- Email: [email protected]
The Role of Autopsy in the Intensive Care Unit
Mayo Clin Proc, August 2003, Vol 78
Role of Autopsy in the ICU
947
Original Article
The Role of Autopsy in the Intensive Care Unit HASSAN F. NADROUS, MD; BEKELE AFESSA, MD; ERIC A. PFEIFER, MD; AND STEVE G. PETERS, MD
• Objective: To identify the frequency and spectrum of clinically relevant diagnoses found at autopsy but not determined before death in adult patients admitted to an intensive care unit (ICU). • Patients and Methods: We retrospectively reviewed medical records and autopsy reports of patients admitted to ICUs from January 1, 1998, to December 31, 2000. Disagreements between autopsy and antemortem diagnoses were classified as type I or type II errors. A new major diagnosis with potential for directly impacting therapy was considered a type I error. Type II errors included important findings that would not have likely changed therapy. • Results: Of 1597 deaths in all ICUs during the study period, autopsies were performed in 527 patients (33%). Autopsy reports were available in 455 patients, of whom 19
(4%) had type I errors and 78 (17%) had type II errors. The most common type I error was cardiac tamponade. There were no significant differences in age, sex, or length of stay in the ICU or hospital among patients with and without diagnostic errors or among patients with type I and II errors. Seventy-eight patients had 81 type II errors. Organ transplant recipients had more type I or II errors than did nontransplant patients (35% vs 20%; P=.04). • Conclusions: Diagnoses with impact on therapy and outcome are missed in approximately 4% of deaths of adult patients admitted to the ICU. Transplant recipients are especially likely to have occult conditions for which additional therapy might be indicated. Mayo Clin Proc. 2003;78:947-950 BMT = bone marrow transplant; ICU = intensive care unit
T
he autopsy is a valuable and reliable tool in contributing to medical education, patient care, and quality assessment and control.1-3 However, the reported autopsy rate has decreased worldwide during the past 4 decades.2,4-11 The decline has been attributed to a variety of factors, including religious and ethical beliefs, cost, fear of litigation, improved clinical accuracy, reluctance to ask the family for permission, fear of infectious diseases, and exclusion of minimum mandatory autopsy rates as one of the accreditation criteria for US hospitals.12,13 The value of the autopsy has been investigated extensively in hospitalized patients, including those admitted to the intensive care unit (ICU).1,14-17 With recent advances in diagnostic technology and therapeutic interventions, the accuracy of antemortem clinical diagnosis may change, and new diseases and complications may emerge. For example, with the availability of computed tomographic angiography of the chest, one might expect a decrease in the antemortem underdiagnosis of pulmonary embolism. With the increasing number of immunocompromised patients, including bone marrow transplant recipients, a new pattern of complications will likely occur in critically ill patients who die in the ICU. The
objectives of this retrospective study are to describe the recent trends in autopsy findings and to identify the frequency and spectrum of clinically relevant findings not identified before death. PATIENTS AND METHODS We reviewed the medical records, Acute Physiology and Chronic Health Evaluation III database, and autopsy reports of patients treated in adult ICUs of the Mayo Medical Center, Rochester, Minn, from January 1, 1998, through December 31, 2000. The ICUs included the medical ICU, coronary care unit, neurologic ICU, surgical and trauma ICUs at Saint Marys Hospital, and a multidisciplinary ICU at Rochester Methodist Hospital. The medical staff make rounds in these units every day, and medical decisions are supervised and approved by the attending physicians. Housestaff routinely requested consent for autopsy from the relatives of the deceased patients. Patients who did not authorize their records to be reviewed for research were excluded from the study. Paper medical charts and electronic databases were reviewed for each patient, including admission notes, progress notes, discharge summaries, radiography reports, laboratory results, and surgical pathology reports. Data collected included age, sex, length of ICU and hospital stay, and major autopsy and clinical findings. The clinical and pathologic causes of death were obtained from chart reviews and autopsy reports, respectively. The clinical cause of death included the immediate cause and the under-
From the Division of Pulmonary and Critical Care Medicine and Internal Medicine (H.F.N., B.A., S.G.P.) and Department of Laboratory Medicine and Pathology (E.A.P.), Mayo Clinic, Rochester, Minn. Address reprint requests and correspondence to Steve G. Peters, MD, Division of Pulmonary and Critical Care Medicine, Mayo Clinic, 200 First St SW, Rochester, MN 55905 (e-mail: peters.steve@mayo .edu). Mayo Clin Proc. 2003;78:947-950
947
© 2003 Mayo Foundation for Medical Education and Research
For personal use. Mass reproduce only with permission from Mayo Clinic Proceedings.
948
Role of Autopsy in the ICU
Mayo Clin Proc, August 2003, Vol 78
Table 1. Patients With Type I and II Errors in Each Intensive Care Unit* Patients with errors
Intensive care unit
Type I
Type II
Type I and II
Medical Neurologic Mixed Surgical Coronary
7 0 5 3 4
26 3 20 16 13
33 (34) 3 (3) 25 (26) 19 (20) 17 (17)
*Values are number (percentage).
lying primary disease. Autopsy diagnoses included histological and microbiological findings as listed in the final autopsy reports. Agreement of autopsy findings with antemortem diagnoses was noted. Disagreements between autopsy and antemortem diagnoses were classified as type I or type II errors. A new major diagnosis with potential for direct impact on therapy was considered a type I error. Type II errors included major unexpected findings that would not have likely changed therapy. A new diagnosis was classified as a type II error if the patient was receiving appropriate treatment despite the missed diagnosis, if effective treatment was unavailable for the condition, or if the patient refused further treatment or investigation. Minor incidental findings were omitted if they were not related to the main diagnosis and did not contribute to the immediate cause of death. Student t, Mann-Whitney U, and χ2 tests were used for statistical comparisons between groups. P<.05 was considered significant. RESULTS During the study period, 1597 deaths occurred in the ICUs. Autopsies were performed in 527 patients (33%). Patients’ mean ± SD age was 59.6±18.7 years. Autopsy reports were available in 455 adult patients, of whom 19 (4%) had type I errors and 78 (17%) had type II errors (Table 1). The other 72 patients did not meet the inclusion criteria because they Table 2. Type I Errors in Patients in Intensive Care Units Diagnosis
No.
Cardiac tamponade Fungal infection Bacterial infection Gastrointestinal bleeding Perforated bowel Acute myocardial infarction Ruptured abdominal aortic aneurysm Ruptured iliac artery aneurysm Ruptured pulmonary artery Tension pneumothorax Cytomegalovirus pneumonia Pulmonary embolus Ischemic cecum
4 3 3 2 2 1 1 1 1 1 1 1 1
died in the pediatric ICU (46); they died in the chronic ventilator unit (5), which is not an ICU; or they did not authorize their medical charts to be reviewed for research (21). The errors were less frequent in neurologic and surgical ICUs (Table 1). The most common type I error was cardiac tamponade (Table 2). The pericardial tamponades were attributed to malignant pericardial effusion, ventricular rupture after myocardial infarction, pacemaker placement, and blind pericardiocentesis. The main type II errors were infectious complications (19), malignancy (12), and bleeding (11) (Table 3). There were no statistically significant differences in age, sex, or length of ICU or hospital stay among patients with and without diagnostic errors (Table 4) or among patients with type I and II errors. Seventy-eight patients had 81 type II errors. Infections comprised the largest group of missed diagnoses (Table 3), of which fungal organisms were the most common. Malignancies diagnosed as type II errors were bronchogenic carcinoma (3), lymphoma (3), and 1 each of cystic duct adenocarcinoma, stomach adenocarcinoma, sarcoma, renal cell carcinoma, adenocarcinoma of the pancreas, and adenocarcinoma of unknown primary tumor. Type II errors in the diagnosis of hemorrhage included retroperitoneal bleeding (4), hemoperitoneum (3), and 1 case each of intracranial bleeding, aortic dissection, aortic laceration, and ruptured intercostal artery. Type II errors listed as “others” were bronchiolitis obliterans organizing pneumonia, deep venous thrombosis, penetrating stomach ulcer, esophagobronchial fistula, neutropenic enterocolitis, pulmonary veno-occlusive disease, systemic amyloidosis, systemic vasculitis, fat embolism, lamotrigine-induced toxicity, acute myocardial infarction, acute cardiac ischemia, atrioventricular node dysplasia, myocytolysis, and acute hemorrhagic pancreatitis. Transplant recipients had more type I or II errors than did nontransplant patients (35% vs 20%; P=.04). The transplant recipients were younger and had a longer hospital stay (Table 5). Of the 17 type I and II errors in the 11 organ transplant recipients, 7 errors occurred in blood and bone marrow transplant (BMT) recipients. Six of the errors were infections, 5 of which were found in BMT recipients. Of these 5 cases, 3 were autopsy diagnoses of disseminated aspergillosis. Two pulmonary emboli were identified in BMT recipients. Other new diagnoses included ruptured gastric ulcer, pulmonary veno-occlusive disease, acute myocardial infarction, ischemic bowel, and tension pneumothorax. Acute myocardial infarction and tension pneumothorax were classified as type I errors. DISCUSSION In this study, we found that one third of patients who died in the ICU underwent autopsy, and type I or II errors were
For personal use. Mass reproduce only with permission from Mayo Clinic Proceedings.
Mayo Clin Proc, August 2003, Vol 78
found in 21%. There were no significant differences in age, sex, or length of hospital stay among patients with and without errors. Major diagnoses with impact on treatment were missed in 4% of patients. Pericardial tamponade was the most common type I error, occurring in 4 patients. Infections were the most common type II errors. Errors were more likely to occur in transplant recipients than in other patients. Although Roosen et al17 reported autopsies in 93% of their patients, the 33% autopsy rate in our study is comparable to that in other postmortem studies from adult ICUs.14-16,18 The 21% discordance between clinical and postmortem diagnoses is also comparable to other ICUbased studies.14-17 Type I errors were found in 4% of our patients. Although this frequency is lower than that reported in other series, the pattern of new diagnoses is comparable to that in previous reports. Specifically, diagnoses of infection, hemorrhage, and cardiovascular disease were identified that might have led to different therapy before death. Pericardial tamponade was the most common type I error in our study. The diagnosis of pericardial effusion can be difficult, and identification at autopsy only has been noted in other series.17 These findings argue for consideration of cardiac tamponade and early use of echocardiography in the evaluation of unexplained shock states. In our study, infections were the most frequently missed major diagnoses, with a total of 26 infections, of which 13 were fungal organisms. Of these fungal infections, 3 were categorized as type I errors. Previous studies have suggested that transplant recipients may be more prone to missed infections.14,15 Mort and Yeston14 found that 85% of missed diagnoses in a surgical population were of infectious etiology. Although it might be argued that terminal infections, especially disseminated aspergillosis, would not have been treated more effectively by earlier diagnosis, as new therapies emerge,19 it remains important to recognize atypical presentations of infection in immunocompromised patients. Intra-abdominal and retroperitoneal hemorrhage represented the second most common category of missed diagnosis. Although the diagnosis of hemorrhage can be suspected in patients with recent invasive procedures or trauma, the clinical diagnosis is often difficult, bedside ultrasonographic findings are frequently inconclusive, and angiography or computed tomography is often not possible in unstable ICU patients.16,17 In autopsy series before 1970, undiagnosed malignancies were common and represented up to 35% of missed diagnoses.16 Recognition of neoplastic diseases improved with progress in imaging techniques, and the reported frequency of undiagnosed malignancy decreased similarly.6,10,20 Because some techniques may not be feasible in critically ill patients, tumors remain a diagnostic challenge in the ICU setting.15,17 In our series, malig-
Role of Autopsy in the ICU
949
Table 3. Type II Errors in Patients in Intensive Care Units Diagnosis
No.
Infections Cancers Bleeding Pulmonary embolus Ischemic bowel Perforated bowel Myocarditis Others
19 12 11 10 7 3 4 15
nancies found at autopsy, of which lung cancers were most common, were all considered type II errors. Pulmonary embolism has been a commonly reported misdiagnosis in autopsy studies.4,6,14,18 The low frequency of undiagnosed major pulmonary embolism in our study and other recent reports may be due to the availability of computed tomographic angiography and a high index of suspicion of this potentially fatal and treatable condition.6,14,17,18 Acute abdominal complications in the medical ICU may be underdiagnosed and can substantially increase risk of death. In a recent study from our institution, Gajic et al21 found that a delay in surgical intervention and a diagnosis of ischemic bowel were associated with increased mortality. Reasons for delay or failure to diagnosis acute abdomen in patients in the ICU may include altered mental state, narcotic use, immunosuppression, and mechanical ventilation. In our study, 12 cases of acute abdomen were missed, of which 2 episodes of bowel perforation were classified as type I errors. Nearly 20 years ago, Goldman20 reported that the percentage of missed diagnoses had not decreased in the previous 70 years. This observation remains valid, although the spectrum of diseases identified at autopsy has evolved. Advances in medical therapy, including ICU support, organ transplantation, and chemotherapy, may prolong the lives Table 4. Differences Among Patients With and Without Type I or II Errors* Characteristic Mean ± SD age (y) Sex, No. (%) Male Female Median ICU LOS (range) (d) Median hospital LOS (range) (d) ICU LOS, No. (%) ≤2 d >2 d
Error I or II (n=97)
No error (n=358)
62±16.5
58±19
48 (49) 49 (51)
203 (57) 155 (43)
P value .07 .20
4 (1-57)
4 (1-75)
.72
8 (1-64)
6 (1-53)
.19 .6
39 (40) 58 (60)
133 (37) 225 (63)
*Values may not equal 100% because of rounding. ICU = intensive care unit; LOS = length of stay.
For personal use. Mass reproduce only with permission from Mayo Clinic Proceedings.
950
Role of Autopsy in the ICU
Mayo Clin Proc, August 2003, Vol 78
REFERENCES
Table 5. Differences Among Transplant Recipients and Nontransplant Patients* Characteristic Mean ± SD age (y) Sex, No.(%) Male Female Type I or II error, No. (%) Median ICU LOS (range) (d) Median hospital LOS (range) (d) ICU LOS, No. (%) ≤2 d >2 d
Transplant (n=31)
Nontransplant (n=424)
47±12.7
60±18.7
22 (71) 9 (29)
229 (54) 195 (46)
11 (35)
86 (20)
1.
P value
2.
<.001 .06
3. 4.
.04 5.
12 (1-46)
9 (1-75)
.67
18 (1-83)
6 (1-83)
<.001 .53
13 (42) 18 (58)
159 (37) 265 (62)
6. 7. 8.
*Values may not equal 100% because of rounding. ICU = intensive care unit; LOS = length of stay. 9.
of patients and lead to the discovery of new diseases and new complications. Our findings confirm that nosocomial and opportunistic infections are currently the diagnoses most likely to be missed in the ICU, especially in organ transplant recipients.4,14,15,17 Although the age of our patients did not differ between those with and without missed diagnoses, transplant recipients were younger and more likely to have a new diagnosis at autopsy. In several studies, the number of missed diagnoses increased with the age of the patients,22,23 although a higher frequency of missed diagnoses has been observed in patients younger than 40 years.20 Length of stay in the ICU or hospital did not alter the likelihood of new diagnoses at autopsy in our study or in other series.2,13,15,17,24 In 1 study of surgical patients, an ICU stay of more than 48 hours was associated with a higher likelihood of dying of missed nosocomial infections.14 Our study has several limitations. It was retrospective, and autopsy was performed in a minority of deaths. Because the study is limited to 1 medical center, the results may not apply in other patient populations. Despite these limitations, we found that a substantial number of clinically important diagnoses are missed before death. This study confirmed that organ transplant recipients are especially likely to have occult clinical conditions for which additional therapy might be indicated. It also showed that the diagnosis of pericardial tamponade is missed antemortem despite the wide availability of bedside echocardiography. Recognition of the evolution of these patterns is important for the management of all critically ill patients. The autopsy remains the ultimate tool of accountability for clinical evaluation and management. We thank Joanna L. Meyer, Rebecca W. Antony, and Patricia L. Kasey for secretarial assistance.
10. 11. 12. 13. 14. 15.
16.
17. 18. 19.
20. 21. 22. 23. 24.
Fernandez-Segoviano P, Lazaro A, Esteban A, Rubio JM, Iruretagoyena JR. Autopsy as quality assurance in the intensive care unit. Crit Care Med. 1988;16:683-685. Scottolini AG, Weinstein SR. The autopsy in clinical quality control. JAMA. 1983;250:1192-1194. Mollo F, Bertoldo E, Grandi G, Cavallo F. Reliability of death certifications for different types of cancer: an autopsy survey. Pathol Res Pract. 1986;181:442-447. Landefeld CS, Chren MM, Myers A, Geller R, Robbins S, Goldman L. Diagnostic yield of the autopsy in a university hospital and a community hospital. N Engl J Med. 1988;318:1249-1254. Roberts WC. The autopsy: its decline and a suggestion for its revival. N Engl J Med. 1978;299:332-338. Goldman L, Sayson R, Robbins S, Cohn LH, Bettmann M, Weisberg M. The value of the autopsy in three medical eras. N Engl J Med. 1983;308:1000-1005. Welsh TS, Kaplan J. The role of postmortem examination in medical education. Mayo Clin Proc. 1998;73:802-805. Kleiner DE, Emmert-Buck MR, Liotta LA. Necropsy as a research method in the age of molecular pathology. Lancet. 1995;346:945948. Battle RM, Pathak D, Humble CG, et al. Factors influencing discrepancies between premortem and postmortem diagnoses. JAMA. 1987;258:339-344. Stevanovic G, Tucakovic G, Dotlic R, Kanjuh V. Correlation of clinical diagnoses with autopsy findings: a retrospective study of 2,145 consecutive autopsies. Hum Pathol. 1986;17:1225-1230. Pastores SM, Halpern NA. Autopsies in the ICU: we still need them! [editorial]. Crit Care Med. 1999;27:235-236. Landefeld CS, Goldman L. The autopsy in clinical medicine. Mayo Clin Proc. 1989;64:1185-1189. McPhee SJ, Bottles K. Autopsy: moribund art or vital science? Am J Med. 1985;78:107-113. Mort TC, Yeston NS. The relationship of pre mortem diagnoses and post mortem findings in a surgical intensive care unit. Crit Care Med. 1999;27:299-303. Tai DY, El-Bilbeisi H, Tewari S, Mascha EJ, Wiedemann HP, Arroliga AC. A study of consecutive autopsies in a medical ICU: a comparison of clinical cause of death and autopsy diagnosis. Chest. 2001;119:530-536. 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. Roosen J, Frans E, Wilmer A, Knockaert DC, Bobbaers H. Comparison of premortem clinical diagnoses in critically ill patients and subsequent autopsy findings. Mayo Clin Proc. 2000;75:562-567. Blosser SA, Zimmerman HE, Stauffer JL. Do autopsies of critically ill patients reveal important findings that were clinically undetected? Crit Care Med. 1998;26:1332-1336. Herbrecht R, Denning DW, Patterson TF, et al, Invasive Fungal Infections Group of the European Organisation for Research and Treatment of Cancer and the Global Aspergillus Study Group. Voriconazole versus amphotericin B for primary therapy of invasive aspergillosis. N Engl J Med. 2002;347:408-415. Goldman L. Diagnostic advances v the value of the autopsy: 19121980. Arch Pathol Lab Med. 1984;108:501-505. Gajic O, Urrutia LE, Sewani H, Schroeder DR, Cullinane DC, Peters SG. Acute abdomen in the medical intensive care unit. Crit Care Med. 2002;30:1187-1190. Cameron HM, McGoogan E. A prospective study of 1152 hospital autopsies, II: analysis of inaccuracies in clinical diagnoses and their significance. J Pathol. 1981;133:285-300. Veress B, Alafuzoff I. A retrospective analysis of clinical diagnoses and autopsy findings in 3,042 cases during two different time periods. Hum Pathol. 1994;25:140-145. Stambouly JJ, Kahn E, Boxer RA. Correlation between clinical diagnoses and autopsy findings in critically ill children. Pediatrics. 1993;92:248-251.
For personal use. Mass reproduce only with permission from Mayo Clinic Proceedings.
Role of Autopsy in the ICU
947
Original Article
The Role of Autopsy in the Intensive Care Unit HASSAN F. NADROUS, MD; BEKELE AFESSA, MD; ERIC A. PFEIFER, MD; AND STEVE G. PETERS, MD
• Objective: To identify the frequency and spectrum of clinically relevant diagnoses found at autopsy but not determined before death in adult patients admitted to an intensive care unit (ICU). • Patients and Methods: We retrospectively reviewed medical records and autopsy reports of patients admitted to ICUs from January 1, 1998, to December 31, 2000. Disagreements between autopsy and antemortem diagnoses were classified as type I or type II errors. A new major diagnosis with potential for directly impacting therapy was considered a type I error. Type II errors included important findings that would not have likely changed therapy. • Results: Of 1597 deaths in all ICUs during the study period, autopsies were performed in 527 patients (33%). Autopsy reports were available in 455 patients, of whom 19
(4%) had type I errors and 78 (17%) had type II errors. The most common type I error was cardiac tamponade. There were no significant differences in age, sex, or length of stay in the ICU or hospital among patients with and without diagnostic errors or among patients with type I and II errors. Seventy-eight patients had 81 type II errors. Organ transplant recipients had more type I or II errors than did nontransplant patients (35% vs 20%; P=.04). • Conclusions: Diagnoses with impact on therapy and outcome are missed in approximately 4% of deaths of adult patients admitted to the ICU. Transplant recipients are especially likely to have occult conditions for which additional therapy might be indicated. Mayo Clin Proc. 2003;78:947-950 BMT = bone marrow transplant; ICU = intensive care unit
T
he autopsy is a valuable and reliable tool in contributing to medical education, patient care, and quality assessment and control.1-3 However, the reported autopsy rate has decreased worldwide during the past 4 decades.2,4-11 The decline has been attributed to a variety of factors, including religious and ethical beliefs, cost, fear of litigation, improved clinical accuracy, reluctance to ask the family for permission, fear of infectious diseases, and exclusion of minimum mandatory autopsy rates as one of the accreditation criteria for US hospitals.12,13 The value of the autopsy has been investigated extensively in hospitalized patients, including those admitted to the intensive care unit (ICU).1,14-17 With recent advances in diagnostic technology and therapeutic interventions, the accuracy of antemortem clinical diagnosis may change, and new diseases and complications may emerge. For example, with the availability of computed tomographic angiography of the chest, one might expect a decrease in the antemortem underdiagnosis of pulmonary embolism. With the increasing number of immunocompromised patients, including bone marrow transplant recipients, a new pattern of complications will likely occur in critically ill patients who die in the ICU. The
objectives of this retrospective study are to describe the recent trends in autopsy findings and to identify the frequency and spectrum of clinically relevant findings not identified before death. PATIENTS AND METHODS We reviewed the medical records, Acute Physiology and Chronic Health Evaluation III database, and autopsy reports of patients treated in adult ICUs of the Mayo Medical Center, Rochester, Minn, from January 1, 1998, through December 31, 2000. The ICUs included the medical ICU, coronary care unit, neurologic ICU, surgical and trauma ICUs at Saint Marys Hospital, and a multidisciplinary ICU at Rochester Methodist Hospital. The medical staff make rounds in these units every day, and medical decisions are supervised and approved by the attending physicians. Housestaff routinely requested consent for autopsy from the relatives of the deceased patients. Patients who did not authorize their records to be reviewed for research were excluded from the study. Paper medical charts and electronic databases were reviewed for each patient, including admission notes, progress notes, discharge summaries, radiography reports, laboratory results, and surgical pathology reports. Data collected included age, sex, length of ICU and hospital stay, and major autopsy and clinical findings. The clinical and pathologic causes of death were obtained from chart reviews and autopsy reports, respectively. The clinical cause of death included the immediate cause and the under-
From the Division of Pulmonary and Critical Care Medicine and Internal Medicine (H.F.N., B.A., S.G.P.) and Department of Laboratory Medicine and Pathology (E.A.P.), Mayo Clinic, Rochester, Minn. Address reprint requests and correspondence to Steve G. Peters, MD, Division of Pulmonary and Critical Care Medicine, Mayo Clinic, 200 First St SW, Rochester, MN 55905 (e-mail: peters.steve@mayo .edu). Mayo Clin Proc. 2003;78:947-950
947
© 2003 Mayo Foundation for Medical Education and Research
For personal use. Mass reproduce only with permission from Mayo Clinic Proceedings.
948
Role of Autopsy in the ICU
Mayo Clin Proc, August 2003, Vol 78
Table 1. Patients With Type I and II Errors in Each Intensive Care Unit* Patients with errors
Intensive care unit
Type I
Type II
Type I and II
Medical Neurologic Mixed Surgical Coronary
7 0 5 3 4
26 3 20 16 13
33 (34) 3 (3) 25 (26) 19 (20) 17 (17)
*Values are number (percentage).
lying primary disease. Autopsy diagnoses included histological and microbiological findings as listed in the final autopsy reports. Agreement of autopsy findings with antemortem diagnoses was noted. Disagreements between autopsy and antemortem diagnoses were classified as type I or type II errors. A new major diagnosis with potential for direct impact on therapy was considered a type I error. Type II errors included major unexpected findings that would not have likely changed therapy. A new diagnosis was classified as a type II error if the patient was receiving appropriate treatment despite the missed diagnosis, if effective treatment was unavailable for the condition, or if the patient refused further treatment or investigation. Minor incidental findings were omitted if they were not related to the main diagnosis and did not contribute to the immediate cause of death. Student t, Mann-Whitney U, and χ2 tests were used for statistical comparisons between groups. P<.05 was considered significant. RESULTS During the study period, 1597 deaths occurred in the ICUs. Autopsies were performed in 527 patients (33%). Patients’ mean ± SD age was 59.6±18.7 years. Autopsy reports were available in 455 adult patients, of whom 19 (4%) had type I errors and 78 (17%) had type II errors (Table 1). The other 72 patients did not meet the inclusion criteria because they Table 2. Type I Errors in Patients in Intensive Care Units Diagnosis
No.
Cardiac tamponade Fungal infection Bacterial infection Gastrointestinal bleeding Perforated bowel Acute myocardial infarction Ruptured abdominal aortic aneurysm Ruptured iliac artery aneurysm Ruptured pulmonary artery Tension pneumothorax Cytomegalovirus pneumonia Pulmonary embolus Ischemic cecum
4 3 3 2 2 1 1 1 1 1 1 1 1
died in the pediatric ICU (46); they died in the chronic ventilator unit (5), which is not an ICU; or they did not authorize their medical charts to be reviewed for research (21). The errors were less frequent in neurologic and surgical ICUs (Table 1). The most common type I error was cardiac tamponade (Table 2). The pericardial tamponades were attributed to malignant pericardial effusion, ventricular rupture after myocardial infarction, pacemaker placement, and blind pericardiocentesis. The main type II errors were infectious complications (19), malignancy (12), and bleeding (11) (Table 3). There were no statistically significant differences in age, sex, or length of ICU or hospital stay among patients with and without diagnostic errors (Table 4) or among patients with type I and II errors. Seventy-eight patients had 81 type II errors. Infections comprised the largest group of missed diagnoses (Table 3), of which fungal organisms were the most common. Malignancies diagnosed as type II errors were bronchogenic carcinoma (3), lymphoma (3), and 1 each of cystic duct adenocarcinoma, stomach adenocarcinoma, sarcoma, renal cell carcinoma, adenocarcinoma of the pancreas, and adenocarcinoma of unknown primary tumor. Type II errors in the diagnosis of hemorrhage included retroperitoneal bleeding (4), hemoperitoneum (3), and 1 case each of intracranial bleeding, aortic dissection, aortic laceration, and ruptured intercostal artery. Type II errors listed as “others” were bronchiolitis obliterans organizing pneumonia, deep venous thrombosis, penetrating stomach ulcer, esophagobronchial fistula, neutropenic enterocolitis, pulmonary veno-occlusive disease, systemic amyloidosis, systemic vasculitis, fat embolism, lamotrigine-induced toxicity, acute myocardial infarction, acute cardiac ischemia, atrioventricular node dysplasia, myocytolysis, and acute hemorrhagic pancreatitis. Transplant recipients had more type I or II errors than did nontransplant patients (35% vs 20%; P=.04). The transplant recipients were younger and had a longer hospital stay (Table 5). Of the 17 type I and II errors in the 11 organ transplant recipients, 7 errors occurred in blood and bone marrow transplant (BMT) recipients. Six of the errors were infections, 5 of which were found in BMT recipients. Of these 5 cases, 3 were autopsy diagnoses of disseminated aspergillosis. Two pulmonary emboli were identified in BMT recipients. Other new diagnoses included ruptured gastric ulcer, pulmonary veno-occlusive disease, acute myocardial infarction, ischemic bowel, and tension pneumothorax. Acute myocardial infarction and tension pneumothorax were classified as type I errors. DISCUSSION In this study, we found that one third of patients who died in the ICU underwent autopsy, and type I or II errors were
For personal use. Mass reproduce only with permission from Mayo Clinic Proceedings.
Mayo Clin Proc, August 2003, Vol 78
found in 21%. There were no significant differences in age, sex, or length of hospital stay among patients with and without errors. Major diagnoses with impact on treatment were missed in 4% of patients. Pericardial tamponade was the most common type I error, occurring in 4 patients. Infections were the most common type II errors. Errors were more likely to occur in transplant recipients than in other patients. Although Roosen et al17 reported autopsies in 93% of their patients, the 33% autopsy rate in our study is comparable to that in other postmortem studies from adult ICUs.14-16,18 The 21% discordance between clinical and postmortem diagnoses is also comparable to other ICUbased studies.14-17 Type I errors were found in 4% of our patients. Although this frequency is lower than that reported in other series, the pattern of new diagnoses is comparable to that in previous reports. Specifically, diagnoses of infection, hemorrhage, and cardiovascular disease were identified that might have led to different therapy before death. Pericardial tamponade was the most common type I error in our study. The diagnosis of pericardial effusion can be difficult, and identification at autopsy only has been noted in other series.17 These findings argue for consideration of cardiac tamponade and early use of echocardiography in the evaluation of unexplained shock states. In our study, infections were the most frequently missed major diagnoses, with a total of 26 infections, of which 13 were fungal organisms. Of these fungal infections, 3 were categorized as type I errors. Previous studies have suggested that transplant recipients may be more prone to missed infections.14,15 Mort and Yeston14 found that 85% of missed diagnoses in a surgical population were of infectious etiology. Although it might be argued that terminal infections, especially disseminated aspergillosis, would not have been treated more effectively by earlier diagnosis, as new therapies emerge,19 it remains important to recognize atypical presentations of infection in immunocompromised patients. Intra-abdominal and retroperitoneal hemorrhage represented the second most common category of missed diagnosis. Although the diagnosis of hemorrhage can be suspected in patients with recent invasive procedures or trauma, the clinical diagnosis is often difficult, bedside ultrasonographic findings are frequently inconclusive, and angiography or computed tomography is often not possible in unstable ICU patients.16,17 In autopsy series before 1970, undiagnosed malignancies were common and represented up to 35% of missed diagnoses.16 Recognition of neoplastic diseases improved with progress in imaging techniques, and the reported frequency of undiagnosed malignancy decreased similarly.6,10,20 Because some techniques may not be feasible in critically ill patients, tumors remain a diagnostic challenge in the ICU setting.15,17 In our series, malig-
Role of Autopsy in the ICU
949
Table 3. Type II Errors in Patients in Intensive Care Units Diagnosis
No.
Infections Cancers Bleeding Pulmonary embolus Ischemic bowel Perforated bowel Myocarditis Others
19 12 11 10 7 3 4 15
nancies found at autopsy, of which lung cancers were most common, were all considered type II errors. Pulmonary embolism has been a commonly reported misdiagnosis in autopsy studies.4,6,14,18 The low frequency of undiagnosed major pulmonary embolism in our study and other recent reports may be due to the availability of computed tomographic angiography and a high index of suspicion of this potentially fatal and treatable condition.6,14,17,18 Acute abdominal complications in the medical ICU may be underdiagnosed and can substantially increase risk of death. In a recent study from our institution, Gajic et al21 found that a delay in surgical intervention and a diagnosis of ischemic bowel were associated with increased mortality. Reasons for delay or failure to diagnosis acute abdomen in patients in the ICU may include altered mental state, narcotic use, immunosuppression, and mechanical ventilation. In our study, 12 cases of acute abdomen were missed, of which 2 episodes of bowel perforation were classified as type I errors. Nearly 20 years ago, Goldman20 reported that the percentage of missed diagnoses had not decreased in the previous 70 years. This observation remains valid, although the spectrum of diseases identified at autopsy has evolved. Advances in medical therapy, including ICU support, organ transplantation, and chemotherapy, may prolong the lives Table 4. Differences Among Patients With and Without Type I or II Errors* Characteristic Mean ± SD age (y) Sex, No. (%) Male Female Median ICU LOS (range) (d) Median hospital LOS (range) (d) ICU LOS, No. (%) ≤2 d >2 d
Error I or II (n=97)
No error (n=358)
62±16.5
58±19
48 (49) 49 (51)
203 (57) 155 (43)
P value .07 .20
4 (1-57)
4 (1-75)
.72
8 (1-64)
6 (1-53)
.19 .6
39 (40) 58 (60)
133 (37) 225 (63)
*Values may not equal 100% because of rounding. ICU = intensive care unit; LOS = length of stay.
For personal use. Mass reproduce only with permission from Mayo Clinic Proceedings.
950
Role of Autopsy in the ICU
Mayo Clin Proc, August 2003, Vol 78
REFERENCES
Table 5. Differences Among Transplant Recipients and Nontransplant Patients* Characteristic Mean ± SD age (y) Sex, No.(%) Male Female Type I or II error, No. (%) Median ICU LOS (range) (d) Median hospital LOS (range) (d) ICU LOS, No. (%) ≤2 d >2 d
Transplant (n=31)
Nontransplant (n=424)
47±12.7
60±18.7
22 (71) 9 (29)
229 (54) 195 (46)
11 (35)
86 (20)
1.
P value
2.
<.001 .06
3. 4.
.04 5.
12 (1-46)
9 (1-75)
.67
18 (1-83)
6 (1-83)
<.001 .53
13 (42) 18 (58)
159 (37) 265 (62)
6. 7. 8.
*Values may not equal 100% because of rounding. ICU = intensive care unit; LOS = length of stay. 9.
of patients and lead to the discovery of new diseases and new complications. Our findings confirm that nosocomial and opportunistic infections are currently the diagnoses most likely to be missed in the ICU, especially in organ transplant recipients.4,14,15,17 Although the age of our patients did not differ between those with and without missed diagnoses, transplant recipients were younger and more likely to have a new diagnosis at autopsy. In several studies, the number of missed diagnoses increased with the age of the patients,22,23 although a higher frequency of missed diagnoses has been observed in patients younger than 40 years.20 Length of stay in the ICU or hospital did not alter the likelihood of new diagnoses at autopsy in our study or in other series.2,13,15,17,24 In 1 study of surgical patients, an ICU stay of more than 48 hours was associated with a higher likelihood of dying of missed nosocomial infections.14 Our study has several limitations. It was retrospective, and autopsy was performed in a minority of deaths. Because the study is limited to 1 medical center, the results may not apply in other patient populations. Despite these limitations, we found that a substantial number of clinically important diagnoses are missed before death. This study confirmed that organ transplant recipients are especially likely to have occult clinical conditions for which additional therapy might be indicated. It also showed that the diagnosis of pericardial tamponade is missed antemortem despite the wide availability of bedside echocardiography. Recognition of the evolution of these patterns is important for the management of all critically ill patients. The autopsy remains the ultimate tool of accountability for clinical evaluation and management. We thank Joanna L. Meyer, Rebecca W. Antony, and Patricia L. Kasey for secretarial assistance.
10. 11. 12. 13. 14. 15.
16.
17. 18. 19.
20. 21. 22. 23. 24.
Fernandez-Segoviano P, Lazaro A, Esteban A, Rubio JM, Iruretagoyena JR. Autopsy as quality assurance in the intensive care unit. Crit Care Med. 1988;16:683-685. Scottolini AG, Weinstein SR. The autopsy in clinical quality control. JAMA. 1983;250:1192-1194. Mollo F, Bertoldo E, Grandi G, Cavallo F. Reliability of death certifications for different types of cancer: an autopsy survey. Pathol Res Pract. 1986;181:442-447. Landefeld CS, Chren MM, Myers A, Geller R, Robbins S, Goldman L. Diagnostic yield of the autopsy in a university hospital and a community hospital. N Engl J Med. 1988;318:1249-1254. Roberts WC. The autopsy: its decline and a suggestion for its revival. N Engl J Med. 1978;299:332-338. Goldman L, Sayson R, Robbins S, Cohn LH, Bettmann M, Weisberg M. The value of the autopsy in three medical eras. N Engl J Med. 1983;308:1000-1005. Welsh TS, Kaplan J. The role of postmortem examination in medical education. Mayo Clin Proc. 1998;73:802-805. Kleiner DE, Emmert-Buck MR, Liotta LA. Necropsy as a research method in the age of molecular pathology. Lancet. 1995;346:945948. Battle RM, Pathak D, Humble CG, et al. Factors influencing discrepancies between premortem and postmortem diagnoses. JAMA. 1987;258:339-344. Stevanovic G, Tucakovic G, Dotlic R, Kanjuh V. Correlation of clinical diagnoses with autopsy findings: a retrospective study of 2,145 consecutive autopsies. Hum Pathol. 1986;17:1225-1230. Pastores SM, Halpern NA. Autopsies in the ICU: we still need them! [editorial]. Crit Care Med. 1999;27:235-236. Landefeld CS, Goldman L. The autopsy in clinical medicine. Mayo Clin Proc. 1989;64:1185-1189. McPhee SJ, Bottles K. Autopsy: moribund art or vital science? Am J Med. 1985;78:107-113. Mort TC, Yeston NS. The relationship of pre mortem diagnoses and post mortem findings in a surgical intensive care unit. Crit Care Med. 1999;27:299-303. Tai DY, El-Bilbeisi H, Tewari S, Mascha EJ, Wiedemann HP, Arroliga AC. A study of consecutive autopsies in a medical ICU: a comparison of clinical cause of death and autopsy diagnosis. Chest. 2001;119:530-536. 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. Roosen J, Frans E, Wilmer A, Knockaert DC, Bobbaers H. Comparison of premortem clinical diagnoses in critically ill patients and subsequent autopsy findings. Mayo Clin Proc. 2000;75:562-567. Blosser SA, Zimmerman HE, Stauffer JL. Do autopsies of critically ill patients reveal important findings that were clinically undetected? Crit Care Med. 1998;26:1332-1336. Herbrecht R, Denning DW, Patterson TF, et al, Invasive Fungal Infections Group of the European Organisation for Research and Treatment of Cancer and the Global Aspergillus Study Group. Voriconazole versus amphotericin B for primary therapy of invasive aspergillosis. N Engl J Med. 2002;347:408-415. Goldman L. Diagnostic advances v the value of the autopsy: 19121980. Arch Pathol Lab Med. 1984;108:501-505. Gajic O, Urrutia LE, Sewani H, Schroeder DR, Cullinane DC, Peters SG. Acute abdomen in the medical intensive care unit. Crit Care Med. 2002;30:1187-1190. Cameron HM, McGoogan E. A prospective study of 1152 hospital autopsies, II: analysis of inaccuracies in clinical diagnoses and their significance. J Pathol. 1981;133:285-300. Veress B, Alafuzoff I. A retrospective analysis of clinical diagnoses and autopsy findings in 3,042 cases during two different time periods. Hum Pathol. 1994;25:140-145. Stambouly JJ, Kahn E, Boxer RA. Correlation between clinical diagnoses and autopsy findings in critically ill children. Pediatrics. 1993;92:248-251.
For personal use. Mass reproduce only with permission from Mayo Clinic Proceedings.