The value of histological examination in the audit of hospital autopsies: a quantitative approach.

The aims of this study were to compare the clinical with autopsy diagnoses, to evaluate the role of histological examination in the pathological diagnoses and to assess the new pathological diagnoses uncovered by autopsy. We aimed to obtain quantitative assessment of the sensitivity, specificity and accuracy of clinical diagnoses. The guidelines for postmortem reports by the Royal College of Pathologists (1993) were implemented for reports used in this study. These guidelines are similar in intent to those of the College of American Pathologists. Complete macroscopic and histological studies of 108 (53 females) autopsies were analysed. The mean age was 78.0+/-9.0 (SD) years (range 54-94 years). The interquartile range (25%ile 75%ile) was 72-84 years, with a median of 79.5 years. Seventy per cent of all causes of death were confirmed by macroscopical and histological examination. Sixty-one clinical diagnoses were inconsistent with the pathological findings. Histological examination contributed significantly to the final diagnosis in major (5%) and minor (6%) clinicopathological as well as new pathological findings (23%). The most common causes of death not suspected clinically were pulmonary embolism (23%), bronchopneumonia (22%), ischemic heart disease (13%) and malignancies (10%). The clinical sensitivity of antemortem diagnoses was 25% for peritonitis and 24% for pulmonary embolism. The overall clinical sensitivity was 54% and specificity 92%. The accuracy of positive diagnosis was 69% and accuracy of negative diagnosis 88%. Unexpected causes of death represented a third of all causes of death reported. Histological examination is an important tool in hospital autopsy audit. A quantitative approach can be used to assess the accuracy of postmortem clinical diagnoses, to identify the possible source of clinical diagnostic weakness, and provide data that may be of use for diagnostic precision in the more difficult clinical subjects.