Methods

All deaths that occurred at our institution between 1992 and in 2002 were identified from the hospital database. In addition, patients dying in the intensive care unit (ICU) between 1997 and 2002 were identified to provide a cohort of acutely ill patients for comparison with our bone marrow transplant program. Recipients of hematopoietic grafts who died between 1992 and 2002 were identified from the BMT database at our institution. Autopsy lists were provided from the Pathology department and confirmed using the hospital database. Hospital chart numbers were used to crosscheck the lists. All lists were kept in confidential files and patient names were omitted from all research documents. Hospital charts of deceased patients who had an autopsy in 2002 were reviewed to compare clinical diagnosis with pathological diagnosis at the time of death. Differences between clinical and pathological diagnoses or causes of death were classified according to the classification system described by McKelvie and Rode.¹⁴ A 'major' discordance was defined as a difference in the chief diagnosis, which would have led to alternative therapeutic options that may have impacted on clinical outcome, and a 'minor' discordance was assigned if such changes in clinical management were unlikely to affect clinical outcome. Statistical analysis of proportions was performed using the ² test.

Results

The total number of autopsies performed at our hospital, including all clinical services, declined steadily between 1992 and 2002. A total of 208 autopsies were performed out of 870 deaths (24%) in 1992, whereas only 99 autopsies were recorded despite 1094 deaths (9%) in 2002 (P=0.001). A gradual decline in autopsy rates was observed between 1992 and 2002 (Figure 1). Concerning the bone marrow transplant program between the same period, 78 autopsies were performed for 242 total deaths occurring at our hospital (32%). Annual autopsy rates remained relatively stable, ranging from 24 to 46% during the 11-year period (Figure 1). A trend suggesting an increased proportion of autopsies performed on recipients of allogeneic transplants (66 of 193 deaths, 34%) in comparison with autologous transplant patients (12 of 49 deaths, 24%) was observed (P=0.2) (Table 1). Similarly, consent for autopsy may be slightly increased for deaths occurring on weekdays (60 of 173 deaths, 35%) compared with weekend deaths (18 of 69, 26%), although the difference is not statistically important (P=0.2). HSCT recipients who died in the ICU had similar autopsy rates in comparison with deaths occurring on the transplant ward (32% vs 31%, P=1.0) (Table 1). In contrast to the HSCT program between 1997 and 2002 (46 autopsies for 147 deaths, 31%), the percentage of autopsies performed following all deaths in the ICU during the same time period was 23% (283 autopsies for 1199 deaths, P=0.05).

Figure 1.

(a) Autopsy rates at our hospital (HMR) and within the HSCT service, 1992–2002. (b) Number of deaths and autopsies in the HSCT program 1992–2002.

Full figure and legend (27K)

Table 1 - Autopsy rates in the HSCT program 1992–2002.

Full table

Of the 78 autopsies performed between 1992 and 2002, 61 autopsy reports were available for review. A total of 15 charts had been destroyed due to an institutional obligation to retain records for only 10 years following a patient's death and two charts could not be found in medical archives. In comparison with clinical notes and information available in the patient chart, autopsies confirmed major clinical diagnoses and/or suspected cause of death in 45 of 61 autopsies (74%). In accordance with the classification of diagnostic discordance utilized by McKelvie and Rode,¹⁴ autopsies yielded major disagreements in clinical diagnosis or cause of death that may have altered therapy in 10 cases (16%) and uncovered minor disagreements in seven cases (11%) for an overall discordance rate of 27% (Table 2). The diagnostic disagreements are listed in Table 3. Important fungal infections were discovered at autopsy in seven cases, which were not previously suspected and not treated prior to the patient's death. In three cases, relapsed disease or unsuspected cancer was discovered at autopsy.

Table 2 - Audit of HSCT autopsies 1992–2002.

Full table

Table 3 - Audit of autopsies on HSCT recipients.

Full table

Illustrative cases

Two illustrative cases are provided, which highlight the potential benefits of performing an autopsy in recipients of hematopoietic grafts.

Illustrative Case 1: an example of diagnostic discordance

A 55-year-old male was diagnosed with myelodysplasia (refractory anemia with excess blasts) and underwent allogeneic transplantation from his matched sister using a mobilized peripheral blood graft. The patient was admitted to the ICU 72 days after transplantation because of progressive hyperbilirubinemia and upper gastrointestinal bleeding. A liver biopsy performed on day +77 demonstrated parenchymal hepatic cholestasis without evidence of hepatic GVHD, supporting the clinical suspicion of medication-induced cholestasis or veno-occlusive disease (VOD) of the liver. Upper gastrointestinal bleeding was attributed to probable GVHD based on biopsies of the rectosigmoid colon suggestive of GVHD. However, upper endoscopy was not performed. Despite treatment for suspected GVHD and supportive treatment in the ICU, the patient succumbed to progressive deterioration on day +87 and an autopsy was performed. Surprisingly, pathological examination revealed severe mycosis with branching filamentous mycelia due to probable aspergillus in the gastric mucosa and evidence of hepatic GVHD. The patient had not received systemic antifungal therapy in the weeks preceding his death.

Mycotic infections in the gastrointestinal tract are uncommon before day 100 in patients undergoing allogeneic transplantation. The pattern of injury is instructive and the autopsy findings raise awareness of acute infectious causes of acute gastrointestinal bleeding. In addition, the autopsy findings highlight the difficulty in diagnosing the precise cause of hyperbilirubinemia in this patient population.

Illustrative Case 2: a missed opportunity

A 46-year-old man underwent matched unrelated bone marrow transplantation with myeloablative conditioning as treatment for therapy-related acute myeloid leukemia (AML). AML occurred as a result of chemotherapy for lymphoproliferative disease developing after cardiac allografting for end-stage familial cardiomyopathy. At 1 day prior to infusion of the marrow graft, the patient suffered an acute cerebrovascular occlusion in the region of the right middle cerebral artery. Investigations revealed critical stenosis of the right carotid artery. His left-sided paraparesis partly resolved by day +50 after BMT and he regained independent ambulatory function. On day +85, the patient expired from rapidly progressive hypotensive shock and died in the ICU. All cultures were negative. An autopsy was not performed.

Studies from autopsy series in sex-mismatched recipients of hematopoietic grafts have demonstrated that bone marrow-derived cells can migrate to areas of ischemic injury and participate in reparative processes. Few studies have examined this effect in human stroke. Whether cells of the allogeneic marrow donor migrated to the region of cerebral ischemia and facilitated repair of the injured brain in this patient remains unanswered due to the lack of an autopsy and microscopic examination.

Discussion

Several factors likely contribute to the overall decline in autopsy rates in recent decades. Advances in diagnostic testing, most notably in the areas of radiology and molecular biology, have reduced the frequency of uncertain diagnoses at the time of death. However, discordance between clinical and pathologic diagnoses still persists. Three large studies have revealed a constant rate of diagnostic discordance in approximately 35% in cancer patients throughout the 20th century, and undiagnosed or misdiagnosed conditions were frequently attributed to the immediate cause of death.^{3, 15, 16} Our own data suggest that discordance between clinical and pathological causes of death also occurs in recipients of bone marrow transplantation, further underscoring the utility of post-mortem examinations. The impact of more aggressive pre-mortem diagnostic testing on the frequency of discordant diagnoses remains to be determined.

The general decline in autopsies performed between 1992 and 2002 at our hospital is consistent with trends observed worldwide. The current rate of approximately 10% is consistent with other reports.^{1, 17} The approximately 30% autopsy rate on deaths at our hospital in the HSCT program over a 11-year period represents a rare example of preserving high autopsy rates. To the best of our knowledge, there are no previous reports detailing rates of autopsies performed in bone marrow transplantation programs. It should be noted that some patients who received transplants at our center died at other hospitals or at home with palliative support. As the bone marrow transplant service is a referral center, many patients return to the care of referring physicians for follow-up and longer-term care. It is unknown whether these patients underwent autopsy and the factors contributing to decisions concerning autopsy at the referring centers probably deserve attention.

Neonatal autopsy rates are commonly in excess of 50% but may be in decline, according to one report.² Few other specialized services, however, have reported autopsy rates in excess of 30%. Rates of autopsy among ICU deaths at our hospital were also increased relative to the overall hospital rate but to a lesser degree than the HSCT program, perhaps reflecting the number and variety of different clinical services involved in the ICU. Among HSCT patients who died in the ICU, the elevated autopsy rate associated with our service was preserved, suggesting that the involvement of the HSCT service may have facilitated consent for autopsy. The suggestion that day of the week may affect autopsy rates has been reported previously,¹ but was not confirmed in the present study. In some instances, it may be possible that weekend medical staff are less familiar with family members, have less time or are less motivated to obtain consent for an autopsy. Alternatively, fewer autopsies may be requested on weekends due to perceptions regarding its availability. The increased trend in the number of autopsies performed in cases of allogeneic vs autologous transplants may reflect the greater complexity of diagnostic challenges following allogeneic transplantation.

Methods of obtaining consent are often underemphasized in academic centers as the autopsy has become a less common tool for medical education. One group addressed possible mechanisms for improving methods of obtaining consent and reported an increase from 10 to 27% by implementing a formal education tool and requiring that senior residents rather than junior housestaff complete the consent for autopsy at the time of death.¹⁸ Legislation requiring informed consent from family members has likely contributed to increased rates of refusal for autopsies in France and Australia.^{14, 17} On the other hand, one medical center reports preservation of high autopsy rates through emphasizing collaborative interactions between the Department of Pathology and clinical services and by involving residents in regular rotations with the autopsy service.¹⁹ The process for obtaining consent to perform an autopsy at our institution requires that next-of-kin provide informed written consent to a member of the treating medical team, which may include housestaff, the staff physician or the doctor on-call. All residents receive a general orientation at the beginning of their training, but are not obligated to spend time with the autopsy service. Within our hospital, all medical programs are obligated to hold regular morbidity and mortality rounds to review causes of death and to provide feedback for the treating physicians. Information from autopsy findings are frequently presented and discussed at these rounds.

Several factors related to the delivery of health care may influence autopsy rates. Hospital costs and methods of remunerating physicians for conducting autopsies may affect the number of autopsies requested in different jurisdictions. At our institution, pathologists are remunerated through the provincial government's fee-for-service system. In Canada, all hospital-based health-care expenses are paid for by provincial governments. In contrast to health-care systems where autopsies are paid for from hospital budgets or by private insurance plans, no disincentives exist at our institution with respect to ordering or performing post-mortem examinations. As more care is transferred to the outpatient setting, particularly owing to nonmyeloablative transplants, portable intravenous pumps and transfusion clinics, it is unclear as to how this will influence rates of autopsy in the HSCT program. Whether a concomitant increase in the availability of diagnostic testing will parallel the increased trend in outpatient clinical care remains to be seen. Further, as families and caregivers become more comfortable supporting patients at home, more deaths may occur outside the hospital, which may diminish overall rates of autopsy.

Our data suggest that the HSCT service at our hospital continues to regard autopsy as an important diagnostic tool, representing a useful educational and research resource. Despite a global trend of decreasing autopsy rates, it appears feasible to maintain consistently high rates of post-mortem exams within highly specialized programs.

References

References

Acknowledgements

We wish to acknowledge the assistance of Ms Karine Pelletier and Ms Michèle Joseph for their help in searching hospital databases. David Allan is a recipient of an unrestricted BMT Fellowship Award from Fujisawa Canada Inc.