A model to aid in the prediction of discharge location for stroke rehabilitation patients

Objective: To determine which demographic and medical factors recorded on admission to a rehabilitation unit best predict discharge accommodation outcomes. Design: Retrospective chart review. Setting: Inpatient rehabilitation unit in an academic hospital in southwestern Ontario, Canada. Participants: One hundred four stroke patients (54 women, 50 men: mean age, 72.0y) admitted to the rehabilitation unit over a 4-year period. Interventions: All patients underwent evaluations by the physical therapy, occupational therapy, social work, speech pathology, and psychology departments. Patients were divided into 2 groups: (1) no change in premorbid accommodation and (2) change in premorbid accommodation. Main Outcome Measures: Demographic, clinical, and housing information (premorbid, discharge) and functional data (FIM(TM) instrument, Chedoke-McMaster Stroke Assessment [CMSA] Impairment Inventory, Berg Balance Scale [BBS]) were recorded for each patient. Results: Of 104 patients, 24 were discharged with a change in premorbid accommodation. Change in discharge location was significantly associated with age, gender, and the presence of premorbid social support (P<.01), but not with type of premorbid living arrangement. Statistically significant differences were noted between total FIM scores (P<.001), BBS scores (P<.001), and the postural component of the CMSA Impairment Inventory (P<.03). A logistic regression model, predicting 67% of the variance, was created to predict discharge accommodations. Conclusions: Patients admitted to the rehabilitation unit can be scored to obtain their predicted chance of being discharged with a change from their premorbid accommodations. The equation is relatively easy to calculate and is based on data that are commonly collected in rehabilitation.