Comorbidity measures for stroke outcome research: a preliminary study.

Abstract

To develop standardized comorbidity measures for use in stroke outcome research. Retrospective review of medical records to analyze comorbidities and to study reliability and validity of the newly developed measures, comorbidity index (CI), and weighted comorbidity index (w-CI). Tertiary rehabilitation center in Japan. 106 stroke patients, age 56.5 +/- 13.2 yr, admitted and discharged during the year from May 1994 to December 1995. The median days of duration of stroke, onset to admission, and length of stay (LOS) were 199, 83, and 105.5, respectively. The median admission and discharge Functional Independence Measure (FIM) raw scores were 85 and 110, respectively. Assessment of interrater reliability with intraclass correlation coefficient (ICC) for total scores and weighted kappa for subscores; assessment of concurrent validity by relating the measures to Charlson's comorbidity index, total numbers of medications, laboratory studies, therapeutic interventions, consultations, and days of interruption (Spearman's rank correlation method); study of predictive validity with discharge FIM score and LOS as dependent variables. The ICCs were .896 for CI and .997 for w-CI, and weighted kappa ranged from .615 to 1.00. CI and w-CI correlated significantly with Charlson index and the above indices of validity. They also correlated negatively with discharge FIM scores and positively with LOS. With stepwise multiple regression analysis, 79.8% of the variance of discharge FIM scores could be explained by w-CI, days from onset to admission, admission FIM score, and deviation in tape bisection task. The newly developed comorbidity measures are reliable and valid for use in stroke outcome research.