How to score a 3

The playbook

How to ace any case

The rubric page tells you what is being measured. This one tells you what to actually say and do — the sentence templates, the interview technique, the sweep, the traps, and the drills that make all of it automatic.

01 — The phrasebook

Six sentences that carry the whole rubric

Most lost points are not knowledge gaps — they are formatting. Learn these six shapes and the reasoning becomes visible without you having to think about the grader at all.

Problem representation

[age and sex] with [relevant context], presenting over [tempo] with [syndrome in abstract terms], accompanied by [two or three findings] — which separates this from [nearest rival] because [feature].

In use

A 49-year-old man with no psychiatric history, presenting over days with a first-episode paranoid psychosis accompanied by autonomic hyperactivity, vertical nystagmus, and rigidity — which separates this from a primary psychotic disorder because the age and the physical findings both point to a toxidrome.

A differential entry

[Diagnosis] — supported by [finding 1] and [finding 2]; against it, [finding 3]. Ranked [above/below] [other diagnosis] because [discriminator].

In use

Serotonin syndrome — supported by three serotonergic agents within 72 hours and inducible clonus with lower-limb-predominant hyperreflexia; against it, nothing on the current data. Ranked above NMS because the tempo is hours rather than days and there has been no dopamine antagonist exposure.

Disposing of a can't-miss

[Diagnosis] is on the list because [what makes it lethal]. I asked [question]; the answer was [answer], which makes it unlikely — but I am [testing / monitoring] with [order] because [reason a negative history is not enough].

In use

Subdural haematoma is on the list because he is anticoagulated and it can fluctuate without focal signs. I asked about falls and head strikes; his daughter reports none, which makes it unlikely — but I am imaging him with a non-contrast head CT because an unwitnessed strike in a confused patient is exactly what a negative history misses.

An order cluster

[Cluster name], to test [hypothesis]: [test], [test], [test].

In use

End-organ damage, to test for rhabdomyolysis from sustained hypertonia: creatine kinase, comprehensive metabolic panel, urinalysis for myoglobin.

Stating uncertainty

What would change my mind: [specific finding or result].

In use

What would change my mind: a chest radiograph showing a mass, or a mood that turns out to be reactive on review.

The closing formulation

[Diagnosis] in a [patient descriptor]. [Can't-miss] was considered and set aside on [evidence]. Tripwire: [the finding that would reopen it].

In use

Major depressive disorder with melancholic features in a 68-year-old man fourteen months after bereavement. Occult malignancy was considered and set aside on a clear chest radiograph and an unremarkable metabolic panel. Tripwire: weight that does not recover as the mood does.

02 — Interview technique

Six rules for the history

Open first, then close

Begin with "tell me what has been happening" and let them run. Closed questions after that are aimed; closed questions before it are guesses. You also get the patient's own words, which are worth quoting later.

One concept per question

"Any new medications, and do you drink?" gets you a single answer to whichever half they heard. Compound questions lose findings and read as unfocused. Ask them separately.

Ask the can't-miss questions expecting a no

The value is not in the answer, it is in having asked. A can't-miss you listed and never pursued is the most reliably detected leak there is, and it costs you Directed History every time.

Get collateral whenever the patient cannot be their own historian

Confused, psychotic, manic, intoxicated, or cognitively impaired — in every one of those, the patient cannot tell you their own baseline, and the baseline is the diagnosis. Ask the family what they were like a week ago.

Negatives are findings

A documented "no antipsychotic exposure" is what kills NMS. Pertinent negatives only count when the diagnosis they exclude is on your list — which is another reason to write the list first.

Ask the question that could falsify, not the one that confirms

The strongest question is the one whose answer could destroy your leading diagnosis. Questions whose "yes" supports your favourite feel productive and teach you nothing. If every question you asked confirmed what you already thought, you were not taking a history — you were collecting agreement.

03 — The sweep

Seven categories, every altered patient

Run this on any confusion, agitation, or new psychosis. It produces a broad, clean differential and the questions to go with it, in about ninety seconds.

CategoryWhat lives thereWhat you ask
StructuralHaemorrhage, mass, trauma, hydrocephalusAnticoagulated? Fall or head strike? Focal deficit? Headache?
InfectiousMeningitis, encephalitis, sepsis, urinary tract infectionFever? Neck stiffness? Any source — urine, chest, skin, dental?
MetabolicGlucose, sodium, calcium, thyroid, uraemia, hepatic, hypoxiaFingerstick first. Then electrolytes, TSH, ammonia, saturation.
ToxicIntoxication, withdrawal, anticholinergic, serotonergic, NMSFull medication list including OTC. Last use. Last drink. Anything started or increased in two weeks?
NutritionalThiamine, B12How long has intake been poor? Any alcohol history?
Vascular / cardiacStroke, arrhythmia, silent ischaemiaECG. Focal exam. Rate and rhythm.
Risk statesSuicidality, homicidality, inability to care for selfIdeation, plan, intent, means. Access to a firearm. Who is at home?

04 — Traps

Six ways strong students lose points

Mistaking what the case volunteers for what you sought

A case will hand you findings whether or not you asked well. Capturing all of them proves you read; it does not prove you reasoned. This is why a perfect finding count sits happily beside a 2.

Treating a risk factor as a diagnosis

Trauma history is not PTSD. A family history of bipolar disorder is not bipolar disorder. Bereavement is not depression. Each raises prior probability and none of them satisfies a criterion.

Inheriting the previous clinician's diagnosis

The referral says depression; the prescription says SSRI. Both are findings to explain, not conclusions to adopt. Ask what the prescription was actually treating, and at what dose — that question has broken more cases than it has confirmed.

Reflexive concerns that do not apply here

Teratogenicity on a 49-year-old man. Ruling out anaemia in a case with no anaemia. These read as padding rather than thoroughness, and they actively cost you Initial Differential. A concern earns its place from this patient and this drug.

Certainty ahead of the workup

"Definitely", "obviously", "classic case", "I bet". Every one of these is scanned for and quoted back. They mark the moment you stopped generating hypotheses.

Speed as a virtue

Finishing in half the average time means you skipped the steps that carry the marks. There is no bonus for the clock.

05 — Pacing

Thirty minutes, allocated

0–2 minProblem representation, v1
2–4 minInitial differential, built by category
4–12 minHistory — one discriminator per entry, plus the full sweep
12–15 minExam, then rewrite the problem representation
15–21 minRanked differential with evidence attached
21–25 minOrders in named clusters
25–30 minManagement, risk, and a re-read of your own notes

Roughly a third of that budget is spent writing rather than gathering. That ratio looks wrong until you notice that three of the four criteria only ever read the writing.

06 — Drills

Five minutes a day

None of these require a case simulator. They work on any vignette you already have.

The 60-second one-liner

Take any question-bank vignette. Set a timer and write the problem representation in one sentence, ending on the discriminator. Do ten. This is the highest-yield sixty seconds in your study day.

The discriminator drill

Name any three diagnoses that could be confused with each other. Write the single question that separates each from the others. If you cannot, you do not yet know the difference — you only recognise the classic presentations.

The falsification drill

Take your leading diagnosis on any case and write what would kill it. If nothing would, your diagnosis is not a hypothesis and you are not doing clinical reasoning.

The category drill

Ninety seconds, any presenting syndrome: one or two diagnoses in each of psychiatric, substance, medical, neurologic, iatrogenic. Trains breadth without padding.

The leak log

After each debrief, write down which leak fired. The same leak twice is not bad luck, it is your pattern — and a named pattern is a fixable one.