Predicting rehospitalization at a community mental health center: a "double-crossed" validation.

Used a multiple regression equation that contained a variety of demographic and diagnostic measures as predictor variables to predict 5-year recidivism in a group of 100 psychiatric inpatients hospitalized at a community mental health center. This equation then was cross-validated via the prediction of recidivism in a second group of 100 inpatients. A predictive equation derived from an analysis of this second group of patients then was used in a "double-crossed' validation back to the first group. Diagnoses that involved schizophrenia and drug abuse, major psychotropic drug regimens, and a failure to appear for follow-up therapy as an outpatient emerged consistently as predictor variables.