Diagnostic errors in the intensive care unit: a systematic review of ...

Downloaded from qualitysafety.bmj.com on July 23, 2012 - Published by group.bmj.comBMJ Quality & Safety Online First, published on 21 July 2012 as 10.1136/bmjqs-2012-000803Systematic reviewDiagnostic errors in the intensive careunit: a systematic review of autopsystudiesBradford Winters, 1 Jason Custer, 2 Samuel M Galvagno Jr, 1 Elizabeth Colantuoni, 3Shruti G Kapoor, 1 HeeWon Lee, 4 Victoria Goode, 5 Karen Robinson, 6 Atul Nakhasi, 3Peter Pronovost, 1 David Newman-Toker 7< An additional appendix ispublished online only. Toview this file please visit thejournal online (http://dx.doi.org/10.1136/bmjqs-2012-000803/content/early/recent).1 Department ofAnesthesiology and CriticalCare Medicine, JohnsHopkins University School ofMedicine, Baltimore,Maryland, UK2 Department of Pediatrics,Division of PediatricIntensive Care, University ofMaryland, Baltimore,Maryland, UK3 Department of Biostatistics,Bloomberg School of PublicHealth, Johns HopkinsUniversity, Baltimore,Maryland, UK4 Johns Hopkins UniversitySchool of Medicine,Baltimore, Maryland, UK5 Welch Medical Library,Johns Hopkins UniversitySchool of Medicine,Baltimore, Maryland, UK6 Department of Medicine,Johns Hopkins UniversitySchool of Medicine,Baltimore, Maryland, UK7 Department of Neurology,Johns Hopkins UniversitySchool of Medicine,Baltimore, Maryland, UKCorrespondence toDr Bradford Winters,Department ofAnesthesiology and CriticalCare Medicine, JohnsHopkins University School ofMedicine, 600 N. Wolfe St,Baltimore, MD 21287, UK;bwinters@jhmi.eduAccepted 24 June 2012ABSTRACTContext: Misdiagnoses may be an underappreciatedcause of preventable morbidity and mortality in theintensive care unit (ICU). Their prevalence, nature, andimpact remain largely unknown.Objectives: To determine whether potentially fatal ICUmisdiagnoses would be more common than in thegeneral inpatient population (w5%), and wouldinvolve more infections or vascular events.Data sources: Systematic review of studies identifiedby electronic (MEDLINE, etc.) and manual searches(references in eligible articles) without languagerestriction (1966 through 2011).Study selection and data abstraction: Observationalstudies examining autopsy-confirmed diagnosticerrors in the adult ICU were included. Studiesanalysing misdiagnosis of one specific disease wereexcluded. Study results (autopsy rate, misdiagnosisprevalence, Goldman error class, diseasesmisdiagnosed) were abstracted and descriptivestatistics calculated. We modelled the prevalence ofClass I (potentially lethal) misdiagnoses as anon-linear function of the autopsy rate.Results: Of 276 screened abstracts, 31 studiesdescribing 5863 autopsies (median rate 43%) wereanalysed. The prevalence of misdiagnoses ranged from5.5%e100% with 28% of autopsies reporting at leastone misdiagnosis and 8% identifying a Class Idiagnostic error. The projected prevalence of Class Imisdiagnoses for a hypothetical autopsy rate of 100%was 6.3% (95% CI 4.0% to 7.5%). Vascular eventsand infections were the leading lethal misdiagnoses(41% each). The most common individual Class Imisdiagnoses were PE, MI, pneumonia, andaspergillosis.Conclusions: Our data suggest that as many as 40 500adult patients in an ICU in USA may die with an ICUmisdiagnoses annually. Despite this, diagnostic errorsreceive relatively little attention and research funding.Future studies should seek to prospectively measurethe prevalence and impact of diagnostic errors andpotential strategies to reduce them.INTRODUCTIONDiagnostic errors are an underappreciatedcause of preventable mortality and morbidityin hospitalised patients, accounting for anestimated 40 000e80 000 annual deaths inUSA. 1 Patients in an intensive care unit(ICU) may be especially prone to suffer harmfrom diagnostic errors. They have limitedreserve, often require fast diagnosis andtreatment, are cared for by multiple cliniciansand undergo frequent laboratory andimaging evaluations. 2 Harm may result fromdelay or failure to treat the correct underlyingdisease, complications of unnecessarydiagnostic testing or treating a condition thatis not actually present. 1Diagnostic errors are missed, wrong ordelayed diagnoses that are detected by somesubsequent definitive test or finding. 3Autopsy-detected misdiagnoses are typicallyclassified by the Goldman criteria and basedon their clinical relevance and the potentialthat timely therapy would have changed theoutcome. The Goldman criteria for autopsyconfirmed discrepancies stratify misdiagnosiserrors as major (Class I and II misdiagnosis)and minor (Class III and IV misdiagnosis). AClass I error is a major missed diagnosis witha potential adverse impact on survival (likelywas the cause of death or contributed to thepatient’s death) and had that diagnosis beenknown would have changed managementand outcome. A Class II error is a majormissed diagnosis that did not have an impacton survival (was not the cause of death nordirectly contributed to the patient’s death). AClass III error is a minor missed diagnosisrelated to a terminal disease but not relatedto the cause of death, and Class IV errors areother missed minor discrepancies. 4WintersCopyrightB, CusterArticle J, GalvagnoauthorSM, et al.(orBMJtheirQualemployer)Saf (2012). doi:10.1136/bmjqs-2012-0008032012. Produced by BMJ Publishing Group Ltd under licence. 1e9

Downloaded from qualitysafety.bmj.com on July 23, 2012 - Published by group.bmj.comSystematic reviewwould not have changed therapy. Class III and Class IVare minor errors related to the terminal disease but notdeath and completely unrelated to either, respectively.7e16 18e32 34e37The majority of studies reported Class Iand II 7 9 10 12 13 15 16 18e32 34e36 error data, and9 10 12 13 15 16 18e20 22 26 30 31less reported Class IIIand IV 9121316182023 error data.We defined four broad disease categories to assess thedistribution of types of misdiagnoses. These categoriesincluded: (1) vascular (myocardial infarction, pulmonaryembolus, stroke, any organ infarct, any vesselrupture or dissection, and any haemorrhage), (2)infectious (bacterial, aspergillosis, other fungal, viral,and tuberculosis), (3) mechanical pathophysiologicalstates (cardiac tamponade, bowel obstruction, perforatedviscus, congestive heart failure, and cardiac valvefailure), and (4) neoplastic/other. Some studies treatedpneumonia and aspergillosis separately from otherinfections. 10 14 19 20 25 34 We aggregated these resultsduring the analytic phase.Data were tabulated using Microsoft Excel 2008 forMac (Microsoft Corp., Redmond, Washington, USA).Descriptive statistics were calculated using Stata V.11(Stata Corp., College Station). A p value

Systematic reviewDownloaded from qualitysafety.bmj.com on July 23, 2012 - Published by group.bmj.comTable 1Included studies of intensive care unit (ICU) misdiagnoses in generalAuthor Year Country Type of studyLength ofstudy period(months)ICU typeTotal # ofdeathsNo ofautopsiesused foranalysisBerlot 7 1999 Italy Retrospective 36 Mixed 346 159Blosser 8 1998 USA Retrospective 12 Med/132 41coronaryCombes 9 2004 France Prospective 36 Med/surg 315 167Dimopolous 10 2004 Belgium Retrospective 12 Med/surg 489 222Duke 11 1999 Australia Retrospective 24 Not given 238 Not givenFernandez-Segoviano 12 1988 Spain Prospective 30 Mixed Not given 100Fish 13 2000 USA Retrospective 72 Burn 94 88Gerain 14 1990 Belgium Retrospective 11 Oncology 48 34Gut 15 1999 Brazil Retrospective 36 Med/surg 152 30Kallinen 16 2008 Finland Retrospective 72 Burn 74 71Koch 17 2008 Germany Retrospective 120 Medical 1205 1205Magret 18 2006 Spain Retrospective 46 Mixed 525 80Maris 19 2007 Belgium Retrospective 24 Med/surg 786 289Mort 20 1999 USA Retrospective 72 Surg 560 149Nadrous 21 2003 USA Retrospective 24 Multiple 1597 455Ong 22 2002 USA Retrospective 24 Trauma/burn 158 153Palazon-Sanchez 23 1999 Spain Retrospective 12 Med/surg 67 24Papadakis 24 1991 USA Retrospective 24 Medical 401 172Pastores 25 2007 USA Retrospective 69 Med/surg 658 86Perkins 26 2003 Great Britain Retrospective 42 Med/surg 636 38Podbregar 27 2001 Slovenia Retrospective 24 Medical 270 126Podbregar 28 2011 Slovenia Retrospective 20 Medical 373 170Roosen 29 2000 Belgium Retrospective 24 Medical 108 100Saad 30 2007 Brazil Retrospective 24 Coronary Unclear 161Sharma 31 2005 India Retrospective 36 Trauma/burn 249 163Silfvast 32 2003 Finland Retrospective 48 Mixed 388 346Simon 33 2001 Hungary Retrospective 12 Medical 163 110Tai 34 2001 USA Retrospective 24 Medical 401 91Tejerina 2010 Spain Retrospective 300 Mixed 2857 866Twigg 36 2001 Great Britain Retrospective 36 Med/surg 252 97Yalamarthi 37 1998 Great Britain Retrospective 24 Not given 233 70ICU, Intensive Care Unit.potentially lethal), and 15% were Class II (major but notlethal) despite presumably aggressive diagnostic assessmentin the critical care environment. Given the 540 000deaths annually during or immediately following an ICUstay, 48 and extrapolating to a 100% autopsy rate, ourfindings suggest that 34 000 (95% CI ¼22 600 to 40 500)ICU patients in the US may die as a result of a Class Ierror annually, assuming that the error was the cause ofdeath. This predicted number is comparable to the estimateddeaths from central line-associated bloodstreaminfections in the ICU 49 and from breast cancer 50 in theUS, and does not consider misdiagnosis-related morbidityamong intensive care patients. These results also suggestthat despite the dropping autopsy rate over time, autopsyremains a crucial asset especially in ICU patients.We proved our hypothesis that infections and vascularevents would culminate a large majority of the diagnosticerrors. Vascular events and infections were the leadingpotentially lethal misdiagnoses accounting for almost82% of potentially lethal Class I errors and almost 76%of the serious but not lethal Class II errors. Neoplasticdisease was the third most common overall misdiagnosiscategory but was rarely deemed lethal. Mechanicalpathophysiological states were uncommon. Importantly,pulmonary embolus, myocardial infarction, pneumonia,and aspergillosis accounted for over one-third of all ClassI misdiagnoses. These four particular diagnostic errorsmay be useful as outcome measure for designing earlyinterventions to reduce diagnostic errors as a whole.Our results add new information to the field of diagnosticerror research. Shojania et al 5 conducteda systematic review of autopsy-detected diagnostic errorprevalence rates over time, but they did not describespecific misdiagnoses in the ICU or elsewhere. Weincluded the seven ICU studies from their review in ouranalysis to offer refined estimates of the prevalence of4e9Winters B, Custer J, Galvagno SM, et al. BMJ Qual Saf (2012). doi:10.1136/bmjqs-2012-000803

Downloaded from qualitysafety.bmj.com on July 23, 2012 - Published by group.bmj.comSystematic reviewTable 2 Condition/disease specific misdiagnosis studies not analysedDisease ordiagnosisNo ofmisseddiagnosesNo ofautopsiesanalysedTotal noof deathsStudyperiod(mos) ICU typeAuthor Year Country Type of studydeHemptinne 38 2009 Belgium Retrospective 36 Med/surg 169 64 15 ARDS onlyDu 39 1996 China Retrospective 492 Surg Not given 3447 80 Fungal onlyHasani 40 2005 Iran Prospective 12 Mixed Not given 43 2 Meningitis onlyJanson 41 1976 Germany Retrospective 24 Surg 100 100 48 CXR misDx onlyPetersen 43 1999 Denmark Prospective 12 Multiple 659 141 70 Pneumonia onlyPinheiro 44 2007 Brazil Retrospective 96 Not given 592 22 7 ARDS onlyAl-Saidi 42 2002 Canada Retrospective 67 Med/surg 59 28 10 Bone marrowtransplant onlyTejerina 45 2009 Spain Retrospective 150 Not given 1633 253 111 VAP onlyRetrospective 120 Coronary Not given 31 12 PE onlyWidimsky 46 1985 Czech-SlovakiaARDS, Acute Respiratory Distress syndrome; CXR, chest X-ray; ICU, Intensive Care Unit; misDx, misdiagnosis; mos, months; PE, pulmonary embolus; VAP, ventilator-associated pneumonia.misdiagnoses in the ICU by error class and autopsy rate.Shojania et al projected that, despite a reduction of 33%in the rate of class I errors over 4 decades, a US hospitalwould likely continue to observe a class I error at a rateof 4.1%e6.7% across all patients. While a few of thestudies in Shojania’s review included ICU patients inaddition to general ward and emergency departmentpatients, our results suggest that the rate of class I errorsare 16%e48% higher in the ICU patient populationthen the general hospital patient population as a whole.Methodologically, we found that autopsy rates of 30%e40% or higher are likely to produce fairly accurate estimatesof the overall prevalence of potentially lethalmisdiagnoses in the ICU population, despite potentialbiases inherent in autopsy case selection. 51Observed diagnostic error frequencies were highest instudies with the lowest autopsy rates. Clinical selection ofthe most complex cases to undergo autopsy is believed tocontribute to this effect. 51 To counter this effect, we havemodelled the measured error frequency as a function ofautopsy rate, controlled for study size and time andextrapolated out to conditions of 100% autopsy. Nevertheless,it is possible that other factors may contribute tothe difference seen in studies with low autopsy rates suchas a more detailed and rigorous approach to autopsy anderror finding. Additionally, the level of diagnosticcertainty as to the cause of death is likely to influence theautopsy rate. Podbregar et al 27 attempted to measure thisinfluence finding that diagnostic certainty significantlyinfluenced the decision to pursue an autopsy thoughthere were no differences in the class I error rate. Asa consequence, our extrapolated estimate for the totalerror frequency likely represents a conservative,minimum estimate of the true frequency as aggregatedfrom the present data.Although we could not measure morbidity associatedwith misdiagnosis in survivors, the consequences andcosts are likely substantial. ICU-based studies havereported complications of misdiagnosis, such asre-operation, 52 and increased length of stay. 53 The mostfrequently reported non-lethal long-term consequencesof misdiagnosis are neurological. Preventable or treatableneurological complications of critical illness such asstroke may go unrecognised with devastatingconsequences. 54e56 Robust estimates of the morbidconsequences of diagnostic errors will likely requireprospective studies.Risk predictors for misdiagnosis in the ICU at thepatient, organisational or provider level are incompletelydefined and there is sparse data on root causes ofmisdiagnosis in the ICU. Off-hours occurrence appearsto be a risk factor for at least some types of ICU misdiagnosis.53 Patient demographics and institutional factorshave been cited as possible influences. 5 57 58 There isWinters B, Custer J, Galvagno SM, et al. BMJ Qual Saf (2012). doi:10.1136/bmjqs-2012-0008035e9

Systematic reviewDownloaded from qualitysafety.bmj.com on July 23, 2012 - Published by group.bmj.comFigure 2 This figure shows the proportion of all autopsies thatidentified a misdiagnosis, and the proportion of autopsies thatidentified a misdiagnosis stratified by the Goldman criterion(Class I, II, III, and IV). The grey bars represent identifiedmisdiagnosis and the black bars represent all autopsiesreported for total and class. Of the 30 studies included in thisanalysis, 26 reported class I error data, 24 class II, 13 class III,and 7 class IV error data. Thus, the sample sizes by Goldmanclassification varied. Stratified by class, 8% of autopsiesdetected a Class I error, 15% detected a Class II error, 15%detected a Class III error, and 21% detected a Class IV error.evidence that the patient’s clinical presentation such asgreater co-morbidity or illness complexity increases therisk of misdiagnosis. 52 53 56 59 The diagnostic process iscomplex 60 and errors may occur at any stage during theprocess, often at multiple stages. 61 Some misdiagnosesmay be directly linked to limited sensitivity and specificityof individual tests in the critical care setting, 62 butremediable causes are most often associated with failuresFigure 3 This figure illustrates the prevalence of Class Imisdiagnoses (as a per cent of autopsies) and the autopsyrate. Each point on the figure represents data from one study.The plotting symbols are proportional to the number ofautopsies performed in each study (with larger plotting symbolsindicating larger number of autopsies). The solid linerepresents the logistic regression models estimate for reportedvalues. The dotted lines represent the 95% CIs. The curve wasgenerated using a natural spline with two degrees of freedom.in information processing. Cognitive errors are alsoa major contributor to misdiagnoses and a recent reviewby Graber et al 63 identified a large number of possibleapproaches (some tested and some not) that may helpreduce cognitive errors and hence, misdiagnoses.Figure 4 This figure shows themost common misdiagnoses bybroad disease categories for ClassI and II errors. These categoriesare vascular (myocardialinfarction, pulmonary embolus,stroke, any organ infarct, anyvessel rupture or dissection, andany haemorrhage), infectious(bacterial, aspergillosis, otherfungal, viral and tuberculosis),mechanical pathophysiologicalstates (cardiac tamponade, bowelobstruction, perforated viscus,congestive heart failure andcardiac valve failure), andneoplasm and other (includesacute respiratory distresssyndrome (ARDS)). The numberof Class I (black bar) and Class II(grey bar) errors are compared for each diagnostic category. The data for Class I errors represents 24 of the 30 studies in theanalysis and data for Class II errors represents 16 of 30 studies.6e9Winters B, Custer J, Galvagno SM, et al. BMJ Qual Saf (2012). doi:10.1136/bmjqs-2012-000803

Downloaded from qualitysafety.bmj.com on July 23, 2012 - Published by group.bmj.comSystematic reviewBox 1 Breakdown of misdiagnoses by final diagnosisreportedVascular eventsHaemorrhageVascular ruptureAortic aneurysm/dissection (11, 1)Other (2, 1)Abdominal bleed (21, 41)Other (5, 12)ThrombosisPulmonary embolus (37, 64)Other (1, 0)IschemiaMyocardial infarction (31, 33)Bowel infarction (13, 10)Stroke (1, 0)InfectionsMechanical obstructive pathophysiologyCardiac tamponade (10, 0)Perforated viscus not designated peritonitis (9, 4)Bowel obstruction (1, 4)Tension pneumothorax (1, 0)Congestive heart failure (2, 2)Cardiac valvular lesion (not endocarditis) (5, 2)OtherMalignancy (6, 47)Pulmonary oedema (not acute respiratory distresssyndrome) (2, 2)Acute respiratory distress syndrome (3, 1)Pancreatitis (4, 1)Cirrhosis (0, 1)Airway obstruction (2, 0)Other (9, 22)Values in parentheses are the actual total number of class Iand class II identified in studies reporting this data.A major impediment to better understanding ofdiagnostic errors is the declining autopsy rate. 5 Thereasons for this are multi-factorial and likely includemedical culture and societal attitude shifts. One alternativethat might improve the rate of autopsy is the‘virtual autopsy’ using sophisticated radiological techniquesas opposed to dissection. Wichmann et al, 64recently reported a comparison of ‘virtual autopsy’ totraditional autopsy in ICU patients suggesting that itcompares favourably and had a higher acceptance ratein terms of per cent of deaths referred to post-mortemanalysis. With further refinement this approach mayoffer a way to continue to obtain the post-morteminformation required.Our systematic review has significant policy implications.Diagnostic errors have not been the focus ofpatient safety campaigns or systematic studies. 1 65 A 2003report of 93 Agency for Health Research and Quality(AHRQ)-funded safety projects found only one focusedon misdiagnosis. 65 A recent Special Emphasis Notice hassought proposals for these errors in ambulatory andemergency department settings, 66 but it has not beenlinked to a funded request for applications.Our review has limitations. We did not search the greyliterature nor did we follow-up with authors, pursuesecondary references or consult with experts and publicationbias could, in theory, have inflated our measuredprevalence. Virtually all studies were retrospective innature and there is potential for selection bias in thoseautopsied though Podbregar et al 27 found that whilediagnostic certainty did influence the rate of autopsy, itdid not influence the rate of class I errors. Unfortunately,the reliability of determining whether an erroroccurred and its severity could not be determined to anymeaningful extent in the reviewed studies. Thus ourestimate of the class I (lethal) diagnostic error frequencymay either be an over- or under-estimate of the truefrequency. Debate continues on the extent to which thisis a potential problem. 11 There was some heterogeneityin the study populations included in our review and themeasured misdiagnosis prevalence varied even at highautopsy rates, so misdiagnosis prevalence might not beevenly distributed. We did not assign a quality rating tothe studies analysed based on autopsy confirmed diagnosisbeing a ‘gold standard’. Autopsy, as a ‘gold standard’,is potentially flawed as the decision to perform anautopsy or the results of the autopsy may be influencedby the degree of involvement of clinicians in the judgementof diagnostic errors, the degree to which pathologistsknew the clinical history and working diagnoses atthe time that they performed autopsy and the format/context in which requests for autopsy were transmitted.The original study authors may have misclassified thetypes of diagnostic errors, resulting in inaccurate estimatesof lethality or preventability and disease frequencyestimates might be distorted. Nevertheless, given theconsistency with which studies reported missed infectionsand vascular events as a common source of lethaldiagnostic errors, these are likely important.In conclusion, this systematic review and analysissuggests that between 22 600e40 500 ICU patients dieeach year in the US with and potentially from a diagnosticerror and many more suffer a clinically relevantdiagnostic error. These data suggest that diagnosticerrors are a significant cause of preventable harm inhospitalised patients. To this point, diagnostic errorshave received relatively little attention and researchfunding, leaving the methods to measure them immature;this must change. Future studies should seek toprospectively measure the prevalence and direct impactof diagnostic errors, including those that are non-lethal,and potential strategies to reduce them.Acknowledgements The authors wish to thank Christine G. Holzmueller, BLAfor her assistance in editing the manuscript.Winters B, Custer J, Galvagno SM, et al. BMJ Qual Saf (2012). doi:10.1136/bmjqs-2012-0008037e9

Systematic reviewDownloaded from qualitysafety.bmj.com on July 23, 2012 - Published by group.bmj.comContributors Conception and design: BW, KR, DNT, PP. Acquisition of data:BW, VG, SMG, KR, JC, SGK, AN. Analysis and interpretation of data: BW, JC,HWL, EC, SMG, AN, DNT, PP. Drafting of the manuscript: BW, JC, DNT, PP.Critical revision of manuscript: BW, JC HWL, SMG, AN, DNT PP. Statisticalanalysis: BW, EC, SMG, AN. Administrative, technical, or material support: VG,KR, PP. Supervision: BW, DNT, PP.Funding SMG was supported by an NIH training grant awarded to the JohnsHopkins University School of Medicine; DN-T was supported in part by a grantfrom the Agency for Healthcare Research and Quality (HS017755-01).Competing interests BWinters reports fees for expert testimony at severaldefense and plaintiff law firms and speaking honoraria from 3M corporationand various health systems to speak on patient safety and Rapid ResponseSystems. PP reports receiving grant or contract support from the Agency forHealthcare Research and Quality, the National Institutes of Health, the RobertWood Johnson Foundation and The Commonwealth Fund for research relatedto measuring and improving patient safety; honoraria from various hospitalsand healthcare systems and the Leigh Bureau to speak on quality and safety;consultancy with the Association for Professionals in Infection Control andEpidemiology, Inc.; and book royalties for authoring “Safe Patients, SmartHospitals: How One Doctor’s Checklist Can Help Us Change Health Care fromthe Inside Out.” The following authors report no conflicts of interest: JC, EC,SGK, HWL, VG, KR, AN; SMG and DN-T have no further conflicts beyond whatis reported under funding.Provenance and peer review Not commissioned; externally peer reviewed.REFERENCES1. Newman-Toker DE, Pronovost PJ. Diagnostic errorsdthe nextfrontier for patient safety. JAMA 2009;301:1060e2.2. Newman-Toker DE, Pham JC, Winters BD, et al. Diagnostic errors incritical care settingsdmanaging information overload. ICUManagement 2009;9:6e11.3. Graber ML, Franklin N, Ruthanna G. Diagnostic error in internalmedicine. Arch Intern Med 2005;165:1493e9.4. Goldman L, Sayson R, Robbins S, et al. The value of autopsy in threemedical eras. N Engl J Med 1983;308:1000e5.5. Shojania KG, Burton EC, McDonald KM, et al. Changes in rates ofautopsy-detected diagnostic errors over time: a systematic review.JAMA 2003;289:2849e56.6. Landrigan CP, Parry GJ, Bones CB, et al. Temporal trends in rates ofpatient harm resulting from medical care. N Engl J Med2010;363:2124e34.7. Berlot G, Dezzoni R, Viviani M, et al. Does the length of stayin the intensive care unit influence the diagnostic accuracy?A clinical-pathological study. Eur J Emerg Med1999;6:227e31.8. Blosser SA, Zimmerman HE, Stauffer JL. Do autopsies of critically illpatients reveal important findings that were clinically undetected? CritCare Med 1998;26:1332e6.9. Combes A, Mokhtari M, Couvelard A, et al. Clinical and autopsydiagnoses in the intensive care unit: a prospective study. Arch InternMed 2004;164:389e92.10. Dimopoulos G, Piagnerelli M, Berré J, et al. Post mortem examinationin the intensive care unit: still useful? Intensive Care Med2004;30:2080e5.11. Duke GJ, Morley PT, Cooper DJ, et al. Management of severetrauma in intensive care units and surgical wards. Med J Aust1999;170:416e19.12. Fernandez-Segoviano P, Lázaro A, Esteban A, et al. Autopsy asquality assurance in the intensive care unit. Crit Care Med1988;16:683e5.13. Fish J, Hartshorne N, Reay D, et al. The role of autopsy on patientswith burns. J Burn Care Rehabil2000;21:339e44.14. Gerain J, Sculier JP, Malengreaux A, et al. Causes of deaths in anoncologic intensive care unit: a clinical and pathological study of 34autopsies. Eur J Cancer 1990;26:377e81.15. Gut AL, Ferreira AL, Montenegro MR. Autopsy: quality assurance inthe ICU. Intensive Care Med 1999;25:360e3.16. Kallinen O, Partanen TA, Maisniemi K, et al. Comparison ofpremortem clinical diagnosis and autopsy findings in patients withburns. Burns 2008;34:595e602.17. Koch S, Bredahl J, Wirtz SP, et al. How often is bronchopneumoniaoverlooked as the cause of death in intensive care unit patients?(In German) Anaesthetist 2008;57:37e42.18. Magret Iglesias M, Vidaur Tello L, Fernández Olsina S, et al.Discrepancies between clinical and pathological diagnosis ina polyvalent intensive care service (In Spanish). Med Intensiva2006;30:95e100.19. Maris C, Martin B, Creteur J, et al. Comparison of clinical and postmortemfindings in intensive care unit patients. Virchows Arch2007;450:329e33.20. Mort TC, Yeston NS. The relationship of pre mortem diagnoses andpost mortem findings in a surgical intensive care unit. Crit Care Med1999;27:299e303.21. Nadrous HF, Afessa B, Pfeifer EA, et al. The role of autopsy in theintensive care unit. Mayo Clin Proc 2003;78:947e50.22. Ong AW, Cohn SM, Cohn KA, et al. Unexpected findings in traumapatients dying in the intensive care unit: results of 153 consecutiveautopsies. J Am Coll Surg 2002;194:401e6.23. Palazon-Sanchez C, Palazon-Sanchez E, Pardo Talavera JC, et al.Findings on necropsy studies in critically ill patients. Med Intensiva1999;23:6e9.24. Papadakis MA, Mangione CM, Lee KK, et al. Treatable abdominalpathologic conditions and unsuspected malignant neoplasms atautopsy in veterans who received mechanical ventilation. JAMA1991;265:885e7.25. Pastores SM, Dulu A, Voigt L, et al. Premortem clinical diagnoses andpostmortem autopsy findings: discrepancies in critically ill cancerpatients. Crit Care 2007;11:R48e54.26. Perkins GD, McAuley DF, Davies S, et al. Discrepancies betweenclinical and postmortem diagnoses in critically ill patients: anobservational study. Crit Care 2003;7:R129e32.27. Podbregar M, Voga G, Krivec B, et al. Should we confirm our clinicaldiagnostic certainty by autopsies? Intensive Care Med2001;27:1750e5.28. Podbregar M, Kralj E, Cicak R, et al. A triad algorithm for analysingindividual ante- and post-mortem findings to improve the quality ofintensive care. Anaesth Intensive Care 2011;39:1086e92.29. Roosen J, Frans E, Wilmer A, et al. Comparison of premortem clinicaldiagnoses in critically iII patients and subsequent autopsy findings.Mayo Clin Proc 2000;75:562e7.30. Saad R, Yamada AT, Pereira da Rosa FH, et al. Comparisonbetween clinical and autopsy diagnoses in a cardiology hospital.Heart 2007;93:1414e19.31. Sharma BR, Gupta M, Harish D, et al. Missed diagnoses in traumapatients vis-à-vis significance of autopsy. Injury 2005;36:976e83.32. Silfvast T, Takkunen O, Kolho E, et al. Characteristics ofdiscrepancies between clinical and autopsy diagnoses in theintensive care unit: a 5-year review. Intensive Care Med2003;29:321e4.33. Simon J, Rudas L, Iványi B. The autopsy in verifying diagnosticaccuracy at the intensive care unit (In Hungarian). Orv Hetil2001;142:2373e6.34. Tai DY, El-Bilbeisi H, Tewari S, et al. A study of consecutiveautopsies in a medical ICU: a comparison of clinical cause of deathand autopsy diagnosis. Chest 2001;119:530e6.35. Tejerina E, Esteban A, Fernández-Segoviano P, et al. Clinicaldiagnoses and autopsy findings; discrepancies in critically ill patients.Crit Care Med 2012;40:842e6.36. Twigg SJ, McCrirrick A, Sanderson PM. A comparison of post mortemfindings with post hoc estimated clinical diagnoses of patients who diein a United Kingdom intensive care unit. Intensive Care Med2001;27:706e10.37. Yalamarthi S, Ridley S, Barker T. Agreement between ante-mortemdiagnoses, death certificates and post-mortem causes of death incritically ill patients. Clin Intensive Care 1998;9:100e4.38. deHemptinne Q, Remmelink M, Brimioulle S, et al. ARDS:a clinicopathological confrontation. Chest 2009;135:944e9.39. Du B, Zhang H, Chen D. Invasive fungal infection in 3447 autopsycases Natl Med J China 1996;76:352e4.40. Hasani M, Sahraian MA, Motamedi M, et al. Postmortemcerebrospinal fluid analysis in a general intensive care unit. Indian.J Crit Care 2005;9:176e8.41. Janson R, Thelen M, Rommelsheim K, et al. A review of the findingson chest examinations and at autopsy in surgical patients underintensive care. Fortschr Rongenstr 1976;124:427e33.42. Al-Saidi F, Diaz-Granados N, Messner H, et al. Relationship betweenpremortem and postmortem diagnosis in critically ill bone marrowtransplantation patients. Crit Care Med 2002;30:570e3.43. Petersen IS, Aru A, Skødt V, et al. Evaluation of pneumoniadiagnosis in intensive care patients. Scand J Infect Dis1999;31:299e303.44. Pinheiro BV, Mruaoka FS, Assis RV, et al. Accuracy of clinicaldiagnosis of acute respiratory distress syndrome in comparison withautopsy findings. J Bras Pneumol 2007;33:423e8.8e9Winters B, Custer J, Galvagno SM, et al. BMJ Qual Saf (2012). doi:10.1136/bmjqs-2012-000803

Downloaded from qualitysafety.bmj.com on July 23, 2012 - Published by group.bmj.comSystematic review45. Tejerina E, Esteban A, Fernández-Segoviano P, et al. Accuracy ofclinical definitions of ventilator associated pneumonia: comparisonwith autopsy findings. J Crit Care 2009;25:62e8.46. Widimský J, Stanek V. Clinical diagnosis of haemodynamicallysignificant pulmonary embolism in a coronary care unit. Cor Vasa1985;27:337e45.47. Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis ofobservational studies in epidemiology: a proposal for reporting. MetaanalysisOf Observational Studies in Epidemiology (MOOSE) group.JAMA 2000;283:2008e12.48. Angus DC, Black N. Improving care of the critically ill: institutional andhealth-care system approaches. Lancet 2004;363:1314e20.49. Pronovost P, Needham D, Berenholtz S, et al. An intervention todecrease catheter-related bloodstream infections in the ICU. N Engl JMed 2006;355:2725e32.50. Cancer Facts and Figures, 2010. Atlanta: American Cancer Society,2010. http://www.cancer.org/acs/groups/content/@epidemiologysurveilance/documents/document/acspc-026238.pdf.(accessed 10 Mar 2011).51. Shojania KG, Burton EC, McDonald KM, et al. Overestimation ofclinical diagnostic performance caused by low necropsy rates. QualSaf Health Care 2005;14:408e13.52. Janjua KJ, Sugrue M, Deane SA. Prospective evaluation of earlymissed injuries and the role of tertiary trauma survey. J Trauma1998;44:1000e6.53. Kollef MH. Risk factors for the misdiagnosis of pneumothorax in theintensive care unit. Crit Care Med 1991;19:906e10.54. Fabian TC, Patton JH Jr, Croce MA, et al. Blunt carotid injury.Importance of early diagnosis and anticoagulant therapy. Ann Surg1996;223:513e22.55. Levi AD, Hurlbert RJ, Anderson P, et al. Neurologic deteriorationsecondary to unrecognised spinal instability following traumaeamulticenter study. Spine (Phila Pa 1976)2006;31:451e8.56. Drislane FW, Lopez MR, Blum AS, et al. Detection and treatment ofrefractory status epilepticus in the intensive care unit. J ClinNeurophysiol 2008;25:181e6.57. Battle RM, Pathak D, Humble CG, et al. Factors influencingdiscrepancies between premortem and postmortem diagnoses.JAMA 1987;258:339e44.58. Tavora F, Crowder CD, Sun CC, et al. Discrepancies between clinicaland autopsy diagnoses: a comparison of university, community, andprivate autopsy practices. Am J Clin Pathol 2008;129:102e9.59. Goldhaber SZ, Hennekens CH, Evans DA, et al. Factors associatedwith correct antemortem diagnosis of major pulmonary embolism. AmJ Med 1982;73:822e6.60. Singh H, Peterson LA, Thomas EJ. Understanding diagnostic errorsin medicine: a lesson from aviation. Qual Saf Health Care2006;15:159e64.61. Schiff GD, Hasan O, Kim S, et al. Diagnostic error in medicine: analysisof 583 physician-reported errors. Arch Intern Med 2009;169:1881e7.62. Crowther MA, Cook DJ, Griffith LE, et al. Deep venous thrombosis:clinically silent in the intensive care unit. J Crit Care 2005;20:334e40.63. Graber ML, Kissam S, Payne VL, et al. Cognitive interventions toreduce diagnostic error: a narrative review. BMJ Qual Saf2012;21:535e57.64. Wichmann D, Obbelode F, Vogel H, et al. Virtual autopsy as analternative to traditional medical autopsy in the intensive care unit:a prospective cohort study. Ann Intern Med 2012;156:123e30.65. Schiff GD, Kim S, Abrams R, et al; Agency for Healthcare Researchand Quality (AHRQ). Diagnosing diagnosis errors: lessons froma multi-institutional collaborative project. Adv Patient Saf2005;2:255e278. http://www.ahrq.gov/downloads/pub/advances/vol2/Schiff.pdf. (accessed 10 Mar 2011).66. Special Emphasis Notice (SEN): AHRQ Announces Interest inResearch on Diagnostic Errors in Ambulatory Care Settings.http://grants.nih.gov/grants/guide/notice-files/NOT-HS-08-002.html.(accessed 10 Mar 2011).PAGE fraction trail=8.5Winters B, Custer J, Galvagno SM, et al. BMJ Qual Saf (2012). doi:10.1136/bmjqs-2012-0008039e9

Downloaded from qualitysafety.bmj.com on July 23, 2012 - Published by group.bmj.comDiagnostic errors in the intensive care unit:a systematic review of autopsy studiesBradford Winters, Jason Custer, Samuel M Galvagno, Jr, et al.BMJ Qual Saf published online July 21, 2012doi: 10.1136/bmjqs-2012-000803Updated information and services can be found at:http://qualitysafety.bmj.com/content/early/2012/07/20/bmjqs-2012-000803.full.htmlReferencesP