Even in patients with severe stroke, it is necessary to predict in an early stage whether they can be discharged home from hospital in order to offer the later services at home. The purpose of this study was to develop a clinical prediction rule (CPR) for discharge home after rehabilitation in patients with severe stroke.

The subjects were 206 patients with stroke during post-acute phase, registered in the Japan Rehabilitation Database, whose Functional Independence Measure (FIM) scores was 36 points or less. The following potential independent variables at the time of admission to the rehabilitation ward were collected; age, gender, side of lesion, duration since stroke onset, National Institutes of Health Stroke Scale score, cognitive status, speech disorder status, motor/sensory impairment status and scores of each item of FIM. The collected data were evaluated using Classification and Regression Trees (CART) analysis method to develop a CPR for discharge home.

Forty-six percent of the patients were discharged home after rehabilitation. The CART analysis found the CPR included age, motor FIM score, motor impairment status of upper limb, functional status of eating, grooming, and memory (sensitivity=76.6%, specificity=76.8%, positive predictive value=73.5%, negative predictive value=79.6%, accuracy=76.7%). The best predictor for discharge home was the level of functional status of eating, the next predictors were age and motor FIM score.

The CPR with moderate accuracy was developed to predict discharge home after rehabilitation in patients with severe stroke. Further investigation including environmental factors is necessary to improve accuracy of the CPR.