Canonical correlation analysis of residential and vocational outcomes following psychiatric rehabilitation

This analysis uses canonical correlation techniques to explore client outcomes following psychosocial rehabilitation services in two rehabilitation areas: vocational and residential. At issue are the ways in which these two outcomes are affected by a series of eight client characteristics and service delivery features. The overarching question is the effectiveness of the program in establishing independence vocationally and residentially, and whether the same set of factors influences these two outcomes for persons with mental illness. Subjects were 212 ex-clients of a psychosocial rehabilitation agency who were working six months after leaving the program. The analysis derived two canonical functions both of which were significant. Canonical loadings and cross loadings indicated that client level of functioning and program tenure loaded most strongly with residential status on the first canonical variate while education loaded most strongly with salary on the second. These results support the notions that different predictors account for the variance associated with different outcomes following psychosocial rehabilitation. Discussion of the relevance of strong predictor variables highlights ways they can impact program design. Finally, the usefulness of canonical correlation in longitudinal program evaluation research is explored. Copyright (C) 1996 Elsevier Science Ltd