A program for quality assessment of medical diagnostics is proposed, based on detailed, ongoing comparison of clinical diagnoses with autopsy diagnoses. The model requires that the autopsy service itself be subjected to quality control to assure the reliability of the autopsy findings, including the quality of diagnoses made postmortem, and the quality of autopsy reports. Discrepancies between premortem and postmortem diagnoses are classified as to cause and magnitude; for each disease, the total diagnostic experience is collected, and sensitivity and specificity of clinical diagnostics are computed. These are compared with control ranges of sensitivity and specificity, prospectively arrived at through analysis of thousands of cases. Statistically acceptable sampling for autopsy, and review of findings by an independent body are also discussed. The findings generated by this program would be of great value in programs for quality improvement of medical care. The model has the following advantages: (1) it recognises that there is an unavoidable baseline of diagnostic errors; (2) the control levels, based on current medical practice, are prospectively established; (3) it avoids placing blame in a single case; and (4) it provides a fairly high power scrutiny that can focus on diagnostics for single diseases rather than the entire spectrum of medical practice.