The lengths of psychiatric hospital stays and community stays.
No abstract. Conclusion: This paper has demonstrated the value of including observed covariates, duration dependence, and unobserved heterogeneity in a survival model of psychiatric hospital and community stays. In particular regression analysis would have been inappropriate because a) there is censoring, and b) there is nontrivial duration dependence. Kaplan-Meier estimation is deficient because of significant covariates and unobserved heterogeneity. Cox regression would be hard to implement because of unobserved heterogeneity. We find that sex, marital status, employment status, diagnosis, age, committed days, and urban/rural codes help explain hospital stay lengths, and that marital status, employment status, and some county medical resources measures help explain community tenure. Previous history helps explain both stay lengths. There is significant evidence of interesting duration dependence and of unobserved heterogeneity.