A proportional hazards model for the prediction of psychiatric rehospitalization

Cox regression and logistic regression models were used to explore the predictors of psychiatric rehospitalization in 558 Medicaid managed care beneficiaries. The Chronic Disability Payment System (CDPS) was used to create measures of psychiatric morbidity based upon the subjects’ diagnostic history the year prior to the study year. Younger subjects, who possessed a larger number of co-morbid psychiatric diagnoses, as well as a diagnosis of schizophrenia in the year before rehospitalization behavior was measured, were found to be rehospitalized at a greater rate and to have less community tenure from discharge to rehospitalization than subjects that did not have these characteristics. Subjects who had outpatient psychiatric care in the interim between discharge and rehospitalization were also found to be at greater risk for rehospitalization. No significant interactions were discovered. Theoretical as well as practical implications of these findings are discussed.