A comparison of two comorbidity instruments in arthritis

Comorbidity (CM) is a powerful predictor of health outcome and cost, as well as an important confounder in many epidemiologic studies. However, choosing the most appropriate CM measurement instrument is difficult because comparative data on how the available instruments perform in various disease settings are limited. We collected CM data (from the complete medical records) for two population-based prevalence cohorts with rheumatoid arthritis (RA) and osteoarthritis (OA) and a comparison cohort without arthritis (NA), using two different CM instruments: the Charlson CM index (Charl), which is based on 17 diagnoses each weighted by mortality risk, and the Index of Coexistent Diseases (ICED), which estimates the severity and frequency of 14 comorbid conditions and provides an assessment of the impairment or disability caused by each. Cox proportional hazards modeling was used to assess the impact of the two types of comorbidity scores (Charl and ICED) on survival after prevalence (index) date, adjusting for the age, sex, and disease status. There were 450, 441, and 889 individuals in the RA, OA, and NA groups, respectively, with a mean follow-up period of 10.6 years. During the follow-up, 293, 307, and 546 deaths occurred in the RA, OA, and NA groups, respectively. The mean age and percent females were: 63.3 years, 74%; 70.7 years, 74%; and 67.5 years, 75% for the RA, OA, and NA groups, respectively. Comorbidity was highest in RA, intermediate in OA, and lowest in NA by both Charl and ICED. Cox proportional hazards modeling demonstrated that both Charl and ICED were highly statistically significant predictors of mortality (P<0.0001) after adjusting for age, sex, and disease state (RA, OA, or NA) and that ICED remained highly significant as a predictor of mortality, even after adjusting for Charl. We conclude that estimating CM from medical records using ICED, an instrument that incorporates an assessment of impairment and disability, is feasible and that such as assessment provides information that independently predicts mortality, even after adjusting for the results of traditional diagnosis-based CM measures, such as Charl.