Guess early, then ask the questions that tell your guesses apart.
The idea
Skilled clinicians don't gather every fact and then decide. They form a few hypotheses the moment they hear the complaint, then choose each history question to discriminate between them — the answer should change what they believe. Keep two lists in mind: what's most likely, and what you can't afford to miss. A pertinent negative — a symptom that's absent — prunes the tree as sharply as a positive.
The case. A 58-year-old arrives with chest pain that began two hours ago. You have a small history budget. Ask the questions that best separate your hypotheses.
History questions
Most likely — ranked
Bars show relative suspicion for teaching, not real probabilities.
Can't-miss — keep pinned until excluded
Form your hypotheses, then ask a discriminating question to begin.
questions used 0 of 6high-yield 0 · low-yield 0discrimination
How it works
Generate early. From the presenting complaint alone, name three to five plausible diagnoses before you ask anything. Chest pain instantly suggests cardiac, musculoskeletal, reflux, and the dangerous few.
Split into two lists. "Most likely" drives your working plan; "can't-miss" is a separate watchlist of low-probability, high-lethality diagnoses that stay pinned until a test excludes them.
Ask to discriminate. A good question is one whose answer would move your ranking either way. "Does pressing on it reproduce the pain?" separates chest-wall pain from ischemia; "worse with breathing?" speaks to clot or lung.
Prune with pertinent negatives. No leg swelling, no recent immobility, not pleuritic — each absent feature lowers a hypothesis. Absence is evidence.
Confirm the can't-miss. History raises or lowers suspicion; it rarely excludes a killer on its own. The dangerous diagnoses come off the list only when the right test says so.
When to use it
Fits well
Key trade-off
Any diagnostic reasoning under time or resource limits — you can't ask everything.
Early hypotheses can anchor you; stay willing to reorder when the data pushes back.
Presentations with a handful of common causes and a few lethal ones.
Rare presentations may not fit your first hypotheses — keep a "none of these" branch open.
Teaching and structuring a focused history.
Discrimination is only as good as your hypothesis set; a missing hypothesis can't be ranked.
Watch out for
. Latching onto the first fit and stopping. The can't-miss list exists precisely so a tidy story doesn't hide a dangerous one.
Shotgun questioning. Asking everything hoping something sticks burns your budget and buries the signal. Prefer questions that separate your top two hypotheses.
Confirmation bias. Asking only questions that support your favourite. Deliberately seek the answer that would disprove it.
Excluding a killer on history alone. "It's probably not a clot" is a suspicion, not a rule-out. Match the level of certainty to the stakes.
Ignoring pertinent negatives. A missing red-flag symptom is high-value information — record it, don't skip it.
Worked example
Interviewer: "Walk me through your first three minutes with this chest-pain patient."
You'd say your hypotheses out loud first: cardiac ischemia, chest-wall pain, and reflux as likely, with heart attack, clot, aortic tear, and collapsed lung on a separate can't-miss list. Then you'd ask to discriminate: character and radiation (pressure to the arm and jaw raises ischemia), whether it's reproduced by palpation (rules chest-wall up or down), pleuritic quality and leg findings (prunes the clot), and cardiac risk factors (raises pre-test probability). Note each pertinent negative. You'd close by naming the plan: pursue ischemia as most likely and get an ECG and troponin, because the can't-miss items come off only with a test.
Check yourself
Your working diagnosis is reflux and the story fits. What still belongs in your plan?