![[ Visit Client Website ]](images/banner.gif)
PURPOSE. To develop and validate a comorbidity index (CI) suitable for outpatient rheumatic disease (RD) care and research that is useful and valid across different RDs.
METHODS. We examined the association of 22 comorbid conditions and 6 outcomes in 28,495 patients participating in a longitudinal study of rheumatic disease outcomes. At six month intervals, patients reported comorbid conditions as “current” or as “past” from which a third category, “either” current or in the past, was derived. We determined the ability of each comorbid condition to predict direct medical costs, work disability, social security disability (SSD), HAQ functional disability, hospitalization and mortality, adjusting for age, sex and ethnicity in multivariable regressions. Based on these data we constructed 30 candidate CIs, and compared the predictive ability of each using the Bayesian Information Criterion (BIC).
RESULTS. For outcomes studied the “current” category produced the greatest odds ratios though the best model fits were obtained by the “either” category. For semi self-evaluative outcomes such as HAQ, work disability and SSD, depression dominated the multivariate predictors (Table 1). Pulmonary disorder was the best predictor of mortality and costs, and hospitalization was predicted strongly by MI, pulmonary, other CV conditions (OCV), hypertension and depression. We identified the 5 best CIs based on average BIC in all outcomes from which we selected the CI with best mortality and hospitalization response. By BIC criteria the best CI was “much better” than the next (28 difference). The selected CI had a 0-9 range and included the following comorbidities (“either” as default): pulmonary, MI, OCV, stroke, hypertension, diabetes (current), spine/hip/leg fracture, depression, GI ulcer, other GI, cancer (current). The final CI had little difference in predictive value among different RDs (eg., mortality ANOVA p=0.41). The CI’s mean (SD) score by RD was: RA 1.67 (1.53), OA 1.97 (1.53), SLE 2.56 (1.86) and FMS 2.22 (1.56). The addition of specific conditions (noted by “*” in Table 1) with the CI improved predictive ability of the outcome (change in BIC > 10).
CONCLUSIONS. Comorbid conditions such as depression and pulmonary problems, which may not be specifically documented in physician records, are powerful predictors of adverse outcomes in RDs. An 11-item CI for use in rheumatology was identified as the best overall comorbidity predictor of major outcomes. Even so, indices dilute the effect of individual comorbidities. Therefore, for studies of specific outcomes, the index should be supplemented by up to 3 specific comorbid predictors.
Significant (p<0.001) comorbidity predictors by outcome (odds ratio/coefficient, z/t-score)
*Combined with final CI, greatly improves model fit for outcome
METHODS. We examined the association of 22 comorbid conditions and 6 outcomes in 28,495 patients participating in a longitudinal study of rheumatic disease outcomes. At six month intervals, patients reported comorbid conditions as “current” or as “past” from which a third category, “either” current or in the past, was derived. We determined the ability of each comorbid condition to predict direct medical costs, work disability, social security disability (SSD), HAQ functional disability, hospitalization and mortality, adjusting for age, sex and ethnicity in multivariable regressions. Based on these data we constructed 30 candidate CIs, and compared the predictive ability of each using the Bayesian Information Criterion (BIC).
RESULTS. For outcomes studied the “current” category produced the greatest odds ratios though the best model fits were obtained by the “either” category. For semi self-evaluative outcomes such as HAQ, work disability and SSD, depression dominated the multivariate predictors (Table 1). Pulmonary disorder was the best predictor of mortality and costs, and hospitalization was predicted strongly by MI, pulmonary, other CV conditions (OCV), hypertension and depression. We identified the 5 best CIs based on average BIC in all outcomes from which we selected the CI with best mortality and hospitalization response. By BIC criteria the best CI was “much better” than the next (28 difference). The selected CI had a 0-9 range and included the following comorbidities (“either” as default): pulmonary, MI, OCV, stroke, hypertension, diabetes (current), spine/hip/leg fracture, depression, GI ulcer, other GI, cancer (current). The final CI had little difference in predictive value among different RDs (eg., mortality ANOVA p=0.41). The CI’s mean (SD) score by RD was: RA 1.67 (1.53), OA 1.97 (1.53), SLE 2.56 (1.86) and FMS 2.22 (1.56). The addition of specific conditions (noted by “*” in Table 1) with the CI improved predictive ability of the outcome (change in BIC > 10).
CONCLUSIONS. Comorbid conditions such as depression and pulmonary problems, which may not be specifically documented in physician records, are powerful predictors of adverse outcomes in RDs. An 11-item CI for use in rheumatology was identified as the best overall comorbidity predictor of major outcomes. Even so, indices dilute the effect of individual comorbidities. Therefore, for studies of specific outcomes, the index should be supplemented by up to 3 specific comorbid predictors.
Significant (p<0.001) comorbidity predictors by outcome (odds ratio/coefficient, z/t-score)
| Mortality (OR, z) | Hospitalization (OR, z) | SS Disability (OR, z) | Disabled (OR, z) | HAQ (b, t) | Medical Costs (b, t) |
| Lung (2.2, 14.7)* | MI (1.8, 6.8)* | Depression (1.6, 14.6)* | Depression (2.1, 18.7)* | Depression (0.19, 14.3)* | Lung (1200, 5.5) |
| MI (1.5, 5.9)* | Lung (1.4, 4.8) | Fracture (1.6, 10.5)* | Diabetes (1.8, 10.6)* | Fracture (0.18, 7.7)* | Liver (1300, 3.8) |
| Fracture (1.4, 4.6) | Other CV (1.4, 4.6) | Lung (1.5, 9.0) | Lung (1.5, 7.9) | Diabetes (0.15, 8.0) | Diabetes (760, 3.5) |
| Stroke (1.5, 4.4) | Hypertension (1.3, 4.3) | Diabetes (1.4, 8.2) | Hypertension (1.3, 7.0) | Ulcer (0.13, 7.8)* | |
| Diabetes (1.3, 4.0) | Depression (1.3, 4.2) | Ulcer (1.3, 6.1) | GI (1.3, 6.2) | GI (0.11, 7.7) | |
| Hypertension (1.2, 5.5) | Fracture (1.4, 5.8) | Cataract (0.11, 7.7) | |||
| GI (1.2, 5.5) | Ulcer (1.3, 5.2) | Hypertension (0.08, 7.5) | |||
| Neurological (1.3, 4.1) | Neurological (1.5, 5.0) | Lung (0.12, 6.6) | |||
| Other CV (1.3, 4.7) | Allergies (0.10, 5.9) | ||||
| Neurological (0.13, 4.2) | |||||
| Other CV (0.07, 4.1) | |||||
| Stroke (0.10, 3.6) |
K. Michaud, None; F. Wolfe, None.
See more of: Epidemiology and Health Services Research III
See more of: ACR Abstract Submission Poster Sessions
See more of: ACR Abstract Submission Poster Sessions