A mathematical instrument was developed to supplement the diagnostic information available to physicians in the emergency room to improve physicians' diagnostic accuracy in managing patients with acute ischemic heart disease and thereby reduce inappropriate coronary care unit admissions. The instrument was empirically derived and is based on nine clinical, historical, and electrocardiographic predictive variables. Probabilities of acute ischemic heart disease generated by the instrument were given to the house staff in an emergency room during alternate months. Comparison of the control months (455 patients) with the experimental months (401 patients) showed the following: The overall diagnostic accuracy increased from 83% to 91% (P less than 0.005), the overdiagnostic accuracy increased from 51% to 33% (P less than 0.01), and the admission rate to the coronary care unit fell from 26% to 14% (P less than 0.001), while the inappropriate discharge rate from the emergency room did not change, 3% versus 3% (not significant).