Secondary Logo

Journal Logo

Hospital autopsy audit: discordant primary clinical diagnoses are found in 20% of cases in a reducing autopsy case load. Selection bias or significant findings?

Jackett, Louise1,2; McLean, Catriona3,4

Author Information
Pathology - Journal of the RCPA: October 2015 - Volume 47 - Issue 6 - p 499–502
doi: 10.1097/PAT.0000000000000297
  • Free

The post-mortem examination is an effective tool that provides insight into a patient's medical care and assists in auditing clinical decision making.1,2 Thus, post-mortems have the potential to improve quality of care, educate the medical community, ensure accurate epidemiological health statistics and assist bereft relatives in coming to terms with end of life issues.1–8

The post-mortem examination, also known as autopsy or necropsy, became an influential medical procedure in the Middle Ages that significantly advanced our understanding of human disease.7 Standardised by Virchow, ‘the dissection of the dead body to determine, through direct observation, the cause of death or the nature of the disease’ was common place and an integral part of medical curricula until the mid-twentieth century.3–5,7,9

However, Australian and international hospital autopsy rates have declined in recent decades, due in part to advances in diagnostic technologies and possibly an unfamiliarity of their value among doctors with little exposure to autopsies.1,3–5,8–13 Comprehensive Australian data from the mid-2000s reported that the nation's hospital autopsy rate declined from 21% to 12% over a 10 year period.1 Similar observations have been made elsewhere, including in the USA, UK, Europe, Hong Kong and China.2,4,8,9 Despite this decline, discrepancies between clinical and autopsy diagnoses continue to be found in limited studies published over the last 15 years.2,6,8,10,11,14–18 A meta-analysis of autopsy studies by Roulson et al. found that discrepancies in major diagnoses and potentially treatable conditions occurred at rates of 15–41% and 10–44%, respectively.8 Therefore, many authors contend that the autopsy remains a relevant and valuable tool with the potential to inform and educate the medical community.1,4,6,8,15,19

Advances in diagnostic technologies are understandably a factor that has contributed to the decline.6,11 Other reasons include unfavourable public attitudes, restrictions placed on the extent of examination by next-of-kin, lack of resources, poor training of medical staff in autopsy consent, time-consuming consent processes and unfamiliarity among the medical community of the value of autopsy.1–6,9,12,13

As far as the authors are aware, hospital autopsy data have not been published from Australian sources for a decade.1 Therefore, we undertook an audit of non-coronial autopsies between August 2011 and August 2013 at a major Australian tertiary centre for trauma and heart-lung transplantation to examine concordance of clinical and autopsy findings.

Using retrospective data from death certificates, hospital records and autopsy reports, the clinical cause of death (COD) and major co-morbid diagnoses were compared with autopsy findings for all hospital post-mortems performed across a 2 year period (August 2011–August 2013). Cases referred to the coroner were excluded from the study. In every case a death certificate was completed by the clinical team prior to the autopsy being conducted. The type of autopsy (partial or full) was recorded for each case and a comprehensive microscopic examination was carried out where possible within the limitations of consent.

Based on the autopsy findings, the death certificate and clinical diagnoses were assessed as either correct or discordant. Significant missed diagnoses that may have impacted clinical care were also noted.8,20,21 Minor discrepancies that were considered not to have impacted on the patient's acute episode of care were excluded from the final results.20,21 Thus, a discordant autopsy was one where the autopsy findings were considered a substantially different pathological process contributing to death when compared to the death certificate and major ante-mortem diagnoses. The findings were then rated in accord with past studies.1,2,6,8,10,14–19,21

Of 1298 hospital deaths between August 2011 and August 2013, 86 patients were referred for autopsies (autopsy rate 6.6%). Full autopsies accounted for 70% (60/86) and partial for 30% (26/86). Ages ranged from 14 to 91 (mean age 61). Male cases accounted for 52% (45/86) and organ transplant recipients for 29% (25/86).

Cause of death was assessable at post-mortem in all but one partial autopsy. In the latter case the COD (multiple myeloma) was unable to be fully evaluated since the next-of-kin consented only to examination of the heart, in which amyloid of the myocardial vessels was diagnosed.

Of conclusive autopsies, correct diagnoses were made in 80% (68/85) while discordant diagnoses accounted for 20% (17/85).

The most common correct causes of death were infections in 20% (17/85), respiratory aetiologies in 20% (17/85) and cardiovascular conditions in 16% (14/85) (Table 1).

Table 1
Table 1:
Table 1 Causes of death in assessable autopsies grouped by system

The most common missed COD was an infective aetiology (frequently fungal) occurring in 9% of cases (8/85), of which half were transplant patients. Respiratory and cardiovascular causes each accounted for 4% of cases (3/85). In 12% (10/85) the discordance was a major change from one body system to another (Table 2).15 Examples of discordant diagnoses are shown in Fig. 1–4.

Table 2
Table 2:
Table 2 Comparison of discordant clinical and autopsy causes of death (COD) grouped by system

Even when the mechanism of death was correct, a further 31% (26/85) yielded significant missed diagnoses that were valuable to the understanding of a patient's pathological process. For example, a 66-year-old male renal transplant patient was diagnosed with a ruptured abdominal aortic aneurysm during hospital admission for a surgical wound infection; a CT scan suggested mycotic aneurysm and the autopsy was able to confirm this diagnosis in the setting of disseminated aspergillosis. Another example involved a 52-year-old male whose death was attributed to right middle cerebral artery stroke; at autopsy he was found to have an intracerebral haemorrhage in the setting of undiagnosed intravascular large B cell lymphoma. Overall, at least one unsuspected diagnosis was revealed in 51% of cases.

Fig. 1
Fig. 1:
Fig. 1. Small and large bowel, 80-year-old male (macroscopic image). Clinical cause of death was possible abdominal sepsis and ischaemic heart disease. Autopsy revealed small bowel infarction secondary to sigmoid diverticular adhesion.
Fig. 2
Fig. 2:
Fig. 2. Hippocampus, 42-year-old male (H&E; inset, GMS). Clinical cause of death was liver failure secondary to presumed drug toxicity and acute hydrocephalus. Autopsy revealed intracerebral and pulmonary nocardiosis.
Fig. 3
Fig. 3:
Fig. 3. Lungs, 51-year-old female (H&E; inset, higher power). Clinical cause of death was cardiac amyloidosis in the setting of multiple myeloma and renal failure. Autopsy revealed angioinvasive aspergillosis involving lungs, trachea, oesophagus and stomach. There was no evidence of amyloidosis or visceral multiple myeloma.
Fig. 4
Fig. 4:
Fig. 4. Right pulmonary artery, 45-year-old female (macroscopic image). Clinical cause of death was multi-organ failure in the setting of septic shock, necrotising fasciitis and systemic lupus erythematosis. Autopsy revealed a large right pulmonary artery occlusive pulmonary thromboembolus and multiple organising emboli. There was no evidence of necrotising fasciitis.

Organ transplant recipients accounted for 29% (25/86) of all autopsies, including lung (16 cases), heart (6), renal (2) and bone marrow (1). The percentage of correct COD was similar between the transplant and general populations (77% versus 80%). Transplant patients with discordant diagnoses most commonly had an infectious aetiology (4/6), which accounted for half of discordant infectious diagnoses overall.

Several studies have shown that evaluating discrepancies between clinical and autopsy diagnoses may enhance diagnostic accuracy, improve our knowledge of disease and assist relatives with grieving and knowledge of potentially heritable diseases.1–8 Equally, autopsies may be beneficial when findings are concordant by providing feedback on treatment decisions as well as educating clinicians and medical students.4,6

In our audit we found that clinical cause of death was incorrect in 20% of cases and that at least one unsuspected significant diagnosis was revealed in a further 31%. These rates appear consistent with those previously reported in the literature.1,2,8,10,14–19,21

The types of discordant diagnoses in our study are also consistent with other investigators in which pulmonary conditions and infectious diseases, especially nosocomial and opportunistic infections such as invasive aspergillosis, were found to be among the most common unexpected findings.6,8,10,14,17,19,21 While other studies have found that infections occur often in the intensive care setting, our findings are likely to reflect the high proportion of immunosuppressed transplant patients.10,21

We acknowledge several limitations of our audit and other post-mortem studies.22 Autopsy studies suffer inherent selection bias since post-mortems are frequently only requested in cases where there has been a complicated clinical course or there is a specific question to be answered.19,21 Indeed while request for autopsy at our hospital is looked upon favourably among general medical units, like other studies, acceptance is low and random.19 This is in contrast to autopsy acceptance in the lung transplantation unit (18% of our cases), which is high. Hence our work has selection bias towards complicated patients, similar to other autopsy studies.6,10,12,19

Furthermore, most studies, including our own, are retrospective and deal with small cohorts.2,6,10,17,19,21 There is also subjectivity in attributing concordance between clinical information and autopsy findings.22 We addressed this by heeding previous recommendations to define discrepancies as treatable or untreatable major conditions significantly contributing to death, as distinct from discordant minor co-morbidities.6,8,20,21

Given these caveats, 20% discordance remains a finding of our study, as do unsuspected fungal infections, in those patients who have had an autopsy. These rates and types of discordant diagnoses are similar to those reported in recent literature since 2005 despite advances in many areas of medicine.1,2,6,8,10,14–19,21 Further careful evaluation of discrepancies in autopsy studies should enhance our understanding of disease in the current era.

In conclusion, despite the decline of autopsies in modern medical practice, hospital autopsies may provide treating clinicians with significant and unexpected findings, which highlights their effectiveness in auditing clinical diagnoses. Infection is one of the most common undiagnosed conditions and autopsies have a particularly useful role in cases where the clinical history is complex, such as organ transplantation. We advocate that medical staff should consider the continuing practice of autopsy for select patients within their hospital communities as a valuable adjunct for clinical quality assurance and continuing medical education.


The authors gratefully acknowledge the contribution of deceased patients and consenting relatives in assisting our medical knowledge.

Conflicts of interest and sources of funding: The authors state that there are no conflicts of interest to disclose.


1. Royal College of Pathologists of Australasia Autopsy Working Party.. The decline of the hospital autopsy: a safety and quality issue for healthcare in Australia. Med J Aust 2004; 180:281–285.
2. Tse GM, Lee JC. A 12-month review of autopsies performed at a university-affiliated teaching hospital in Hong Kong. Hong Kong Med J 2000; 6:190–194.
3. Ayoub T, Chow J. The conventional autopsy in modern medicine. J R Soc Med 2008; 101:177–181.
4. Burton JL, Underwood J. Clinical, educational, and epidemiological value of autopsy. Lancet 2007; 369:1471–1480.
5. Carr U, Bowker L, Ball RY. The slow death of the clinical post-mortem examination: implications for clinical audit, diagnostics and medical education. Clin Med 2004; 4:417–423.
6. De Vlieger GYA, Mahieu EM, Meersseman W. Clinical review: What is the role for autopsy in the ICU? Crit Care 2010; 14:221.
7. King LS, Meehan MC. A history of the autopsy. A review. Am J Pathol 1973; 73:514–544.
8. Roulson J, Benbow EW, Hasleton PS. Discrepancies between clinical and autopsy diagnosis and the value of post mortem histology; a meta-analysis and review. Histopathology 2005; 47:551–559.
9. Zhu M, Yu D. Medical progress: fluctuations in the rate of autopsy in China. Chin Med J 2011; 124:3403–3407.
10. Tejerina E, Esteban A, Fernandez-Segoviano P, et al. Clinical diagnoses and autopsy findings: discrepancies in critically ill patients. Crit Care Med 2012; 40:842–846.
11. Shojania K, Burton E, McDonald K, et al. Overestimation of clinical diagnostic performance caused by low necropsy rates. Qual Saf Health Care 2005; 14:408–413.
12. Henry J, Nicholas N. Dead in the water – are we killing the hospital autopsy with poor consent practices? J Royal Soc Med 2012; 105:288–295.
13. Lugli A, Anabitarte M, Beer JH. Effect of simple interventions on necropsy rate when active informed consent is required. Lancet 1999; 354:1391.
14. Sington JD, Cottrell BJ. Analysis of the sensitivity of death certificates in 440 hospital deaths: a comparison with necropsy findings. J Clin Pathol 2002; 55:499–502.
15. Alfsen GC, Maehlen J. The value of autopsies for determining the cause of death. Tidsskr Nor Laegeforen 2012; 132:147–151.
16. Rastan AJ, Gummert JF, Lachmann N, et al. Significant value of autopsy for quality management in cardiac surgery. J Thorac Cardiovasc Surg 2005; 129:1292–1300.
17. Twigg SJ, McCrirrick A, Sanderson PM. A comparison of post mortem findings with post hoc estimated clinical diagnoses of patients who die in a United Kingdom intensive care unit. Intensive Care Med 2001; 27:706–710.
18. McKelvie PA. Medical certification of causes of death in an Australian metropolitan hospital. Comparison with autopsy findings and a critical review. Med J Aust 1993; 158:816–818.820-1.
19. Pinto Carvalho FL, Cordeiro JA, Cury PM. Clinical and pathological disagreement upon the cause of death in a teaching hospital: analysis of 100 autopsy cases in a prospective study. Pathol Int 2008; 58:568–571.
20. Goldman L, Sayson R, Robbins S, et al. The value of the autopsy in three medical eras. N Engl J Med 1983; 308:1000–1005.
21. Winters B, Custer J, Galvagno SM Jr, et al. Diagnostic errors in the intensive care unit: a systematic review of autopsy studies. BMJ Qual Saf 2012; 21:894–902.
22. Saracci R. Is necropsy a valid monitor of clinical diagnosis performance? BMJ 1991; 303:898–900.
© 2015 Royal College of Pathologists of Australasia