A differential diagnosis is a working set of possibilities that should change as new information becomes available. Early impressions are often useful, but they can become a problem when a reasonable first hypothesis becomes difficult to dislodge after the clinical picture begins to contradict it.
The National Academies describes diagnosis as an iterative process in which clinicians gather information, generate diagnostic possibilities, and refine those possibilities as additional information becomes available. A working diagnosis may consist of a single provisional diagnosis or a differential diagnosis that reflects remaining uncertainty (National Academies of Sciences, Engineering, and Medicine [NASEM], 2015).
Start With a Working Differential
Early diagnostic reasoning happens with incomplete information. The history may still be developing, examination findings may be nonspecific, and laboratory or imaging results may not yet be available. A useful differential identifies the most likely explanation while preserving other possibilities that would substantially change management if they were true.
The differential should become more focused as information accumulates. Each active possibility should have a reason for staying on the differential, and the clinician should know what information would make it more or less likely. As the encounter progresses, some diagnoses should rise, others should fall, and some should leave the differential entirely.
Do Not Let the First Label Become the Final Answer
Anchoring occurs when an initial impression continues to exert too much influence after information appears that should prompt reconsideration. Premature closure occurs when diagnostic reasoning stops before reasonable alternatives have been adequately considered (Etchells, 2015). These problems can begin before the clinician sees the patient because a triage note, referral diagnosis, previous chart entry, or handoff can frame the encounter around an existing explanation.
In a study of Veterans Affairs emergency department visits among patients with congestive heart failure presenting with shortness of breath, Ly et al. (2023) found that physicians were less likely to test for pulmonary embolism, and testing occurred later, when congestive heart failure was mentioned in the triage documentation. The pattern was consistent with anchoring bias, although the study did not directly measure clinicians’ cognitive processes.
An existing diagnosis should function as information, not as proof. “History of migraine,” “anxiety,” “viral illness,” or “chronic back pain” may accurately describe previous episodes without fully explaining the current presentation.
Ask What Does Not Fit
A diagnosis becomes more convincing when it explains the important features of the case. Findings that do not fit deserve attention because they may indicate that the working diagnosis needs to be reconsidered. Confirmation bias can pull reasoning in the opposite direction. Once a diagnosis seems likely, clinicians may preferentially notice information that supports it and give less weight to evidence that challenges it. Etchells (2015) recommends actively seeking information that could refute the provisional diagnosis as one strategy for reducing anchoring and premature closure.
That can include asking what findings would be expected if the working diagnosis were correct and whether the patient actually has them. It may also mean identifying the feature that is hardest to explain. A patient presumed to have a routine musculoskeletal problem may have progressive neurologic deficits. A patient with a presumed viral syndrome may have persistent hypotension. A familiar headache may now have a different onset, severity, or neurologic finding.
A brief diagnostic time-out can make that reconsideration explicit: What diagnosis am I currently assuming? What evidence challenges it? What important alternative would change management if I were wrong? Discordant findings do not automatically make the initial diagnosis incorrect, but they provide a reason to reopen the differential.
Rank Diagnoses by Probability and Consequence
Diagnostic reasoning begins with probability, but the consequences of missing a diagnosis can also influence whether it deserves additional testing, observation, consultation, or follow-up. NASEM describes the differential diagnosis as a set of possible diagnoses that is refined and prioritized as information accumulates (NASEM, 2015).
A diagnosis does not have to be the most likely possibility to remain clinically important. A lower-probability diagnosis may still warrant action when the consequences of missing it are substantial. The most likely diagnosis may remain at the top of the list while another possibility stays active because overlooking it could lead to significant harm.
Let New Information Reopen the Differential
Diagnostic testing is most useful when the information gained can meaningfully refine the working diagnosis or influence the next clinical decision. A normal result may lower the likelihood of one diagnosis while doing little to address another. An abnormal result may support the leading hypothesis without being specific enough to establish it.
The clinical course belongs in the same reasoning process. NASEM describes diagnosis as iterative, with new information incorporated and prior hypotheses revised over time (NASEM, 2015). Failure to improve as expected, worsening symptoms, or new findings should influence whether the original explanation still fits.
A return visit should trigger reconsideration when symptoms are persisting, worsening, or changing rather than automatically inheriting the diagnostic framing of the previous encounter. AHRQ describes anchoring as persistence with an initial impression despite later information that should prompt reconsideration (Agency for Healthcare Research and Quality [AHRQ], 2019).
Diagnostic momentum can reinforce the same problem across encounters. Reilly and Webster (2017) describe how a diagnostic label can be propagated through successive clinicians until a tentative impression begins to function as an established diagnosis. Reopening the differential does not require abandoning the leading diagnosis; it means allowing the current evidence to determine whether that diagnosis should remain in the lead.
Diagnostic Uncertainty Is Part of Clinical Practice
A differential diagnosis is useful partly because certainty is not always available during the first encounter. NASEM emphasizes that diagnosis unfolds over time and that uncertainty may remain even after substantial evaluation (NASEM, 2015). Developing advanced-practice clinical judgment includes learning to make a safe plan with incomplete information while remaining willing to revise that plan as the evidence changes.
A strong differential remains responsive to the patient in front of you. It prioritizes plausible diagnoses, preserves dangerous alternatives when they still require action, and changes when the clinical story no longer supports the assumptions that came first.
References
Agency for Healthcare Research and Quality. (2019). Diagnostic errors. PSNet. https://psnet.ahrq.gov/primer/diagnostic-errors
Etchells, E. (2015, June 1). Anchoring bias with critical implications. PSNet. Agency for Healthcare Research and Quality. https://psnet.ahrq.gov/web-mm/anchoring-bias-critical-implications
Ly, D. P., Shekelle, P. G., & Song, Z. (2023). Evidence for anchoring bias during physician decision-making. JAMA Internal Medicine, 183(8), 818–823. https://doi.org/10.1001/jamainternmed.2023.2366
National Academies of Sciences, Engineering, and Medicine. (2015). Improving diagnosis in health care. The National Academies Press. https://doi.org/10.17226/21794
Reilly, J. B., & Webster, C. (2017, March 1). Diagnosing a missed diagnosis. PSNet. Agency for Healthcare Research and Quality. https://psnet.ahrq.gov/web-mm/diagnosing-missed-diagnosis