Background: Studies comparing the accuracy of clinical diagnosis in unselected patients who died in hospital in different medical eras have shown no decline of errors in the main diagnosis. We assessed changes in diagnostic accuracy over 20 years.
Methods: We analysed retrospectively diagnostic errors, with use of necropsy as the gold standard for diagnosis. We randomly selected 300 patients who died at a tertiary-care teaching hospital in Switzerland--100 in each of 1972, 1982, and 1992. We classified discrepancies between clinical diagnosis and necropsy findings as major and minor errors.
Findings: The overall necropsy rate at the hospital stayed at around 90% for the whole period. During the study, the frequency of major discrepancies declined significantly (1972, 30%; 1982, 18%; 1992, 14%; p=0.007). The rate of minor diagnostic errors increased significantly from 23% in 1972 to 46% in 1992 (p<0.001). The increase in overall diagnostic accuracy occurred mainly because of a significant improvement in specificity for cardiovascular diseases (1972, 85%; 1982, 82%; 1992, 97%; p=0.034) and non-significantly improved sensitivity (1972, 69%; 1982, 82%; 1992, 86%; p=0.061). Sensitivity also improved for infectious diseases (1972, 25%; 1982, 67%; 1992, 86%; p=0.036). Sensitivity and specificity for neoplastic diseases were high originally and did not change. The total number of diagnostic procedures per year increased from 191 in 1972 to 259 in 1992, mainly because of non-invasive techniques, such as ultrasonography, and endoscopies.
Interpretation: The frequency of major diagnostic errors in unselected patients who died in hospital was halved over 20 years, probably because of improved clinical skills and new diagnostic procedures.