• In order to determine the best combination of variables for predicting which patients would be released from state mental hospitals and which would be retained, the 1,500 patients examined in the course of a cross-sectional survey of the residental population of Texas state mental hospitals in 1966 were followed up in 1971.

Eighteen variables were selected for discriminant analysis, comparing the patients never released from the hospital with those known to be current extrainstitutional residents. The stepwise program selected four variables: (1) length of prior stay; (2) relatives' attitudes toward release; (3) social adequacy; and (4) conceptual disorganization. Applying this model to the study groups, 79% were correctly classified with regard to their in hospital/not in hospital status, suggesting that early identification of chronicity in future studies is feasible.