that the strength of association would increase for analyses that used the data on treatment and functional limitations.

We then fitted linear regression models with the Charlson Index score as the dependent variable and the SCQ score as the principal independent variable. Age, sex, educational attainment, race, insurance status, and hospital service were included as covariates. We also included interaction terms between the questionnaire score and each covariate. We fitted models in a stepwise fashion for the comorbidity instrument in total and for a truncated version that contained only items that were also contained in the Charlson Index.

We calculated scores for the eight SF-36 subscales, the Physical Component Summary (PCS) score, and the Mental Component Summary (MCS) score according to published algorithms (24, 25). To assess the relative contribution of comorbidity to the SF-36 subscales and summary scores, we fitted linear regression models with health status (the SF-36 subscales and summary scores) as the dependent variable, comorbidity as the primary independent variable, and covariates including age, sex, race, insurance status, and educational achievement. We fitted models with the total SCQ scores (including treatment and functional limitation) serving as the principal independent variable. Other models used just a count of comorbidities alone, and a score derived from a count of comorbidities plus the data on treatment.

Inclusion of optional comorbid conditions of the SCQ increased the average summary score by 0.63, but did not change the results for reliability and predictive validity. Therefore, the results we present are restricted to the close-ended items.

Data analysis was performed on a personal computer using the SAS statistical package (SAS Institute, Cary, NC).

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Recruitment and baseline characteristics.

During a 6-week period, we screened 453 inpatients for inclusion in the study and randomly selected 229. Of these subjects, 32 patients were excluded, 14 because they were too ill to participate, 8 because they didn't understand English well, and 10 because of cognitive deficits that impeded completion of the questionnaire. Twenty-seven patients refused participation, leaving a final sample of 170.

Baseline characteristics of the 170 patients are presented in Table 1. Fifty-five percent were female, average age was 65.3 ± 8.8 years, 31% were over 70 years old (median 64, range 50–89). The sample was predominantly white (82%) and 50% had at least some college education. Fifty-four percent of the sample was hospitalized on medical services with an average length of stay of 8.8 ± 9.9 days; 46% was hospitalized on surgical services with an average length of stay of 8.7 ± 6.2 days. Excluding charges for physicians, hospital charges totaled $21,314 ± 33,883 for medicine patients and $22,748 ± 13,892 for surgery patients (1999 dollars).

Table 1. Baseline characteristics of study sample (n = 170)
Age, years, mean (SD)65.3 (8.8)
Sex (%) 
 Female94 (55)
Race (%) 
 White140 (82)
 African American22 (13)
 Hispanic5 (3)
 Other3 (2)
Education (%) 
 Less than high school36 (21)
 High school49 (29)
 Some college85 (50)
Insurance status (%) 
 Private78 (46)
 Nonprivate (Medicaid, Medicare)92 (54)
Service (%) 
 Medical92 (54)
 Surgical78 (46)
Mean length of stay, days (SD) 
 Medical service8.8 (9.9)
 Surgical service8.7 (6.2)
Mean total hospital charges, 1999 US dollars (SD) 
 Medical service21,314 (33,883)
 Surgical service22,748 (13,892)
Mean Charlson Index score (SD)1.59 (1.8)
Mean Self-Administered Comorbidity Questionnaire score (SD)5.61 (4.1)

The patients had a range of admitting diagnoses with the following represented by more than 10 individuals: myocardial infarction, angina, other cardiac problems, lower gastrointestinal diseases, infectious diseases, peripheral vascular disorders, osteoarthritis, and other musculoskeletal disorders. The most common surgical interventions included total joint arthroplasty (in 25 patients), other orthopedic surgeries, and gastrointestinal surgery.

The average score of the SCQ was 5.61 (SD 4.1, median 5.0, range 0–25). The mean Charlson Index score was 1.59 (SD 2.13, median 1.0, range 0–9). On the open-ended section of the questionnaire, 34 patients (20.0%) indicated 1 additional disease, 6 patients (3.5%) indicated 2 diseases, and 1 patient indicated 3 diseases. These diagnoses included hyperthyroidism (n = 3), peripheral vascular disease (n = 3), benign prostatic hyperplasia (n = 2), multiple sclerosis (n = 2), osteoporosis (n = 2), systemic lupus erythematosus (n = 2), diverticulits (n = 2), “blood clotting” (n = 2), and gastroesophageal reflux (n = 2). Twenty-six patients who indicated additional comorbid conditions were receiving treatment for their problems and 17 had functional limitations due to those.

Reliability and validity.

The test-retest reliability for the SCQ in 26 patients was 0.94 (95% confidence interval 0.72, 0.99) as calculated by the intraclass correlation coefficient and 0.81 by the Spearman correlation coefficient. (These findings compare with a test-retest reliability of the Charlson instrument of 0.92 as measured with the intraclass correlation coefficient and 0.94 as measured with the Spearman coefficient [15]). The test-retest reliability of specific items was moderate to high, ranging from kappa = 0.40 for back pain to kappa > 0.9 for heart disease, hypertension, lung disease, diabetes, kidney disease, anemia, and depression.

The Spearman correlation between the SCQ and the Charlson Index was 0.32 for the entire instruments and 0.55 for truncated versions of the measures that contained only comparable items for each instrument. The associations between the SCQ and the Charlson Index on an item level are demonstrated in Table 2. Overall agreement exceeded 90% except for lung disease (78%) and heart disease (88%). The kappa statistic, which adjusts for chance agreement, documented low concordance for lung disease (kappa = 0.27), and moderate to high concordance (kappa ≥ 0.46) for the other conditions (Table 2).

Table 2. Prevalence of comorbid conditions and agreement between Charlson Index and Self-Administered Comorbidity Questionnaire (SCQ)*
 Prevalence (%)Agreement, %Kappa (95% CI)
Charlson IndexSCQ
  • *

    Data limited to items covered by both instruments. 95% CI = 95% confidence interval.

Heart disease37 (22)22 (13)780.50 (0.39, 0.65)
Lung disease17 (10)14 (8)880.27 (0.04, 0.50)
Peptic ulcer disease14 (8)15 (9)920.46 (0.21, 0.70)
Liver disease02 (1)990.93 (0.88, 0.99)
Diabetes29 (17)32 (19)970.90 (0.82, 0.99)
Kidney disease8 (5)9 (5)980.79 (0.56, 0.99)
Cancer37 (22)35 (21)900.68 (0.53, 0.82)

Table 3 demonstrates Spearman correlation coefficients between comorbidity instruments and health care utilization. The correlations between the SCQ and prescription and OTC medications were higher than the correlations between the Charlson Index and prescription and OTC medications. Both the SCQ and the Charlson Index had modest correlation with hospitalizations in the previous 12 months for surgical patients and weak correlations with these variables for medical patients. The SCQ correlated weakly with hospital charges and length of stay. The corresponding correlations were higher for the Charlson Index, particularly for hospital charges, which reached correlations of 0.35 (pharmacy charges in medical patients). The difference in correlation between the SCQ and Charlson Index was statistically significant (P < 0.001) for all variables except laboratory charges among medical patients and hospitalizations in the last year and length of stay among surgical patients.

Table 3. Spearman correlation between comorbidity instruments and measures of health care utilization at the time of admission*
 Hospitalizations in previous yearPrescription medsOTC medsAcute inpatient chargesLength of stay
Total chargesPharmacy chargesLaboratory charges
  • *

    OTC = over the counter; Meds = medications; SCQ = Self-Administered Comorbidity Questionnaire.

  • P < 0.05.

  • P < 0.01.

  • §

    P < 0.001.

Medical Service       
 Charlson Index summary score0.220.020.060.260.350.150.20
 SCQ       
  Problem0.210.360.320.06−0.060.110.05
  Problem and treatment0.220.410.390.09−0.010.130.01
  Summary score0.210.400.390.09−0.040.140.03
Surgical Service       
 Charlson Index summary score0.350.300.150.070.150.290.13
 SCQ       
  Problem0.310.50§0.46§0.020.110.120.08
  Problem and treatment0.340.57§0.53§0.090.130.150.13
  Summary score0.370.55§0.54§0.100.090.140.14

None of the covariates age, sex, race, insurance status, hospital service, or educational attainment confounded the association between the Charlson Index and SCQ. None of the interaction terms between these variables and the SCQ was statistically significant.

Predictive validity.

Of the 170 patients, 104 (61.1%) responded to a questionnaire after 12 months, which included items about current health status and health care utilization in the previous year. Six patients had died, 1 person moved away, 1 homeless person could not be traced, 1 person became blind, and 21 patients refused participation in the followup survey. Thirty-five patients (20.6%) did not respond after 3 mailings.

Responders to the followup questionnaire did not differ significantly from nonresponders with respect to age, sex, educational attainment, hospitalizations in the year prior to their admission, number of prescription and OTC drugs at baseline, length of stay, and total hospital charges. However, nonresponders were more likely to be surgical patients (P = 0.03) and white (P = 0.03).

SF-36 scores and correlations with measures of comorbidity are shown in Table 4. The normalized scores for the SF-36 scales PCS and MCS for the general US population are 50 with a standard deviation of 10 (0 indicating worst, 100 indicating perfect health) (24). SF-36 scores for the 8 subscales ranged from 47.9 ± 43.2 for role functioning due to physical limitations to 76.9 ± 27.1 for social functioning. The SCQ had modest correlations with physical function score (r = 0.34) and the physical component score (r = 0.35), and weak associations with the mental health score (r = 0.15). Correlation coefficients between the Charlson Index and all SF-36 subscales were lower than for the SCQ.

Table 4. SF-36 scores at followup and correlation with comorbidity at baseline*
SF-36 scale at followupScore (SD)Spearman correlation with baseline comorbidity scores
SCQCharlson Index
  • *

    SF-36 = Short Form 36; SCQ = Self-Administered Comorbidity Questionnaire.

  • P < 0.001.

  • P < 0.01.

  • §

    P < 0.05.

Physical function58.4 (29.9)−0.34−0.27
Role function physical47.9 (43.2)−0.20§−0.06
Mental function74.7 (18.0)−0.15−0.04
Role function emotional74.1 (36.8)−0.09−0.06
Social function76.9 (27.1)−0.24§−0.14
Bodily pain63.1 (25.3)−0.29−0.13
Energy/vitality54.7 (24.3)−0.18−0.07
General health60.4 (24.3)−0.39−0.33
Physical component summary39.2 (12.6)−0.35−0.23§
Mental component summary51.9 (10.5)−0.03−0.07

The prediction of health status 1 year after assessment of comorbidity is shown in Table 5. Multivariable linear regression models included the SF-36 scales as dependent variables and comorbidity as the primary independent variable. The models controlled for age, sex, ethnicity, educational achievement, and insurance status. We display total model variation explained (R2) only if the association of the independent variable (comorbidity score) was statistically significant (P < 0.05). In these cases, we also show the percent of the total model variation that is explained solely by the comorbidity score. The total score of the SCQ explained substantial variation for all SF-36 subscales except for mental health, role functioning due to emotional distress, and the MCS scale. In general, the amount of variation explained was greater for the total SCQ score (including count of comorbidities and data on treatment and functional limitation) than for the simple count of comorbidities or the count plus the data on treatment. The Charlson Index was a significant predictor of only 3 outcomes: physical function (R2 = 0.22 of which 24.5% was explained by comorbidity), social functioning (R2 = 0.05 of which 81.8% was explained by comorbidity), and general health (R2 0.22 of which 54.2% was explained by comorbidity).

Table 5. Prediction of health status (SF-36) after 1 year by comorbidity in multivariable analyses*
Independent variableCharlson Index, total scoreSCQ, total scoreSCQ, only question about presence of comorbiditySCQ, only question about presence of comorbidity + treatment
Model R2% of variation explained by comorbidityModel R2% of variation explained by comorbidityModel R2% of variation explained by comorbidityModel R2% of variation explained by comorbidity
  • *

    Adjusted for age, sex, ethnicity, insurance status, and educational achievement. SF-36 = Short Form 36; SCQ = Self-Administered Comorbidity Questionnaire; NS = not significant.

Dependent variable        
 Physical function0.2224.50.2547.60.2238.60.2342.6
 Role function physicalNS 0.1420.4NS NS 
 Mental functionNS NS NS NS 
 Role function emotionalNS NS NS NS 
 Social function0.0581.80.1064.7NS NS 
 Bodily painNS 0.1950.00.1638.90.1741.6
 Energy/vitalityNS 0.2030.4NS NS 
 General health0.2254.20.2466.90.2162.50.2264.1
 Physical Component SummaryNS 0.2269.30.1759.20.1963.3
 Mental Component SummaryNS NS NS   

The SCQ score obtained at the index admission correlated moderately with the number of prescription (Spearman r = 0.37) and OTC (r = 0.32) medications taken 1 year later. The Charlson Index correlated poorly (r ≤ 0.14) with these measures of utilization. The frequency of doctors' visits did not correlate with either the SCQ score (r = 0.15) or the Charlson Index score (r = 0.09).

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Coexisting conditions may exert a powerful influence upon outcomes. We have developed a brief, comprehensive, self-administered questionnaire to assess comorbidities. The questionnaire is highly reproducible and its correlation with the Charlson Index, a widely used chart review-based instrument, was moderately strong. Agreement between both scales on an item level was modest, with kappas exceeding 0.46 for all but 1 item.

Methods to assess comorbidities in a standardized manner such as the Charlson Index (17), the CIRS (18), and the ICED (19) rely on medical record abstraction, which is time and resource intensive and infeasible when medical records are not available. Hence, these methods have limited utility in large-scale clinical and epidemiologic studies. The patient is an attractive source of data on comorbid conditions because, after all, the patient is the principal source of this information in his/her medical record. Prior research has indicated that the patient can accurately and reliably report coexisting diseases, particularly for specific conditions such as systemic lupus erythematosus or for surgical procedures (12). In prior research we showed that a questionnaire version of the Charlson Index was reproducible and accurate compared with its chart review-based counterpart (15).

Many comorbidity instruments, including the Charlson Index, were developed for hospitalized patients to adjust for mortality rates. These instruments may have limitations in adjusting for functional status as an outcome. Studies of the effectiveness of health care increasingly focus on functional status and other quality of life outcomes. Greenfield et al have developed a measure of case mix for office practice that uses patients' report on symptoms and diseases, as well as patients' self-perceived disease severity (16). This measure of overall disease burden accounts for severity of 15 different disease groups. The instrument had a statistically significant association with health status as measured by the SF-36. Similarly, our instrument had modest associations with health status 1 year after assessment of comorbidity. Predictive validity was highest for the physical functioning subscale of the SF-36. We hypothesized that adding items about treatment (as a surrogate for disease severity) and functional limitation would improve predictive validity. The findings in Table 5 provide modest support for this hypothesis.

Comorbidity refers to “diseases unrelated in etiology or causality to the principal diagnosis” (1), and must be separated from complications or sequelae of the principal reason for which a patient is treated. Nonetheless, we included the index disease in our analysis. We felt that it would not be possible for individuals to separate index from coexisting disease, particularly by questionnaire. We used only the close-ended responses in these analyses because open-ended responses are difficult to standardize and code.

Several limitations of our study and the instrument itself need to be acknowledged. The questionnaire was studied in inpatients. Further study is needed to assess the performance of the measure in outpatients. Because hospital records are usually more complete than outpatient records, the self-reported data may be even more useful in the outpatient setting. This study was conducted in a large academic center. In smaller, nonteaching centers, the medical records may be less complete, suggesting additional advantages to a self-administered approach. On the other hand, the medical record-based approach would be more advantageous in centers with more complete records. We excluded patients unable to complete the questionnaire because of language barriers or because they were too ill or cognitively impaired. These exclusions further limit generalizability to populations who typically carry a high burden of comorbidity.

We conclude that the SCQ is a reproducible measure of comorbid conditions that has moderately strong associations with a standard medical record-based comorbidity measure. Our data provide initial and preliminary support for the validity of the new measure. The instrument will likely be especially useful in settings in which the medical record is not available. Further work is needed to confirm and expand upon our findings in other clinical settings, overcome the limitations of our study, and define the specific research situations in which the SCQ should be used.

Postscript. Dr. Oliver Sangha, our friend and colleague and the first author of this paper, died in March 2001 at age 39. Oliver's love of the clinical sciences, health policy, family, and friends was complete and unconditional. His vision was prescient and inspired, his intellect sparkling, his laughter infectious, and his generosity boundless. He made a difference.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES
  • 1
    Greenfield S. The state of outcomes research: are we on target. N Engl J Med 1989; 320: 1142.
  • 2
    Greenfield S, Aronow HU, Elashoff RM, Watanabe D. Flaws in mortality data: the hazards of ignoring comorbid disease. JAMA 1988; 260: 22535.
  • 3
    Iezzoni LI, Foley SM, Daley J, Hughes J, Fisher ES, Heeren T. Comorbidities, complications, and coding bias: does the number of diagnosis codes matter in predicting in-hospital mortality. JAMA 1992; 267: 2197203.
  • 4
    Katz JN, Lipson SJ, Larson MG, McInnes JM, Fossel AH, Liang MH. The outcome of decompression laminectomy for degenerative lumbar spinal stenosis. J Bone Joint Surg Am 1991; 73: 80916.
  • 5
    Roos NP, Wennberg JE, Malenka DJ, et al. Mortality and reoperation after open and transurethral resection of the prostate for benign prostatic hyperplasia. N Engl J Med 1989; 320: 11204.
  • 6
    Greenfield S, Apolone G, McNeil BJ, Cleary PD. The importance of co-existent disease in the occurrence of postoperative complications and one-year recovery in patients undergoing total hip replacement. Med Care 1993; 31: 14154.
  • 7
    Katz JN, Wright EA, Guadagnoli E, Liang MH, Karlson EW, Cleary PD. Differences between men and women undergoing major orthopedic surgery for degenerative arthritis. Arthritis Rheum 1994; 37: 68794.
  • 8
    Jencks SF, Williams DK, Kay TL. Assessing hospital-associated deaths from discharge data: the role of length of stay and comorbidities. JAMA 1988; 260: 22406.
  • 9
    Cleary PD, Greenfield S, Mulley AG, Pauker SG, Schroeder SA, Wexler L, et al. Variations in length of stay and outcomes for six medical and surgical conditions in Massachusetts and California. JAMA 1991; 266: 739.
  • 10
    Deyo RA, Cherkin D, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol 1992; 45: 61319.
  • 11
    Iezzoni L, editor. Risk adjustment for measuring health care outcomes. Ann Arbor (MI): Health Administration Press; 1997.
  • 12
    Mechanic D. The experiences and reporting of common physical complaints. J Health Soc Behav 1980; 21: 14655.
  • 13
    Linet M, Harlow S, McLaughlin JK, McCaffrey LD. A comparison of interview data and medical records for previous medical conditions and surgery. J Clin Epidemiol 1989; 42: 120713.
  • 14
    Colditz GA, Martin P, Stampfer MJ, Willett WC, Sampson L, Rosner B, et al. Validation of questionnaire information on risk factors and disease outcomes in a prospective cohort study of women. Am J Epidemiol 1986; 123: 894900.
  • 15
    Katz JN, Chang LC, Sangha O, Fossel AH, Bates DW. Can comorbidity be assessed by questionnaire rather than medical record review? Med Care 1996; 34: 7384.
  • 16
    Greenfield S, Sullivan L, Dukes KA, Silliman R, D'Agostino R, Kaplan SH. Development and testing of a new measure of case mix for use in office practice. Med Care 1995: 33: AS4755.
  • 17
    Charlson ME, Pompei P, Ales K, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chron Dis 1987; 40: 37383.
  • 18
    Linn BS, Linn MW, Gurel L. Cumulative illness rating scale. J Am Geriatrics Soc 1968; 16: 6226.
  • 19
    Greenfield S, Apolone G, McNeil BJ, Cleary PD. The importance of co-existent disease in the occurrence of postoperative complications and one-year recovery in patients undergoing total hip replacement. Med Care 1993; 31: 14154.
  • 20
    Ware JE, Sherboune CD. The MOS 36-item short-form health survey (SF-36). I. Conceptual framework and item selection. Med Care 1992; 30: 47381.
  • 21
    Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull 1979; 86: 4208.
  • 22
    Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977; 33: 15974.
  • 23
    Kleinbaum DG, Kupper LL, Muller KE. The correlation coefficient and straight-line regression analysis. In: PayneM, editor. Applied regression analysis and other multivariate methods. Belmont (CA): Wardsworth Publishing Company; 1988. p. 80.
  • 24
    Ware JE, Kosinski M, Keller SD. SF-36 Physical and Mental Health Summary Scales: a users manual. Boston: The Health Institute, New England Medical Center; 1994.
  • 25
    Ware JE, Kosinski M, Bayliss MS, McHorney CA, Rogers WH, Reczek A. Comparison of methods for the scoring and statistical analysis of SF-36 profile and summary measures: summary of results from the Medical Outcomes Study. Med Care 1995; 33: AS26479.