Residency · Residency · Emergency Medicine
Diagnostic Error and Cognitive Bias in the ED
Introduction
Diagnostic error is the leading cause of malpractice claims in emergency medicine and a significant source of patient harm. Studies estimate that diagnostic errors occur in 5 to 10% of ED encounters, with approximately 1 in 18 emergency patients experiencing a diagnostic adverse event. Emergency physicians make hundreds of decisions per shift under conditions of uncertainty, time pressure, cognitive load, and interruption, an environment that amplifies the natural cognitive shortcuts (heuristics) that can lead to systematic errors (biases). Understanding these biases and implementing debiasing strategies is essential for patient safety.
The Scope of Diagnostic Error in the ED
Prevalence and Impact
An estimated 5.7% of ED patients experience a diagnostic error, translating to approximately 7.4 million per year in the US. Missed diagnoses account for the majority, as opposed to wrong or delayed diagnoses. The high-risk conditions most commonly missed include acute MI, PE, stroke, aortic dissection, appendicitis, subarachnoid hemorrhage, spinal epidural abscess, meningitis, ectopic pregnancy, and testicular torsion. Contributing factors span patient factors (atypical presentations, communication barriers), system factors (crowding, interruptions, inadequate follow-up), and cognitive factors (bias, knowledge gaps).
Dual Process Theory of Clinical Reasoning
System 1 thinking is intuitive and fast, relying on pattern recognition, heuristics, and automatic processing. It is efficient for experienced clinicians in familiar situations but error-prone when patterns are incomplete or misleading. System 2 thinking is analytical and slow, involving deliberate, effortful, logical reasoning. It is resource-intensive and used for novel, complex, or high-stakes decisions. Most clinical reasoning uses a blend of both systems, and errors occur when System 1 operates without appropriate System 2 override. The goal is not to eliminate System 1 thinking, which is essential for ED efficiency, but to recognize when to engage System 2.
Common Cognitive Biases in Emergency Medicine
Anchoring Bias
Anchoring bias involves fixating on an initial impression or piece of information and failing to adjust the diagnosis as new data emerges. For example, a 45-year-old presents with chest pain and a recent COPD exacerbation, and the physician anchors on the respiratory diagnosis without evaluating for ACS despite risk factors and ECG changes. The debiasing strategy is to actively ask "What else could this be?" at key decision points and to reassess the differential when test results return.
Premature Closure
Premature closure means accepting a diagnosis before it is fully verified, stopping the diagnostic process too early. For example, a patient with right lower quadrant pain and mild leukocytosis is diagnosed with appendicitis, but the elevated beta-hCG is missed because the clinician stopped looking after the first plausible diagnosis. The debiasing strategy is to always consider at least three diagnoses on the differential and to review all results before committing to a diagnosis.
Availability Bias
Availability bias involves judging the likelihood of a diagnosis based on how easily examples come to mind rather than actual prevalence. After treating a rare case of aortic dissection, a physician may order CTA on the next several chest pain patients with very low pre-test probability, or conversely, a physician who has never seen a dissection may fail to consider it. The debiasing strategy is to use evidence-based clinical decision rules and epidemiologic data rather than personal recall.
Confirmation Bias
Confirmation bias involves seeking and valuing information that confirms the working diagnosis while ignoring or dismissing contradictory evidence. A patient diagnosed with migraine whose normal CT head is noted but whose stiff neck and fever suggesting meningitis are ignored illustrates this bias. The debiasing strategy is to actively seek disconfirming evidence and ask "What findings would make me change my diagnosis?"
Search Satisficing (Satisfaction of Search)
Search satisficing means stopping the diagnostic search after finding the first abnormality and missing additional pathology. A rib fracture identified on chest X-ray may cause the small pneumothorax underneath to be missed because the search ended after the fracture was found. The debiasing strategy is to complete a systematic review of all imaging and data using structured reading patterns for radiographs.
Framing Effect
The framing effect occurs when the diagnostic assessment is influenced by how clinical information is presented. When EMS reports "anxiety attack" for a young woman with chest tightness and hyperventilation, the physician may not evaluate for PE because the case has been framed as psychiatric. The debiasing strategy is to obtain one's own history directly from the patient and not accept diagnostic labels uncritically from triage, EMS, or referring providers.
Diagnostic Momentum
Diagnostic momentum occurs when a diagnosis assigned early in the patient's course gains momentum and is perpetuated from provider to provider without reassessment. A patient admitted three times for "COPD exacerbation" who actually has recurrent heart failure that was never reconsidered illustrates this bias. The debiasing strategy is to re-examine the patient and question prior diagnoses, especially when the expected clinical trajectory does not match.
<image>Infographic showing six common cognitive biases in emergency medicine (anchoring, premature closure, confirmation bias, availability bias, search satisficing, and framing effect) with a brief definition and clinical example for each, arranged in a circular layout with the central theme of diagnostic error</image>
Attribution Bias
Attribution bias occurs when the physician's perception of the patient based on stereotypes, appearance, or behavior influences the clinical evaluation. A disheveled, intoxicated patient's headache attributed to alcohol withdrawal may lead to a missed subdural hematoma from a fall. The debiasing strategy is to provide the same standard of diagnostic evaluation regardless of patient demographics or behavior and to recognize when personal judgments are influencing clinical decisions.
Posterior Probability Error (Zebra Retreat)
Posterior probability error involves dismissing a rare but serious diagnosis because "it's probably not that." Dismissing the possibility of meningitis in a febrile young adult because "it's probably just the flu" exemplifies this bias. The debiasing strategy is to recognize that for high-consequence conditions, a low probability still warrants evaluation, using the "worst-case scenario" approach for high-risk presentations.
Visceral Bias
Visceral bias involves countertransference, where the physician's emotional response to a patient affects clinical judgment. Frustration with a frequent-flyer patient may lead to an abbreviated workup, while excessive empathy for a colleague's family member may lead to premature reassurance. The debiasing strategy is to recognize emotional responses and consciously separate them from clinical decision-making.
System-Level Contributors to Diagnostic Error
Environmental Factors
ED crowding and boarding increase cognitive load, interruptions, and the pressure to disposition patients quickly. Emergency physicians experience an interruption every 6 to 9 minutes on average, and each interruption increases the risk of losing the cognitive thread. Shift work and fatigue from circadian disruption, sleep deprivation, and end-of-shift fatigue impair judgment. High patient volume forces reliance on System 1 thinking and reduces time for deliberate analysis.
Communication Failures
Handoff errors involve incomplete or inaccurate patient information transfer at shift change. Radiology over-read discrepancies may not be communicated when significant changes occur from preliminary to final reads. Consultant miscommunication occurs when verbal orders and recommendations are not documented or followed through. Discharge instruction failures happen when patients do not understand return precautions or follow-up plans.
<image>Swiss cheese model of diagnostic error adapted for the emergency department showing multiple layers of defense (clinical reasoning, diagnostic testing, second opinions, safety net systems, patient follow-up) with holes representing cognitive biases, system failures, and communication gaps that align to allow diagnostic error to reach the patient</image>
Debiasing Strategies
Individual Strategies
Metacognition, or thinking about thinking, involves actively monitoring one's own reasoning process and recognizing when shortcuts are being used. A diagnostic timeout at key decision points, particularly before disposition and when something does not fit, involves pausing to reassess the differential. Universal checklists for high-risk presentations ensure consistency: always check beta-hCG in reproductive-age women with abdominal pain, always consider PE in dyspnea, always consider ACS in diabetes with vague symptoms. The "worst-case scenario" question asks what is the most dangerous diagnosis that cannot be afforded to miss. Second opinions, presenting the case to a colleague especially when uncertain, provide fresh eyes that may see what anchoring has obscured. The teach-back method of explaining diagnostic reasoning out loud forces System 2 engagement.
System Strategies
Structured handoff tools such as I-PASS and SBAR reduce information loss during transitions. Automated safety nets including follow-up systems for pending labs, cultures, and radiology over-reads, with abnormal result callback protocols, catch errors. Decision support tools embedded in the EMR provide clinical decision rules, drug interaction alerts, and critical value notifications. A culture of safety with non-punitive error reporting, regular morbidity and mortality conferences, and psychological safety for raising diagnostic concerns supports improvement. Simulation training with cases designed to trigger specific biases provides practice recognizing and overriding cognitive errors.
Disclosure and Learning from Error
Open disclosure of diagnostic errors to patients is both an ethical obligation and a risk-reduction strategy, as studies show disclosure reduces litigation. Root cause analysis and morbidity and mortality conferences should focus on system contributors, not individual blame. Error reporting systems at institutional and national levels aggregate data to identify patterns and drive system improvements. A growth mindset accepts that diagnostic error is inherent to the practice of emergency medicine, with the goal being to build systems and habits that minimize its frequency and impact.
Clinical Pearls
Diagnostic error occurs in approximately 1 in 18 ED encounters, meaning it is not rare and every emergency physician will make diagnostic errors in their career. Anchoring and premature closure are the two most common cognitive biases contributing to diagnostic error in the ED, and actively maintaining a broad differential is the most effective countermeasure. The "worst-case scenario" question is a powerful debiasing tool that prompts consideration of what dangerous diagnosis cannot be afforded to miss. System factors including crowding, interruptions, fatigue, and communication failures amplify individual cognitive biases, making system improvement as important as improving individual reasoning. Structured handoffs, follow-up systems for pending results, and a culture of psychological safety are the most impactful system-level interventions for reducing diagnostic error.
References
- Newman-Toker DE, et al. "Burden of Serious Harms from Diagnostic Error in the USA." BMJ Quality & Safety. 2024;33(2):109-120.
- Croskerry P. "The Importance of Cognitive Errors in Diagnosis and Strategies to Minimize Them." Academic Medicine. 2003;78(8):775-780.
- Graber ML, et al. "Diagnostic Error in Internal Medicine." Archives of Internal Medicine. 2005;165(13):1493-1499.
- Croskerry P. "Cognitive Forcing Strategies in Clinical Decision Making." Annals of Emergency Medicine. 2003;41(1):110-120.
