OBJECTIVE Identifying substance use disorders in persons with mental illness is often difficult. In this study prevalence rates of substance use disorders among state psychiatric hospital patients were obtained by six different methods: DSM-III-R substance use diagnoses and five additional strategies based on frequency of use and past substance abuse treatment. Overlaps and differences between patients identified by the six methods were examined. METHODS Chart review and a structured substance use screening interview were used with a random sample of 20 percent (N = 474) of the population of the Maryland state hospital system. Comparisons focused on cohorts identified by two of the methods: DSM-III-R substance use diagnoses and recent regular use (any past period of daily or weekly use plus any use during the 30 days before hospitalization). RESULTS The prevalence rates of substance use identified by the six strategies ranged from 23 percent to 55 percent. The recent-regular-use criteria identified 176 patients, and DSM-III-R diagnoses identified 111. The recent-regular-use criteria also identified a greater number of patients as likely to benefit from substance use treatment. Patients identified by both methods were significantly younger and more likely to be male and nonschizophrenic than patients without substance use disorders. CONCLUSIONS The need for substance use treatment may be underestimated if discharge planners consider only DSM-III-R diagnoses. A brief screen for recent regular use may be a better way to assess treatment needs in a state hospital population.