Autopsies in a modern teaching hospital. A review of 2,537 cases.

We reviewed the clinical charts and the postmortem findings of 2,537 autopsies, and correlated them by constructing a "concordance score." In 2,293 cases (90%), the primary disease, as diagnosed at autopsy, had been recognized as such clinically, but in 244 cases (10%) it had not. In 2,300 cases a second disease was present, which had been diagnosed clinically in 1,795 patients. A third disease had been recognized clinically in 71% of 1,622 patients. In 1,601 patients (64%) the autopsy disclosed one or more unexpected important findings, which ranged from the main disease (244 cases) to a variety of significant incidental findings. The high incidence of surprises indicates that it is fallacious to preselect deaths for postmortem examination even when the clinical-autopsy correlation reaches 90%, and diagnostic error is only 10%.