OBJECTIVE: This study measured the overall prevalence of homelessness and tested a priori hypothesized risk factors for homelessness among patients admitted to a state hospital. The risk factors included male gender, age under 40 years, black race, urban residence, schizophrenia- related diagnosis, alcohol abuse, and drug abuse. METHOD: For 377 patients admitted to a New York state mental hospital, the 3-month, 3- year, and lifetime prevalences of homelessness were assessed. The associations between these prevalences and the hypothesized risk factors were measured by relative risks in univariate analyses and by odds ratios derived from a logistic regression in multivariate analyses. RESULTS: The 3-month prevalence of homelessness was 19%, the 3-year prevalence was 25%, and the lifetime prevalence was 28%. In univariate analyses, significant associations included drug abuse with 3-month prevalence, 3-year prevalence, and lifetime prevalence; urban residence with 3-year prevalence and lifetime prevalence; and age under 40 years with 3-month prevalence. In the logistic regression analyses, the only significant associations were urban residence with 3-year prevalence and lifetime prevalence. Male gender, black race, alcohol abuse, and schizophrenia-related diagnosis had little or no relation to homelessness. CONCLUSIONS: The overall prevalence of homelessness in these patients was remarkably high. Several strong risk factors for homelessness in the general population had only a moderate effect or no effect on homelessness in this population. Risk factors for homelessness in psychiatric patients may be somewhat different from those in the general population.