Objective: To identify the frequency and spectrum of clinically relevant diagnoses found at autopsy but not determined before death in adult patients admitted to an intensive care unit (ICU).
Patients and methods: We retrospectively reviewed medical records and autopsy reports of patients admitted to ICUs from January 1, 1998, to December 31, 2000. Disagreements between autopsy and antemortem diagnoses were classified as type I or type II errors. A new major diagnosis with potential for directly impacting therapy was considered a type I error. Type II errors included important findings that would not have likely changed therapy.
Results: Of 1597 deaths in all ICUs during the study period, autopsies were performed in 527 patients (33%). Autopsy reports were available in 455 patients, of whom 19 (4%) had type I errors and 78 (17%) had type II errors. The most common type I error was cardiac tamponade. There were no significant differences in age, sex, or length of stay in the ICU or hospital among patients with and without diagnostic errors or among patients with type I and II errors. Seventy-eight patients had 81 type II errors. Organ transplant recipients had more type I or II errors than did nontransplant patients (35% vs 20%; P = .04).
Conclusions: Diagnoses with impact on therapy and outcome are missed in approximately 4% of deaths of adult patients admitted to the ICU. Transplant recipients are especially likely to have occult conditions for which additional therapy might be indicated.