Comorbidity weights have become an important tool in longitudinal outcome studies. They should be tailored toward the population and the disease state under investigation. OBJECTIVES: The objectives of the study were to develop and validate a comorbidity index for ischemic stroke patients for use in longitudinal studies. METHODS: A 5-year retrospective review of all Georgia Medicaid claims data from 1990 to 1994 was used to detect first time ischemic stroke patients. Ischemic strokes were defined by three ICD-9-CM code series (433.XX, 434.XX, and 436.XX). Comorbid conditions were measured from all claims submitted within 12 months prior the first ischemic stroke event. Half of the stroke cohort was randomly selected and multivariate logistic regression was used to derive a mortality stroke-specific weighted-index, controlling for age and gender. The Charlson and stroke specific indexes were then tested on the second half of the stroke cohort for their ability to predict risk of death. RESULTS: We identified 3,784 ischemic stroke patients with a mean age of 65 years (range 40 B 102). Of all patients, 40% died within the three-year follow-up and 73% were women. A more concise index with 7 comorbid disease states was identified. The original Charlson index has 16 comorbidities. The stepwise multiple logistic regression integer weights for the 7 comorbidities were 2 for CHF, dementia, neoplasia, and renal disease, and 6 for metastatic solid tumor, liver diseases, and AIDS. Finally, when tested on the second group, the stroke-specific index showed stepwise increases in the cumulative mortality attributable to comorbid diseases (p log rank ÷2<0.001), whereas the Charlson index did not. CONCLUSION: This shorter stroke-specific index allows for the development of more highly discriminant comorbidity models for risk adjustment.