Stroke rehabilitation outcome variation in Veterans Affairs rehabilitation units: accounting for case-mix.

Abstract

To assess variation in stroke outcomes and create a case-mix adjustment model for stroke rehabilitation in Veterans Affairs Medical Centers. Observational Within Veteran's Health Administration hospitals, there are 63 acute rehabilitation bedservice units that care for approximately 2,000 stroke patients annually. Functional gain in FIM points, length of stay (LOS), LOS efficiency (FIM gain/LOS). Significant variation in average patient functional gain, LOS, and LOS efficiency was observed among the 37 highest-volume rehabilitation units. Using analysis of covariance, a model was developed that adjusted functional gain and LOS (logged LOS) unit means using 10 potential covariates identified in a literature review and in pilot studies. Four and six covariates, respectively, were retained in the final models for FIM gain and LOS. The R2 for FIM gain and LOS accounted for by rehabilitation unit alone increased from .07 to .31 (FIM gain) and from .13 to .34 (logLOS) with the addition of the significant covariates to each model. As much as 24% of the variation in two important stroke rehabilitation outcomes is attributable to largely immutable patient and system characteristics (eg, patient function on admission, age, days since stroke onset, year of discharge, marital status, and referral source). Hence, controlling for case-mix is critical for accurate comparison of unit outcomes. Further, the variation in LOS efficiency between VA rehabilitation units suggests a large potential for cost and resource utilization savings system-wide.