This study estimated the ASD prevalence in a psychiatric hospital and evaluated the Social Responsiveness Scale (SRS) combined with other information for differential diagnosis. Chart review, SRS and clinical interviews were collected for 141 patients at one hospital. Diagnosis was determined at case conference. Receiver operating characteristic (ROC) curves were used to evaluate the SRS as a screening instrument. Chi-squared Automatic Interaction Detector (CHAID) analysis estimated the role of other variables, in combination with the SRS, in separating cases and non-cases. Ten percent of the sample had ASD. More than other patients, their onset was prior to 12 years of age, they had gait problems and intellectual disability, and were less likely to have a history of criminal involvement or substance abuse. Sensitivity (0.86) and specificity (0.60) of the SRS were maximized at a score of 84. Adding age of onset <12 years and cigarette use among those with SRS <80 increased sensitivity to 1.00 without lowering specificity. Adding a history substance abuse among those with SRS >80 increased specificity to 0.90 but dropped sensitivity to 0.79. Undiagnosed ASD may be common in psychiatric hospitals. The SRS, combined with other information, may discriminate well between ASD and other disorders.