Nonparametric mixture logistic regression models for clinical disposition

Psychiatric hospitalization is one of the most important decisions in the care of patients. This is attributed to the complexity and intensity of treatment and isolation of patients from the family and community into a typically highly controlled and supervised setting. Logistic regression models are fitted to assess the relationship between DSM-III axes and psychiatric hospitalization decisions. Since the apparent error rate tends to underestimate the true error rate, the estimate for the downward bias of the apparent error was computed. A generalization of the logistic regression model is fitted where the intercept is assumed to be a random parameter.