The 60-item General Health Questionnaire (GHQ) and the National Institute of Mental Health Diagnostic Interview Schedule (DIS) were administered to 501 patients presenting for assessment or treatment of alcohol or drug problems at the Addiction Research Foundation, Toronto, Canada. The GHQ proved to be a useful, if limited, instrument for routine screening for psychiatric disorders other than substance abuse in this population. The sensitivity or "utility" of the test at the optimum threshold score was considerably lower in this sample than in medical or community samples. Receiver Operating Characteristic (ROC) analysis suggests a need for an upward revision of the cut-off score to 23/24 inpatients with alcohol or drug problems. At this threshold, sensitivity was 69%, specificity 75%, positive predictive value 77%, negative predictive value 66%, and overall misclassification rate 29%. A definition of "caseness" was used that included all DIS Axis I disorders with the exception of substance use and "cognitive impairment." Sensitivity of the GHQ was highest for the affective disorders (79%). False negatives accounted for the bulk (61%) of the misclassification rate. At the threshold score of 23/24, positive predictive values were greater for the unemployed and for patients with either no current substance use disorder or with both alcohol and drug disorders.