The Incidence of Diagnostic Error in Medicine
Total Page:16
File Type:pdf, Size:1020Kb
BMJ Quality & Safety Online First, published on 7 August 2013 as 10.1136/bmjqs-2012-001615NARRATIVE REVIEW BMJ Qual Saf: first published as 10.1136/bmjqs-2012-001615 on 15 June 2013. Downloaded from The incidence of diagnostic error in medicine Mark L Graber ▸ Additional material is ABSTRACT wrong, missed, or egregiously delayed? published online only. To view A wide variety of research studies suggest that How often do diagnostic errors cause please visit the journal online (http://dx.doi.org/10.1136/bmjqs- breakdowns in the diagnostic process result in a harm? In this report, we briefly summar- 2012-001615). staggering toll of harm and patient deaths. ise the methods that have been used to These include autopsy studies, case reviews, estimate the rate of diagnostic error, Correspondence to Dr Mark L Graber, RTI surveys of patient and physicians, voluntary and comment on their relative merits and International, SUNY Stony Brook reporting systems, using standardised patients, limitations. A more comprehensive pres- School of Medicine, 1 Breezy second reviews, diagnostic testing audits and entation of studies using each of these Hollow, St James, NY 11780, closed claims reviews. Although these different methodologies has been presented USA; [email protected] approaches provide important information and elsewhere.6 Received 5 November 2012 unique insights regarding diagnostic errors, each Revised 28 April 2013 has limitations and none is well suited to Accepted 30 April 2013 THE INCIDENCE OF DIAGNOSTIC establishing the incidence of diagnostic error in ERROR actual practice, or the aggregate rate of error Arthur Elstein, a cognitive psychologist and harm. We argue that being able to measure interested in ‘how doctors think’, studied the incidence of diagnostic error is essential to clinical decision making for his entire enable research studies on diagnostic error, and career and concluded the diagnosis is to initiate quality improvement projects aimed at wrong 10–15% of the time.7 A diverse reducing the risk of error and harm. Three range of research approaches that have approaches appear most promising in this focused on this issue over the past several regard: (1) using ‘trigger tools’ to identify from decades suggest that this estimate is very http://qualitysafety.bmj.com/ electronic health records cases at high risk for much on target.6 diagnostic error; (2) using standardised patients The incidence of diagnostic error has (secret shoppers) to study the rate of error in been estimated using eight different practice; (3) encouraging both patients and research approaches (table 1). physicians to voluntarily report errors they Autopsy studies identify major diagnos- encounter, and facilitating this process. tic discrepancies in 10–20% of cases. Most cases in autopsy series derive from inpatient settings, but they also include In God we trust, all others bring data1 deaths from the emergency department on September 28, 2021 by guest. Protected copyright. The patient safety movement in the USA which, for many reasons, is considered to has entered its second decade. A wide be the natural laboratory for studying range of important safety concerns have diagnostic error. Although autopsies have been studied, and to this point, including virtually disappeared in the USA, autop- medication errors, hospital-acquired sies are still common in many other coun- infections, wrong-site surgery and a host tries, and despite the availability of of other issues. Strangely lacking, modern imaging, continue to show diag- however, is a concerted effort to find, noses being missed that might have been understand and address diagnostic lifesaving, particularly infections and car- – errors.2 4 One factor that may contribute diovascular conditions. to its relative neglect is that the true inci- Although autopsy data is considered dence of diagnostic error is not widely the ‘gold standard’ in terms of providing appreciated. Measuring the rate of error the most definitive data on the accuracy To cite: Graber ML. BMJ 5 Qual Saf Published Online and, in particular, error-related harm, of diagnosis, only a subset of cases ever First: [please include Day would provide the necessary motivation reach autopsy, and in many cases, the Month Year] doi:10.1136/ to begin addressing this large and silent relationship between clinical diagnoses bmjqs-2012-001615 problem. How likely is a diagnosis to be and autopsy findings remains unclear. Graber ML. BMJ Qual Saf 2013;0:1–7. doi:10.1136/bmjqs-2012-001615 1 Copyright Article author (or their employer) 2013. Produced by BMJ Publishing Group Ltd under licence. Narrative review BMJ Qual Saf: first published as 10.1136/bmjqs-2012-001615 on 15 June 2013. Downloaded from Table 1 Research approaches used to estimate the incidence of diagnostic error Suitable for Suitable for evaluating evaluating Research approach Findings—examples incidence aetiology Autopsies Major unexpected discrepancies that would have changed the management are Yes No found in 10–20%89 Patient and provider surveys One-third of patients relate a diagnostic error that affected themselves, a family Limited Limited member, or close friend10; Over half the surveyed paediatricians report making a diagnostic error at least once or twice a month11 Standardised patients Internists misdiagnosed 13% of patients presenting with common conditions to Yes Yes clinic (COPD, RA, others)12 Second reviews 10–30% of breast cancers are missed on mammography13;1–2% of cancers are Yes No misread on biopsy samples14 Diagnostic testing audits Errors related to laboratory testing are the most common reason for a diagnostic Very limited Limited error15 16 Malpractice claims Problems relating to diagnostic error are the leading cause for paid malpractice Very limited Limited suits in every large system Case reviews (cross-sectional Patients with asthma—median delay in making the correct diagnosis was Yes Limited studies by symptom, disease, or 3 years, or 7 visits17;12–51% of patients with subarachnoid haemorrhage are condition); (may be enriched by misdiagnosed in the emergency department.18 Of 1000 hospital deaths, 5% trigger tools) were considered preventable, and the most frequent aetiology was diagnostic error.19 Voluntary reports 1674 reports of diagnostic error were submitted to the UK’s National Reporting Yes Yes and Learning System over a 2-year period, 0.5% of all incidents reported20 COPD, chronic obstructive pulmonary disease; RA, rheumatoid arthritis. Autopsies also discover a large number of incidental smaller subset of conditions than would be seen in findings that were not suspected during life, but that usual practice. Moreover, in studies seeking to also were clinically irrelevant. study the factors relevant to accurate diagnosis, these Surveys have found that diagnostic errors are a patients may present with comorbid conditions or major concern of both patients and physicians. contextual complexities that are not representative of A survey of over 2000 patients found that 55% listed typical patients, and case complexity is a major factor 27 a diagnostic error as their chief concern when seeing a in determining diagnostic accuracy. http://qualitysafety.bmj.com/ physician in an outpatient setting.21 Similarly, phys- Second reviews refers to research protocols in the ician surveys have consistently found that approxi- visual subspecialties (eg, radiology, pathology, derma- mately half the respondents encounter diagnostic tology) where a second radiologist examines the same errors at least monthly.11 22 23 Moreover, compared films after a first radiologist, or a second pathologist with the many different safety concerns encountered reviews the same biopsy or cytology specimen as in practice, physicians perceive diagnostic errors to be another pathologist. These second review studies may more likely to cause serious harm or death compared be performed under controlled conditions, involving with other safety concerns.24 the review of many or mostly abnormal cases. This Standardised patients studies using ‘secret shoppers’ approach has advantages from a research perspective, have also been used to estimate the accuracy of diag- but substantially increases the possibility for diagnostic on September 28, 2021 by guest. Protected copyright. – nosis. In these studies, real or simulated patients with error, which can range from 10% to 50%.28 31 classical presentations of common diseases, like Interestingly, the studies also show that a diagnostician rheumatoid arthritis, asthma, or chronic obstructive will also disagree with his or her own prior interpret- pulmonary disease (COPD) are sent anonymously into ation in a small fraction of cases. real practice settings. In ‘real world’ situations, the majority of examina- The diagnostic error rates reported (13–15%) are tions are normal. Under these conditions, a critical very much in line with estimates from the other types abnormality is detected by a second expert reviewer of research approaches, and have substantial ‘face’ val- in the range of 2–5%. idity in that the studies are being carried out prospect- Diagnostic testing audits are used to estimate the ively in real-world settings. In addition to providing incidence of error in the clinical laboratory. Thanks to an estimate of diagnostic error rates, this approach impressive advances in quality control procedures, offers the unique ability to probe the various factors diagnostic errors in the modern age are rarely the that promote or detract from optimal diagnosis.25