Availability Bias In Diagnosis

Availability bias in diagnosis is when doctors think a disease is more likely because they remember seeing it recently. This makes them pick familiar diagnoses even if other tests point to a different cause.

Availability bias in clinical diagnosis is a cognitive heuristic where recent or memorable cases disproportionately influence diagnostic probability estimates. Clinicians overweight salient memories relative to base rates, skewing differential diagnosis away from less memorable but more probable conditions.

An emergency doctor recently treated three patients in a row with a rare blood clot condition. The next week, a patient comes in with leg pain from a pulled muscle, but the doctor immediately suspects a blood clot and orders expensive tests—not because the evidence points that way, but because the recent string of clot cases is still vivid in their memory.

A hospitalist diagnoses two consecutive cases of heparin-induced thrombocytopenia (HIT) within a single rotation. When a third patient presents with a falling platelet count, the clinician assigns high pre-test probability to HIT and initiates argatroban therapy before 4T-score calculation, despite the patient's 4T score being 2 (low-probability) and drug-induced thrombocytopenia from a newly added antibiotic being the statistically dominant differential. The clinician's asymmetric exemplar salience—driven by recency weighting of the two prior HIT cases—suppresses integration of the epidemiological prior (HIT incidence ~0.5% in heparin-exposed patients) and the base-rate-weighted posterior probability, resulting in unnecessary anticoagulation risk and instantiating textbook availability bias within the diagnostic memory network.

When a doctor recently saw a case, that memory feels important and influences new decisions. The bright memory pushes the doctor toward that same diagnosis next time.

Recent case exemplars create elevated activation in memory traces, which disproportionately bias evidence accumulation toward those diagnoses; this involves asymmetric weighting of exemplar salience within the diagnostic memory network. The temporal recency constraint reduces reliance on population base rates, producing a predictable skew in posterior probability assignment.

Pause and list other possible diagnoses before deciding. Check local disease rates and test results to guide choices.

Implement a structured differential checklist and force consideration of base rates and decision thresholds before finalizing diagnosis. Use audit feedback and decision support to recalibrate exemplar weighting and reduce recency-driven bias.

Overdiagnosis of recent conditions; Missed rare but probable diseases; Inappropriate treatment selection

Pharmaceutical and device manufacturers can seed clinician memory through vivid sponsored case reports or conference presentations featuring their product's target condition, artificially elevating exemplar salience to nudge prescribers toward that diagnosis without manipulating hard evidence. Health misinformation actors can flood clinical communities with dramatic anecdotal cases of rare conditions to systematically inflate perceived prevalence, causing coordinated overdiagnosis and resource diversion from more probable differentials. In medicolegal settings, opposing counsel or expert witnesses can strategically introduce memorable near-miss narratives to anchor diagnostic-standard expectations around atypical high-salience cases, exploiting the jury's own availability bias to undermine the defendant clinician's actual decision-making fidelity.

Implement structured differential diagnosis checklists that mandate explicit base-rate retrieval and epidemiological priors before exemplar recall is applied, directly disrupting recency-driven activation weighting. Deploy audit-and-feedback loops that surface a clinician's own historical diagnosis distribution against local prevalence data, enabling recalibration over time. Embed clinical decision support at the point of care presenting calibrated population statistics and explicit pre-test probability anchors, reducing reliance on memorable but unrepresentative exemplars. Use case-randomization protocols and blinded outcome reviews to desensitize clinicians to salient anecdotes.