Improvements in Clinical Diagnostic Accuracy After a 5-Year Systematic Analysis of Clinical and Autopsy Discrepancies

Abstract

To the Editor.—Clinical autopsies have contributed considerably not only to the development of medical knowledge1 but also to the quality assurance of medical care.2 Surprisingly, in the last 6 decades, autopsy rates decreased from 50% during the 1950s to around 5% nowadays.3Some authors have demonstrated that current discrepancy rates between clinical and autopsy diagnosis remain similar to those obtained 30 years ago,4 but others3 have shown a decrease inversely related to the autopsy rates. Unfortunately, most autopsy series have been collected retrospectively, which prompted us to conduct a 5-year prospective study with intervention. The basis of the intervention was double: first, the complete lack of knowledge of the postmortem diagnoses when clinical diagnoses were stated in every included patient (blinded to the autopsy findings); and second, the subsequent analysis of cases in which discrepancies were found, performed jointly by clinicians and pathologists involved in this intervention. The aim was to determine whether clinical diagnostic accuracy improved after the systematic analysis of preautopsy diagnoses stated by the clinicians involved, and the resulting analysis of clinical and autopsy discrepancies for a long period of time. The study was centered in 3 medical hospital areas (internal medicine, emergency medicine, and medical intensive care unit) of a 1000-bed teaching hospital.From June 1995 through June 2000 a total of 2984 consecutive dying patients were included. Autopsy was performed in 266 (8.9%); the mean age of autopsied patients (62.8 years; SD 16.6) was significantly lower than that of patients in whom autopsy was not performed (68.2 years; SD 16.5), (P < .001). Finally, 256 patients were available for analysis.The discrepancy rate in the underlying and/or immediate cause of death showed a statistically significant decrease from 43.3% at baseline (first year of the study) to 25.9% in the fifth year (P = .001; see the Figure). Furthermore, the class I discrepancy (supposed adverse impact in prognosis) also showed a significant decrease from 30% to 17.2% (P = .03; see the Figure). Advanced age was the only factor independently associated with a higher risk of discrepancy (relative risk = 1.04; 95% confidence interval, 1.01–1.07). As happened in other studies,5 pneumonia was found to be the leading disease in discrepancy, occurring in 30 (40%) of 75 cases with this autopsy diagnosis, followed by pulmonary thromboembolism, which occurred in 12 (50%) of 24 cases. When the cause of discrepancy was analyzed, the most frequent explanation was a misinterpretation of clinical data, which occurred in 39.1% of these cases. On the other hand, in 12.4% of cases the final diagnosis was not pursued due to the patient's poor clinical status and irreversibility of the disease.This study emphasizes the high value that clinical autopsies still have for clinicians. We believe that good clinical practice still needs the benefit of establishing routine meetings for preautopsy clinical diagnosis discussion before knowing autopsy diagnoses, and also needs continuous feedback between clinicians and pathologists for a systematic revision of autopsy-detected diagnostic discrepancies. To provide a useful tool for improving clinical diagnostic accuracy, these procedures should be maintained over time.