Diagnostic error has received little attention in the patient safety field. Urmimala Sarkar, MD, and Kaveh Shojania, MD, summarize recent literature around the scope of diagnostic errors, review causes such as interruptions and system factors, and discuss strategies to enhance diagnostic accuracy.
Annual Perspective 2014
by Urmimala Sarkar, MD, and Kaveh Shojania, MD
Until very recently, diagnostic errors received relatively little attention in the field of patient safety, particularly when compared with other safety topics such as medication errors, surgical complications, and health care–associated infections. This is particularly surprising for a number of reasons. First, diagnostic errors have always attracted the attention of clinicians. Second, delayed or missed diagnoses constitute the most common (and costly) reasons for a malpractice claim, especially in the outpatient setting. Third, in hospital settings, a substantial literature highlights the frequency with which clinically important diagnoses escape detection untilautopsy.
One explanation for the lack of attention paid to diagnostic errors may lie with the historic interest in the topic primarily from physicians. Nurses, pharmacists, and other health professionals and administrators have often showed greater engagement in patient safety than physicians, but correct diagnosis resonates more with physicians than other health care professionals. Perhaps more importantly, physicians who find diagnostic errors interesting tend to focus on the cognitive aspects of diagnosis—emulating master clinicians (real-life versions of television's Sherlock Holmes–like Dr. House) or studying cognitive biases that can lead one astray. This focus on individual diagnostic acumen and the psychology of cognition may not attract those interested in system problems and the Swiss cheese model of patient safety.
Whatever the explanation for the relative inattention to diagnostic errors within the field, a commentary in the Journal of American Medical Association in 2009 called for making diagnostic errors the next frontier in patient safety. The 5 years since then have seen a gradual increase in publications on the topic. In 2013, BMJ Quality & Safety devoted a special supplement to the subject. And, 2014 saw notable examples of additional traction, including the launch of a new peer-reviewed, open access journal, Diagnosis, published by the Society to Improve Diagnosis in Medicine, and the convening of a committee at the Institute of Medicine onDiagnostic Error in Health Care, which will produce a formal report in 2015. In this review, we summarize important 2014 publications on diagnostic errors.
Epidemiology: Scope of the Problem
A major advance in 2014 was the estimation of the frequency of diagnostic errors for the adult population across the United States by Singh and colleagues. By combining several data sources, including estimates from consecutive lung cancer cases, as well as a trigger approach in which automated screening of the electronic health record for specific patterns of return visits prompted detailed chart reviews, the researchers estimated that 5% of adults in the US (i.e., more than 12 million individuals) experience a serious diagnostic error in the outpatient setting every year.
Newman-Toker and colleagues used data from the AHRQ Healthcare Cost and Utilization Project to show that, among patients with stroke, nearly 13% of patients had a visit to the emergency department (ED) within the prior 30 days that constituted a potential missed diagnosis. Just over 1% of patients had a probable missed diagnosis of stroke. Common (incorrect) diagnoses assigned at the previous ED visit included headache and dizziness. Other studies in 2014 looked at the frequency of diagnostic errors in specific clinical settings, such as missed injuries in pediatric trauma. Another study reported that 8% of deaths at a Level 1 trauma center were definitely or potentially preventable, with inaccurate diagnosis one of the most common contributing factors to the deaths. These studies all lend weight to the long-standing argument that the patient safety field must attend to missed and delayed diagnosis.
Causes of the Problem
As noted at the outset, something about diagnostic errors strikes many clinicians and even patient safety advocates as reflecting individual performance, with less obvious roles for a systems approach to addressing the problem. An interesting pediatric study highlighted this issue. Investigators reviewed the records of medical patients admitted to the pediatric ward or seen in the ED at a United Kingdom district hospital over a 90-day period. Two clinician reviewers judged misdiagnosis to have occurred in 5% of these children presenting with acute illness. The two reviewers identified cognitive factors as the cause of most diagnostic errors. Interestingly, investigators were able to interview the treating clinicians in 14 cases. In contrast to the reviewers, clinicians commonly identified organizational problems such as poor policies and inadequate staffing and equipment, making clear that the problem of diagnostic errors also has a systems component.
An analysis of outpatient physician focus groups similarly saw clinicians attribute diagnostic errors to a variety of system factors, including insufficient information availability, disjointed workflows, and poor communication among providers and with patients. Naturally, one might expect clinicians to preferentially attribute diagnostic errors to system problems rather than individual cognitive errors. But an interesting study in radiologyhighlighted the concrete role that interruptions can play in diagnostic accuracy, in the same vein that other patient safety studies have found that interruptions increase the chances of medication administration errors by nurses.
This study sought to assess the effect of distractions from telephone interruptions on diagnostic performance by radiology residents during 14-hour shifts on-call. A local quality assurance project in this radiology department at a tertiary care pediatric center tracked discrepancies between the diagnostic findings of the on-call residents and the final interpretation of staff radiologists. Major discrepancies occurred in 41 shifts over a 13-month period. Using telephone logs from the radiology reading room, the authors found a slightly greater average number of telephone calls for the "discrepancy shifts" than the "no discrepancy shifts" (48 vs. 44). While this difference was only of borderline significance (P = .057), the study also showed a statistically significant increase in the average number of phone calls in the 1 hour preceding the generation of a discrepant preliminary report versus the "no discrepancy shifts" (4.23 vs. 3.24 calls, P = .027). One additional phone call during the preceding hour was associated with a 12% increased likelihood of error (P = .017).
Even with documented examples of human factor contributors like interruptions, and other system contributors to diagnostic errors (e.g., missed test results, poor communication between providers, poorly designed computer interfaces), cognitive biases remain a fascinating area of study. One study in 2014 looked at the effect on diagnostic performance of distracting salient clinical features—findings that tend to grab physicians' attention because of a well-known association with a particular disease. Participating internal medicine residents (72 in total) each reviewed 12 case vignettes—with or without a salient distracting feature. The authors structured the cases such that some were simple and some complex, and so that the salient distracting feature appeared either near the beginning of the case description or closer to the end. Significantly more diagnostic errors occurred with salient distracting features, especially for complex cases, but only when the distractor appeared near the beginning of the case description.
Another study in 2014 showed that recent media exposure to information about a particular disease can bias physicians toward making that diagnosis in similar looking cases. This study highlights the importance ofavailability bias. Importantly, the authors showed that instructing the participants to reflect further restored performance to the level seen in participants not exposed to the information about a specific disease. This restoration of performance represents a simple example of cognitive debiasing. It also illustrates the augmentation of performance that occurs when clinicians shift from automatic to more reflective thinking. Another 2014 study highlighted the increased diagnostic accuracy that comes from shifting out of the fast, intuitive mode of thinking to a more reflective one. This study of computer keystrokes by residents taking a national certification examination showed that, for cases involving diagnosis, changing an answer from incorrect to correct occurred nearly twice as often as changing an answer from correct to incorrect.
Addressing the Problem
As intellectually appealing as cognitive debiasing may be, addressing systems problems contributing to diagnostic error still seems more likely to deliver demonstrable gains in safety and to serve as a feasible target for interventions and policies. The year 2014 saw one notable report of how two separate health care organizations addressed such system challenges. One system deployed voluntary reporting and root cause analysis for diagnostic errors; another created a system of review for incipient diagnostic delays using the electronic health record. Both strategies yielded promising early results.
Taken together, the literature published in 2014 represents a significant advance in our thinking about diagnostic errors. Importantly, research advances have put us in a position in which we may soon be able to measure the frequency of diagnostic errors in an individual institution in reproducible ways. Further standardization and pilot implementation will be needed in order to be able to create measures that will permit the application of tools such as comparative performance review, differential payment, and accreditation to prompt greater individual and institutional focus on this crucial area of patient safety. Such strategies have become the cornerstone of policy initiatives in the rest of patient safety; the absence of feasible measures of diagnostic accuracy has left diagnosis behind in the policy arena.
Even in the absence of rigorous and standard measures, 2014 provided actionable information about how to address diagnostic safety: using electronic health records and system-wide workflows to intercept diagnostic problems prior to patient harm, applying established patient safety strategies such as reporting and error analysis, and by teaching techniques to improve cognition. The focus of the patient safety field on diagnosis over the past 5 years appears to be bearing fruit.