An extensive literature documents substantial rates of missed diagnoses that are detected at autopsy, including diagnoses that likely affected outcome.1–4 Physicians generally have attributed these findings to selection bias. With the national autopsy rate for nonforensic deaths having fallen to roughly 5 percent,5 physicians may request autopsy precisely in those cases likely to reveal important new diagnoses.
In response to a request by the College of American Pathologists and funded by the Agency for Healthcare Research and Quality, we systematically reviewed the literature to assess time trends in autopsy-detected diagnostic errors and to specifically address the question of whether clinical selection bias explains the persistent error rates that are reported even in contemporary autopsy series.6
We reviewed more than 200 studies of autopsy-detected errors in clinical diagnosis,6 53 of which met prespecified inclusion criteria.1,2 Discrepancies between clinical and autopsy diagnoses were defined as “major errors” when a clinically missed diagnosis involved a principal underlying disease or primary cause of death; and they were defined as “class I errors” when the patient might have survived to hospital discharge if antemortem diagnosis had occurred.
The 53 studies included some that spanned a 40-year period (1959 to 1999) and reported major and class I error rates of 23.5 percent (range, 4.1 to 49.8 percent) and 9.0 percent (range, zero to 20.7 percent), respectively.
Despite steady declines in major and class I error rates, typical U.S. institutions in the year 2000 observed a major error rate ranging from 8.4 to 24.4 percent, based on autopsy rates ranging from 100 percent (the extrapolated extreme at which clinical selection is eliminated) to 5 percent (roughly the national average). The same range of autopsy rates would produce class I error rates of 4.1 to 6.7 percent.7
Although we used the term “error” (in keeping with the rest of the autopsy literature), the discrepancies between clinical and autopsy findings combine true errors with difficult cases. On the other hand, it is worth noting that these error rates do not include less dramatic but potentially important missed diagnoses, nor do they address delayed diagnosis. Diagnoses detected at any time before death were counted as clinically recognized even if earlier detection would have improved outcome.
These findings have important implications, especially for primary care physicians, who are most likely to be caring for patients at the end of life. Autopsy remains a diagnostic gold standard; failure to establish a cause of death is estimated to be less than 5 percent in adult deaths.6
With regard to inpatient deaths, it is tempting to assume that the barrage of sophisticated diagnostic tests readily available in contemporary hospitals obviates the need for autopsy. However, hospitalized patients accounted for the overwhelming majority of patients in the studies included in our analysis. Moreover, clinicians have only a modest ability to identify inpatient deaths in which autopsy will or will not yield important new information.6 With regard to outpatient deaths, the importance of autopsy is presumably greater, even in cases of apparently clear-cut sudden cardiac death.8
Physicians may worry about the possibility of malpractice claims, but the existing literature suggests that autopsy helps physicians more often than not.6 In fact, we found no reported case of malpractice action initiated solely on the basis of unexpected diagnostic errors or complications detected at autopsy. In a recent study assessing the role of autopsy information in malpractice cases, it was found that defendant physicians usually were exonerated, and observance of the standard of care was deemed more important in determining medical negligence than accuracy of clinical diagnosis.9
Although family members of a deceased patient sometimes worry about the potential effect of autopsy on funeral arrangements, it has virtually no effect on any aspect of a funeral. With appropriate notification to the pathologist, autopsy can be completed within 24 to 48 hours, and incisions on the face and head are difficult to detect even by the trained observer, creating no problems for open-casket ceremonies. Unfortunately, one issue of which clinicians in community practice need to remain aware is cost—in some regions, families must pay for nonforensic autopsies. Even in such regions, though, academic centers perform free autopsies and may accept referrals from other hospitals, because dwindling autopsy rates have made it difficult for pathology programs to provide access to sufficient cases for trainees.
Many physicians feel uncomfortable recommending autopsy to family members,10 presumably because of lack of experience because autopsy rates have fallen to such low levels.5 Criteria for autopsy referral are determined by the county or state. At the hospital level, each hospital is required to include autopsy criteria as part of its policy. As with advance directives and preferences for end-of-life care, autopsy discussions are generally less difficult than physicians initially imagine and can be rewarding, as patients express feelings about their illness and goals of care for the time they have left.
In interviews with family members of deceased patients who have undergone autopsy, it has been found that knowing the cause of death with certainty is comforting to the family, as is reassurance that a loved one received all that current medical care could provide.11