The Acute Physiology and Chronic Health Evaluation (APACHE) II classification, a measure of severity of illness in patients requiring intensive care, was devised before the rapid expansion of the acquired immunodeficiency syndrome (AIDS) epidemic. To determine the applicability of the APACHE II system to AIDS, we related observed in-hospital death rates to those predicted by APACHE II in 83 patients with AIDS. In a control group of patients without AIDS (n = 166) mean predicted and observed death rates (34.1 vs 31.3 percent) were similar. For the AIDS group overall observed mortality (63.9 percent) was significantly greater than that predicted by APACHE II (45.8 percent). The subgroup with Pneumocystis pneumonia requiring mechanical ventilation (n = 37) had an observed mortality (86.5 percent) that significantly exceeded the predicted value (44.3 percent), whereas all other AIDS patients (n = 46) showed similar predicted and observed death rates (47.0 vs 45.7 percent). APACHE II prediction of death rate also matched observed mortality in mechanically ventilated patients without Pneumocystis infection. The discrepancy between predicted and observed mortality in patients with Pneumocystis pneumonia requiring mechanical ventilation is most likely due to the lack of an APACHE II diagnostic category that accurately reflects the severity of this disease. A new diagnostic category that provides accurate outcome prediction in this patient group would form a basis for evaluation of new therapeutic interventions.