An algorithm for screening psychiatric patients for physical disease was empirically derived from a comprehensive assessment of 509 patients in California's mental health system. The first 343 patients were used to develop the algorithm, and the remaining 166 were used as a test group. Calculations were made for several versions of the algorithm, and the data were compared with the diagnoses listed in the patients' admission mental health record. The algorithmic procedure was more accurate and more cost-effective than the medical evaluation procedures used by the state mental health system. When applied to the test group, the algorithm detected up to 90 percent of patients who had an active, important physical disease at a cost of $156 per patient. The mental health system had detected 58 percent of test-group patients with a disease at a cost of $230 per patient.