OBJECTIVE We hypothesized that a systematic determination of symptom severity would predict psychiatric admission for non-suicidal patients referred for a psychiatric evaluation in an urban emergency department (ED) setting. METHODS In a pilot study involving consecutive patients referred for a psychiatric evaluation in an urban ED, symptom severity was quantified using the Brief Psychiatric Rating Scale (BPRS). The BPRS scores of all non-suicidal patients were subjected to receiver operator characteristic (ROC) curve analysis to determine the sensitivity, specificity, and optimal cutoff score of the BPRS in predicting admission for non-suicidal patients. RESULTS A BPRS cutoff score of 39 had a sensitivity of 85.71 percent and a specificity of 86.11 percent. The area under the ROC curve was .8671 (Somer's D = .7342) and the standard error of the curve was .1124. The cutoff score of 39 correctly identified six of seven non-suicidal patients who were hospitalized. CONCLUSIONS It is anticipated that use of the BPRS in the ED will be further refined when a larger patient sample is studied. Potentially, a subset of BPRS items could be identified which would be more sensitive in predicting admission than the full BPRS and increase the overall efficiency of administering the BPRS in the ED.