Abstract
Context.—Autopsy rates for patients dying in hospitals have declined from approximately 50% in the 1950s to 5% or less today.
Objective.—To investigate the nature of physician attitudes about autopsy in a large and varied population and to relate these attitudes to certain physician demographic variables.
Design.—A 10-question, anonymous, multiple-choice format questionnaire was distributed to all attending physicians at 4 hospitals: 2 that were university-affiliated (1 private and 1 public), 1 military, and 1 private nonuniversity-affiliated medical center. A total of 2608 surveys were distributed.
Results.—Three hundred eighty-eight (15%) physicians responded to the survey. Respondents agreed (77%) that autopsy results could affect their future medical practice, and disagreed (73%) that the accuracy of modern diagnostic procedures makes autopsy unnecessary. Most respondents (72%) disagreed that litigation concerns play a role in the decision to request autopsy. Spearman correlation coefficients found 9 significant relationships among the survey items, with many correlations reflecting statistically significant relationships between demographic and attitudinal variables.
Conclusions.—The survey data and statistical analysis confirm that respondents value the autopsy as a relevant clinical tool, in spite of declining requests. One of the most crucial factors influencing attitudes is the physician's level of experience with autopsy in training and practice. Among other interesting results was that strength of belief in autopsy relevancy correlates significantly (P = .003) with greater prior exposure to the autopsy. In a sense, the current low autopsy rates may be self-perpetuating because of the paucity of and decreasing experience with autopsy by succeeding generations of clinicians.
BACKGROUND
The autopsy has been an important method of medical research and education for hundreds of years. Although mostly confined to external examination, the autopsy was used in legal cases as early as the time of Charlemagne.1 In the middle 1700s, the first meaningful attempts were made to apply scientific methods and principles of causation to the performance of autopsies. Giovanni Batista Morgagni (1682–1771) brought clinicoanatomic correlation forward, arguing effectively against the humoral and anatomically inaccurate theories of Galen, which had dominated medical thinking for almost a thousand years. “[Physicians] who are not themselves dealing with the very often depressing findings of autopsy are floating in the clouds of an uncontrolled optimism,” he wrote in 1761.2(p215) His philosophical successors, Corvisart and Bichat in France and Baillie in England established accurate anatomic texts and began to correlate tissue structure with physiologic and pathologic organ function. Karl Rokitansky (1804–1878), one of the first full-time professors of pathology, who performed more than 30 000 autopsies, advocated meticulous and regularized procedures for each case. Renowned internist William Osler (1849–1919), who studied for several months with Rokitansky, wrote “To investigate the causes of death, to examine carefully the condition of organs …and to apply such knowledge to the prevention and treatment of disease, is one of the highest objects of the physician.”2(p35)
In the 20th century, autopsy contributed to the discovery and elucidation of myocardial infarction, collagen vascular rheumatoid diseases, Legionnaires disease, acute tubular necrosis, toxic shock syndrome, diabetic complications, and countless other disorders. To a great degree, medicine as a scientific field progressed in tandem with the acceptance of the autopsy and the precision of its performance.
In the 1950s autopsies were performed on approximately 50% of all patients dying in hospitals. Today, that percentage has decreased to 5% or less. A number of reasons are cited for this decline, including physician belief in modern diagnostic accuracy, lack of education and experience regarding autopsy procedure, concerns about litigation, and objections (religious and otherwise) by families. There is evidence suggesting that clinician levels of knowledge about and attitudes toward the autopsy are critical factors in determining whether an autopsy will be requested and if permission will be obtained. Clinicians and pathologists who are “unfavorably disposed towards the autopsy” have had a major negative effect on request rates and success in obtaining permission to perform autopsy.3 In a survey of 145 physicians, the single most important factor influencing clinician requests for autopsy was simply the physician's general interest in autopsies.4 A 2002 audit of autopsy request practices in Jamaica showed a 65% rate of familial agreement when autopsies were requested, in spite of a low request rate that was primarily the result of “reliance on clinical acumen.”5 A 2004 survey of 107 practitioners in Norwich, England, revealed that, although a majority of physicians were aware that autopsy rates overall had decreased, a significant number of physicians overestimated the number of autopsies that had been requested and performed within the past year.6 In contrast, a survey of 135 British general practitioners indicated that a large majority considered autopsies and autopsy reports useful to themselves and to relatives.7 In his 2003 interviews of university hospital staff, Burton8 also found that most have “positive attitudes towards the autopsy” and retain “a firm belief in the value of the procedure,” although most would impose conditions or limitations on the autopsy of themselves or a relative.
Because physician opinions about autopsy may be a major factor contributing to the decline of the use of the procedure, it is useful to investigate the nature of these attitudes in a large and varied population of physicians and to determine to which demographic variables, if any, these attitudes may be related.
MATERIALS AND METHODS
A 10-question survey was designed for distribution to all attending physicians at several institutions (Figure 1). The survey is an anonymous, 1-page, multiple-choice format questionnaire that can be completed in less than 5 minutes. Clinicians were first asked to identify their department, then to select their years of practice, and to indicate the total number of autopsies that they had personally participated in or observed. Respondents then estimated their departmental autopsy rate with “don't know” as one possible choice.
The autopsy research questionnaire distinguishes physician demographic data (questions 1 through 3), from physician knowledge about the autopsy (questions 4 through 6), from the physician attitudes about the autopsy (questions 7 through 10) that provide the basis for this work
The autopsy research questionnaire distinguishes physician demographic data (questions 1 through 3), from physician knowledge about the autopsy (questions 4 through 6), from the physician attitudes about the autopsy (questions 7 through 10) that provide the basis for this work
The remaining questions evaluated physician opinions on factors affecting utilization of the autopsy and are scaled from “strongly agree” to “strongly disagree.” Physicians opined on the sufficiency of departmental autopsy rates and whether autopsy results are available in a timely fashion. They evaluated the relevance of autopsy (its potential effect on future practice) and whether autopsy is needed because of the accuracy of modern diagnosis. Respondents also indicated whether concern over the possibility of litigation stemming from unanticipated findings could have an effect on the decision to request autopsy. The final survey question was used as a flexible item adapted to each institution and included opinions on use of the autopsy as a quality assurance tool, physician comfort level when speaking with families about autopsy, and the importance of family religious objections to requests for autopsy.
A total of 2608 surveys were distributed. The database included physicians from 4 major hospitals, 2 of which are university-affiliated (1 private, with 339 beds, and 1 public, with 300 beds), 1 military (400 beds), and 1 private medical center (850 beds). One hospital is on the East Coast of the United States and 3 are on the West Coast. Institutions were chosen by author affiliation or familiarity, and thus primarily by their accessibility. Surveys were distributed to attending physicians using an organization structure of 14 clinical departments. Department of Medicine subspecialty areas were combined under the heading of “Internal Medicine,” except for Neurology. All surgical subspecialties except Orthopaedics and Urology were grouped together under the heading of “Surgery.” Anesthesiology was considered a separate group. These groupings enabled survey results to be compared across institutions that had different administrative organizations, and facilitated combining departments with fewer than 10 total physicians into larger groups. As the principal focus of the research was on physician opinions about autopsy and how these opinions may influence requesting of autopsy, actual autopsy rates at the 4 hospitals were not considered in this work.
In order to statistically analyze the nonparametric survey data, tests for correlation using the Spearman R test were run in pairs on each of the 9 variables from the survey (item 10 was omitted because it was different in each version of the questionnaire).
RESULTS
A total of 388 (15%) physicians responded to the survey, including 113 military physicians, 94 university physicians, and 181 private hospital physicians (Table 1). Departmental response rates varied from 11% (Anesthesiology and Surgery) to 47% (Urology). In 3 of the 4 hospitals, the Pathology departments produced the lowest response rates. Physicians were fairly evenly distributed across years of practice, from less than 5 up to 20 years of practice, but there were substantially more respondents who had practiced more than 20 years than in any other category (33% of all respondents). Overall exposure to autopsy by observation or participation was low, with 42% of physicians having been involved in fewer than 5 cases, including 9% who had never participated in or seen a single autopsy. When pathologists, who are required to perform at least 50 autopsies for board certification, are removed from the sample, only 22% of remaining respondents had been exposed to more than 20 autopsies during their careers. One hundred forty-six (39%) of respondents reported their departmental autopsy rate as less than 10% of deaths in the hospital, and 196 (51%) responded that they did not know their departmental autopsy rate. With the dataset effectively halved on the question of departmental autopsy rates, it was not possible to usefully analyze opinions on this survey item.
Interestingly, 27% (“strongly agree” plus “agree”) thought that departmental autopsy rates were sufficient, and 29% (“strongly disagree” plus “disagree”) believed that rates were insufficient to meet departmental goals (Table 2). There were 44% of respondents who were neutral on the question. Responses to the question of timeliness of autopsy reporting showed a still more even distribution: 35% agreed, 31% disagreed, and 34% were neutral.
In spite of this apparent lack of overall experience, physicians mostly agreed (77%) that autopsy results could affect their medical practice, and disagreed (73%) that the accuracy of modern diagnostic procedures make autopsy unnecessary. Interestingly, many (72%) did not believe that litigation concerns play a role in the decision to request autopsy. This directly contradicts one explanation consistently offered anecdotally in the literature for declining autopsy rates, that physicians are concerned about unexpected (unanticipated) negative findings at autopsy that might provide cause for later litigation.9
In order to relate demographic variables to physician attitudes about autopsy, mean responses were calculated for the survey questions pertaining to relevance, diagnostic accuracy, and litigation concerns by (1) years of practice, (2) number of autopsy cases observed, and (3) perceived timeliness of reporting autopsy results. Also, quartiles were calculated for data results on relevance, diagnostic accuracy, and litigation concerns (Figure 2). Responses for survey questions 8 and 9 (diagnostic accuracy rendering autopsy unnecessary and litigation concerns affecting autopsy requesting) both showed closely grouped responses with the 50th percentile at a response of “4” (disagree) and both the 75th and 100th percentiles at a response of “5” (strongly disagree). Question 7, concerning whether autopsy could be relevant to future clinical practice, showed an almost mirror image pattern with the 50th percentile at a response of “2” (agree) and the 25th percentile at a response of “1” (strongly agree). These quartile data reflect the strong clustering at the extremes of response choices with responses “1” and “2” for question 7 associated with responses “4” and “5” for questions 8 and 9.
Quartiles for attitude questions demonstrate a clustering at the “strongly agree/agree” or “strongly disagree/disagree” choices, with responses to question 7 presenting a mirror image of those to questions 8 and 9. This relationship of response patterns is consistent with prior expectations for physician attitudes
Quartiles for attitude questions demonstrate a clustering at the “strongly agree/agree” or “strongly disagree/disagree” choices, with responses to question 7 presenting a mirror image of those to questions 8 and 9. This relationship of response patterns is consistent with prior expectations for physician attitudes
Physicians across years of practice agreed that autopsies are capable of providing relevant findings that could change future clinical practice (Table 3). Mean responses for all practice groups were tightly clustered around the “agree” response. Also, all groups disagreed with the statement that diagnostic accuracy may preclude the need for autopsy. Responses varied slightly more across years of practice groups on the question of concerns about litigation affecting autopsy requests, although there was still a strong trend toward disagreement with the statement.
Physician experience with or exposure to autopsy produced wider variations among mean responses to all 3 attitude questions (Table 4). Mean responses for relevancy showed overall agreement that the autopsy is relevant, and strength of belief in autopsy relevancy appears to be strongly related to increasing experience with autopsy. Respondents at the extremes of autopsy experience showed strongest disagreement with the statement regarding litigation. In other words, physicians at the middle levels of autopsy experience showed relatively more concern regarding litigation. Finally, mean responses stratified by belief in the timeliness of autopsy reporting demonstrate the strongest observable relationship to agreement with autopsy relevance (Table 5). As the last questionnaire item was used to address individual institutional interests, the specific question used was different on the 4 surveys. However, a majority of physicians at 2 institutions agreed that they felt comfortable discussing autopsy with families. Nearly equal percentages of physicians at a private hospital agreed, disagreed, and were neutral on the question of whether most families refuse autopsy because of religious objections.
Autopsy Questionnaire Results: Responses to Attitude Questions by Number of Autopsies Observed or Participated*

Autopsy Questionnaire Results: Responses to Attitude Questions by Opinions on Autopsy Timelines*

The results of the statistical analyses are summarized here with the value of the correlation (correlation coefficient) and the P value given for each significant relationship in the data (negative correlation coefficients reflect the wording of survey item 7, which resulted in “1” and “2” responses being associated with “4” and “5” responses on certain other items).
Years of practice is correlated with number of autopsy cases experienced (0.25866; P < .001).
Number of autopsies experienced is correlated with the autopsy rate being seen as not sufficient (0.22146; P < .001).
Number of autopsies experienced is correlated with agreement that autopsy can provide relevant findings to impact future practice (−0.16691; P = .003).
Autopsy rate being perceived as sufficient is correlated with autopsy reporting being seen as timely (0.31888; P < .001).
Autopsy rate being perceived as sufficient is correlated with disagreement that autopsy can provide relevant findings to have an impact on future practice (−0.23157; P < .001).
Autopsy rate being seen as sufficient is correlated with agreement that modern diagnostic accuracy precludes autopsy (0.17120; P = .002).
Agreement that autopsy can provide relevant findings is correlated with disagreement that modern diagnostic accuracy precludes autopsy (−0.36960; P < .001).
Agreement that autopsy can provide relevant findings is correlated with disagreement that litigation is a factor in autopsy requests (−0.17917; P < .001).
Disagreement that modern diagnostic accuracy precludes autopsy is correlated with disagreement that litigation is a factor in autopsy requests (0.25431; P < .001).
Analysis of the data by department or individual hospital facility was not undertaken here because these breakdowns are not central to the focus of this article. Data analysis by department, by autopsy-requesting versus nonautopsy-requesting departments, and by hospital will follow in subsequent research.
CONCLUSIONS
Our study is unusual in that it includes a relatively large sample of attending physicians at multiple types of institutions. The survey data confirm that responding physicians highly value the autopsy as a clinical tool, in spite of declining requests and decreasing exposure to autopsy during their professional lives.
Physicians in practice for more than 20 years would be expected to have seen more autopsy cases and to have a greater belief in autopsy relevance than younger physicians. These older physicians not only have longer overall experience, but were likely trained in a period when there was far more emphasis on the autopsy in education and practice. Many were training or practicing during the time that the Joint Commission on Accreditation of Healthcare Organizations still recommended that autopsy be performed on certain percentages of patients dying in hospitals. Therefore, we would expect that physicians with more years of practice would respond strongly in favor of relevancy because of the greater autopsy exposure they have experienced rather than because of their ages or time in medicine. In fact, statistical data show that the number of autopsy cases experienced is significantly correlated to years of practice (P < .001) and further, to belief in the relevance of autopsy (P = .003). However, years of practice is not independently correlated to belief in relevance.
Physicians in the 5-year to 10-year practice group noticeably differ from physicians of other levels of experience in the areas of diagnostic accuracy and litigation concern. This group may have the strongest belief in diagnostic accuracy because they are most familiar with modern technology, and have less practice experience moderating their expectations for results. Another possibility is an anomaly in the data such that responses in this category are being affected by the disproportionate numbers of physicians from departments in which very few to no autopsies are ordered. For example, the 5-year to 10-year group contains the most anesthesiologists, emergency room physicians, and radiologists of any year category. Future studies should separate out departments that do not order autopsies routinely from those that do, and observe the resultant effects on attitude trends in both types of departments.
Responses showing stronger belief in autopsy relevance correlated with disagreement with the statement that modern diagnosis makes autopsy obsolete (P < .001). These attitudes correspond in an intuitively logical way and support the internal validity of the study. Belief in the relevance of the autopsy significantly correlates with lack of concern about liability in connection with autopsy (P < .001). Physicians who disagreed that diagnostic accuracy makes autopsy unnecessary also significantly disagreed that litigation was a major factor in requesting an autopsy (P < .001). Although one might expect physicians who are convinced of autopsy relevance and less confident about diagnosis to be more concerned about a litigation byproduct of autopsy, the experiences of these respondents may have shown them that autopsy more frequently aids the clinician, as shown in a 2002 review of autopsy roles in medical malpractice cases.9 Experience in practice does not seem to diminish concern about exposure to legal liability, but exposure to autopsy does seem to alleviate this concern.
The autopsy rates at all institutions surveyed are anecdotally extremely low, particularly compared with historic standards. Physicians who had seen more autopsies significantly disagreed that departmental autopsy rates are sufficient (P < .001). Further, respondents who believed that the current autopsy rates are sufficient disagreed that the autopsy is relevant (P < .001) and also agreed that diagnosis is sufficiently accurate to make autopsy unnecessary (P = .002). The results indicate that physicians who do not believe the autopsy is an essential part of clinical practice and who believe strongly in modern diagnosis much more readily accept extremely low autopsy rates.
Belief that autopsy rates were sufficient showed a strong positive correlation with agreement that reporting was timely (P < .001). Respondents who affirmed that autopsy reporting was timely at their institutions believed in the relevance of the autopsy, although not in statistically significant numbers. It could also be that those who think the autopsy rates are sufficient have limited experience with autopsy and are responding without complete knowledge when they say that reporting is timely. Importantly, this finding and those regarding beliefs about sufficiency invite speculation as well as possible future research as to whether improving timeliness of reporting would change attitudes regarding autopsy relevance and perhaps increase autopsy request rates.
The data indicate that physicians overall retain a firm belief that autopsy can be relevant to clinical practice and this belief is maintained across various medical specialties, years of practice, and prior experiences with autopsy. However, some physicians place a great deal of faith in the diagnostic accuracy of modern methods and this may affect their autopsy-requesting behaviors. The literature has suggested that many physicians believe that only selected autopsies with specific types of diagnostic uncertainty should be performed. However, studies such as those by Goldman et al10 in 1983, Podbregar et al11 in 2001, and Shojania et al12 in 2003 have shown that physicians are not reliable predictors of which autopsies will likely show major and minor diagnostic discrepancies. Complicating matters further, many respondents overall show little to no familiarity with the autopsy process and little knowledge of numbers of autopsies within their own departments.
The literature posits that the most important variable in determining whether an autopsy is performed is the requesting physician's general attitude toward autopsy. Our data show that one of the most crucial factors influencing this attitude is the physician's level of experience with autopsy in training and practice. In a sense, the current low autopsy rates are self-perpetuating because of the paucity of and decreasing experience with autopsy of succeeding generations of clinicians. One area to explore in future research would be whether different types of experience— for example, multidepartmental autopsy review conferences—help shape physician attitudes regarding relevance. The autopsy, once such an important part of medical education and knowledge, should not be allowed to become purely a forensic tool and slip entirely out of general medical practice.
Acknowledgments
We thank Professors Paul Hooper and John Page for their advice and general contributions and Mr James Mirocha, a Cedars-Sinai Medical Center consultant, for his help in designing and running the statistical tests used in this article.
References
The authors have no relevant financial interest in the products or companies described in this article.
Author notes
Reprints: Jody E. Hooper, MD, Cedars-Sinai Medical Center, Department of Pathology and Laboratory Medicine, 8700 Beverly Blvd, Los Angeles, CA 90048 ([email protected])