This paper examines correlates of long-term outcome in a sample of 89 stroke patients discharged from medical rehabilitation. Patients in this study were recruited in 1984 and followed prospectively for a 12-month period postdischarge. This analysis builds on previous studies which have called researchers to move beyond single-outcome studies which use only bivariate analysis. Multivariate techniques are used to evaluate the influence of a number of predictor variables on three measures of long-term outcome: 1) a composite variable that includes functional status, mortality and discharge disposition (home v nursing home); 2) life satisfaction; and 3) medical charges. The results show that 1) multivariate analyses present a different picture from that obtained when using bivariate analysis; and 2) the same predictor variables are not equally important in predicting different outcome variables. From 30-42% of the variance in the three dependent variables is explained by severity of illness, function at admission, age, wheelchair use and in- and out-of-house social supports.