Reframing Diagnostic Error: Maybe It's Content, and Not Process, That Leads to Error

Dr. Ritter identifies a key task required of emergency physicians (EPs), the ability to diagnose life-threatening conditions in undifferentiated patients within the chaotic environment of the emergency department. Studies demonstrate that EPs simultaneously care for greater than five patients (range, 2 to 16) at a time, while being interrupted 10 to 20 times per hour.1-3 Thus, diagnostic reasoning (and by implication diagnostic error) has become an increasingly important topic in medicine in general and in emergency medicine (EM) in particular.4, 5 Research in medical education has focused on the process of clinical reasoning in an attempt to develop strategies to prevent diagnostic error. Theories of clinical reasoning include dual processing theory,6-8 which involves an unconscious, heuristic-based System 1 and a conscious, analytical System 2; cognitive continuum theory,9 which argues for a continuum between pure intuition and pure analysis with the clinical problem determining the degree of “quasirationality” required; and fuzzy trace theory,10 which suggests there are two types of memory that inform decision-making, verbatim (literal facts) and gist (representative or relational facts). While a full description of the merits of each of these theories is beyond the scope of this commentary, a dominant argument in the literature is that intuitive or unconscious reasoning (System 1) is the main source of diagnostic error, as a consequence of the cognitive biases presumed inherent in this nonanalytical approach.6, 7 Therefore, it is argued that diagnostic errors can be reduced by understanding and limiting the influence of cognitive biases on the reasoning process. These strategies are collectively referred to as “cognitive forcing strategies.”1, 6 However, our research suggests that cognitive forcing strategies, which provide metacognitive approaches to self-monitoring the diagnostic process, are ineffective.11, 12 While these studies have been criticized for using medical students, who may lack sufficient clinical experience to use automatic, System 1 reasoning and thus do not require inoculation against cognitive biases, these are the only studies to our knowledge that have experimentally evaluated the educational benefit of metacognition to prevent diagnostic error. Other studies from our research group challenge more broadly the assumption that automatic, unconscious, heuristic, System 1 reasoning is the source of diagnostic error (with System 2 failing to correct it). Our research suggests that rapid diagnosis (a surrogate for unconscious, nonanalytical System 1 reasoning) is associated with fewer errors.13 Conversely, encouraging residents to slow down and pay attention (allowing more time for deliberate, analytical System 2 reasoning) did not reduce error; it simply made residents slower.14 These findings have been replicated and extended in a multicenter, randomized trial that specifically incorporated EM trainees and physicians and, showed, not surprisingly, that EPs were both faster and more accurate than trainees.15 Conversely, specific strategies to encourage conscious, reflective, System 2 reasoning had no effect on accuracy, although the strategies increased diagnostic processing time by a factor of 3. There is some evidence that deliberate, intensive reflection, directed at a critical examination of the relation between signs, symptoms, and diagnoses (and not at identification of cognitive biases), can lead to reduction in errors,16 but such a strategy, while it may have educational benefits,17 is not feasible in the clinical practice of EM. With regard to the effect of interruptions, a recent study, which incorporated distractions to mimic authentic clinical practice, showed that diagnostic accuracy correlates with clinical experience and (again) is not affected by instructions to slow down and be consciously systematic (i.e., System 2 reasoning) in the diagnostic process.18 Perhaps surprisingly, the effect of multiple distractions was to add a few seconds to the diagnostic processing time of each case, but again there was no negative effect on accuracy. While these studies have been criticized for the use of written cases,19 which lack face validity to mimic actual diagnostic processes in clinical practice, other studies of medical students have demonstrated that performance on written cases is equivalent to performance on both simulated cases and real patients.20 In any case, with the exception of a few retrospective studies of causes of error,4 the entire empirical basis for the purported effect of cognitive bias on cognitive reasoning is derived from written cases, both within21 and outside of medicine (and entirely from first-year undergraduate psychology students in the latter case).22 With calls from the Accreditation Council for Graduate Medical Education to address patient safety,23 it is insufficient for educators and frontline teachers to ignore the challenges of diagnostic reasoning suggested by Dr. Ritter's portfolio entry. Perhaps the approach to reducing diagnostic error is not in understanding more fully the complexities of the reasoning process, but in reframing the issue. The Canadian Patient Safety Institute provides a framework of safety competencies that describes communication, system, team processes, etc., that can improve patient outcomes.24 The prevailing opinion is that diagnostic error is a cognitive processing error—an inaccurate manipulation of data (transcribed from the clinical encounter) within the complexities of a physician's mind, analogous to the inaccurate transcription of a medication order on a patient chart (i.e., a communication process error). However, the analogy is incorrect. This perspective presupposes that all of the available knowledge is present and it is knowledge processing that leads to error. In contrast, a diagnostic error may reflect not a processing error, but an incomplete knowledge base or inadequate experience. Zwaan et al.,25 in a retrospective review of 247 dyspnea cases, found that the major source of error was a “mistake,” defined as “an intended act, but the physician does not know it is incorrect.” Our research has shown that diagnosis accuracy is correlated with both previous experience with similar true clinical cases and formal knowledge measured by written examinations.13 Moreover, the finding that incorrect diagnoses take the longest time13 suggests that increasing conscious, analytical System 2 reasoning cannot accommodate for knowledge or experience gaps. Diagnostic error may be less about process (e.g., bias) and more about content (e.g., gaps in knowledge or experience). In fact, it may be the case that generic metacognitive skills, generalizable across all types of clinical diagnoses, do not exist.26 Perhaps biases are content (i.e., diagnosis) specific. For example, techniques to avoid search satisficing (stopping a search when a single diagnosis or no diagnosis is found and failing to recognize a second or alternate diagnosis) may not necessarily be generalizable. Admonitions to search for the second fracture when the first fracture is discovered may not help a learner to identify myocardial ischemia on an ECG that simultaneously demonstrates a malignant dysrhythmia. If diagnostic errors have been incorrectly framed as process errors, when in fact they may reflect content gaps, the educational solution to decrease diagnostic errors involves increasing residents’ clinical experience. However, this is not a simple task, particularly in the current age of restricted resident duty hours. Two related instructional techniques offer solutions to this problem. The first is deliberate practice.27 This approach requires the repeated presentation of the same diagnostic task with immediate feedback on diagnostic accuracy28 (e.g., diagnosing a series of pediatric ankle fractures29). The second related instructional method is mixed practice, where a learner is exposed to an (epidemiologically valid) series of cases presentations, relating to a common presentation, and receives feedback on the accuracy of his or her diagnoses (e.g., diagnosing a series of common and uncommon chest pain presentations).30, 31 While the initial development of case banks can be anticipated to be resource-intensive (but mitigated if produced collaboratively), implementation of mixed practice instruction in simulated environments would optimize time away from actual patient care (the most essential element of residency training), while accelerating the acquisition of experience in novice learners without solely relying on ad hoc clinical exposures to gain experience. How should educators and teachers respond to Dr. Ritter's challenge? We suggest that the answer is not found in exclusively learning the expanding taxonomy of cognitive errors, nor exclusively focusing on metacognitive skills. Rather, we suggest that the answer may require an accelerated exposure, via mixed practice, to core EM clinical presentations, ensuring that the required knowledge and experience to distinguish an aortic dissection from acute coronary syndrome is achieved. Experience leads to better patient outcomes.32

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.