OBJECTIVE: To determine whether our practice of requesting an autopsy for patients who die in the medical intensive care unit (MICU) continues to be a valid approach to obtain clinically and educationally relevant findings. METHODS: In this retrospective study conducted in an adult MICU population of a university hospital, the clinical diagnoses and postmortem major diagnoses of 100 patients who died in 1996 (autopsy rate of 93%) were compared. RESULTS: Eighty-one percent of the clinical diagnoses were confirmed at autopsy. In 16%, autopsy findings revealed a major diagnosis that, if known before death, might have led to a change in therapy and prolonged survival (class I missed major diagnoses). The most frequent class I missed major diagnoses were fungal infection, cardiac tamponade, abdominal hemorrhage, and myocardial infarction. Another 10% of autopsies revealed a diagnosis that, if known before death, would probably not have led to a change in therapy (class II error). CONCLUSIONS: Autopsy remains an important tool for education and quality control. In contrast with historical series of 1 to 2 decades ago, there is a clear shift in the type of class I missed major diagnoses toward opportunistic infections. Bedside-applicable techniques such as electrocardiography with supplemental posterior leads, echocardiography, and meticulous abdominal ultrasonography might improve the outcome in selected MICU patients.