Is necropsy a valid monitor of clinical diagnosis performance?

OBJECTIVE: To improve the validity of comparisons between clinical and postmortem diagnoses when postmortem diagnosis is used to monitor clinical diagnosis performance.
DESIGN: Analysis of elementary examples. MAIN OUTCOME MEASURES: Sensitivity and specificity of clinical and postmortem diagnoses and confirmation and agreement rates. Sensitivity and specificity permit valid comparisons of clinical and postmortem diagnoses among different procedures, sites, or times whereas agreement and confirmation rates may be misleading. Estimates of sensitivity and specificity, however, can be severely distorted by factors such as non-random selection of cases for necropsy or by unrecognised errors in postmortem diagnosis. Such distortion may be minimised by (a) estimating the likely magnitude of errors in postmortem diagnosis, (b) specifying standard conditions for performing necropsies, and (c) ensuring an unbiased sample of moderate size rather than a large biased sample.
CONCLUSION: Sensitivity and specificity should be used as measures of agreement between clinical and postmortem diagnoses. IMPLICATION: Monitoring of clinical diagnosis performance by necropsy surveys requires ensuring accuracy of pathological examinations and validity of study design and analysis.