Correlation of clinical diagnosis with autopsy findings

We investigated 110 cases, selected at random out of the total of 1876 autopsies performed in the Institute of Pathological Anatomy and Histology at the University of Ferrara-Arcispedale Santa Anna on patients who had died at the hospital during 1983-87. Clinical data were taken from 'necropsy request forms' filled in by clinicians and from medical records. We then evaluated the extent of agreement and disagreement, expressed as underdiagnosis (false-negative) and overdiagnosis (false-positive), between the clinical and pathological records with regard to primary disease and to cause of death. Agreement between the diagnoses was 81% for primary disease and 58% for cause of death. The diagnoses of neoplastic, cerebrovascular and cardiovascular diseases showed the closest agreement. Among the neoplasms, those of the liver, gall-bladder, pancreas, retroperitoneal space and prostate were most often overlooked in clinical diagnoses. We had conflicting results for cancer of the lung and of the colon-rectum, for which there was a high level of agreement, but also a large number of false-positive cases and cases in which they were found by chance at autopsy. For cerebrovascular diseases, false-negative and false-positive diagnoses were seen most often for cause of death. With regard to cardiovascular diseases, a relatively uniform distribution was found for myocardial infarction among the three categories, and a high rate of agreement was found for pulmonary embolism. Of all diseases, bronchopneumonia was associated with the highest percentage of false-negative diagnoses for cause of death. Our data on digestive diseases show the strongest agreement on diagnosis of primary disease in relation to cirrhosis of the liver; a high rate of agreement on cause of death was confirmed for alimentary bleeding. Active tuberculosis was detected only at autopsy. We conclude that autopsy is a valid tool for investigation, despite the availability of sophisticated diagnostic techniques.