A question bank is necessary for the MRCEM SBA and it is not sufficient. The exam is written entirely in single-best-answer format, so a bank can rehearse the recognition step almost perfectly — but four capabilities it cannot assess well are the ones that separate a comfortable pass from a resit: rapid ED prioritisation across a full department, reading images at speed, interpreting ECGs under time pressure, and sequencing time-critical management. This hub explains what a correct answer does and does not prove, and gives you a concrete way to train the gap.
The official format map: MRCEM SBA on the RCEM 2021 curriculum
The MRCEM SBA (the Intermediate written exam) is 180 single-best-answer questions delivered as two papers of 90, each lasting two hours, with a break in the middle. It is machine-marked with no negative marking — one mark for a correct answer, zero for an incorrect one — and the pass mark is set by a modified Angoff method with a standard-error adjustment. It is mapped to the 2021 RCEM curriculum, drawing on the Year 1–3 Specialty Learning Outcomes (SLOs 1 and 3–7), and historically passes roughly 45–50% of candidates. You need MRCEM Primary before it, and the OSCE follows after. The published domain weighting is worth pinning to the wall:
| Domain | Questions (of 180) |
|---|---|
| Care of the complex/stable adult (Complex Presentations) | 55 |
| Resuscitation and the acutely unwell | 40 |
| Care of the injured patient | 30 |
| Paediatric emergency medicine | 25 |
| Procedural skills / practical procedures | 20 |
| Complex and challenging situations | 10 |
At roughly 80 seconds per item, the paper itself rewards fast, accurate pattern recognition. That is exactly why it can lull you into training only the recognition step. Verify the current weighting and regulations in the RCEM MRCEM SBA information pack before you build a plan around these numbers.
Knowledge versus performance: what a correct answer proves
When you select the right option on an SBA, you have proved something narrow but real: given a clean, complete written stem and five options, you can recognise the best answer. That is knowledge, and it is worth having. What it does not prove is performance. It does not show that you can generate the differential without the options in front of you, extract the salient features from a messy real presentation, decide which of four simultaneously-sick patients to see first, or act inside the minutes that matter. The SBA hands you the discriminating features and the shortlist; the resus room hands you neither. A bank that only ever tests the recognition step will quietly convince you that you have trained the whole task. Reading your bank percentage as a competence figure is the classic error — the same one set out in Your Q-Bank Percentage Is Not Your Exam Score.
The four skills a bank under-tests — and how to train each
For each capability, train an observable behaviour, use a defined deliberate-practice task, take feedback from a defined source, and hold yourself to an exit standard. Vague "do more cases" advice is what leaves these gaps open.
ED prioritisation. The observable behaviour is triaging a board of several undifferentiated patients into a defensible see-first order and justifying it in one sentence each. The deliberate-practice task is a "shop-floor board" drill: take five to eight one-line presentations with observations and sort them by acuity against a clock. The feedback source is a senior emergency clinician who can challenge your ordering, plus the RCEM acuity framing in the curriculum. The exit standard is that you consistently place the genuinely time-critical patient first and can articulate the physiological reason. A bank cannot show you a whole board; it shows you one patient at a time, which is the opposite of prioritisation.
Images. The observable behaviour is naming the abnormality on a radiograph, CT slice, rash or eye within a few seconds and stating the next action. The deliberate-practice task is timed image decks — trauma films, chest radiographs, fundoscopy, dermatology — with a hard rule of five seconds to first impression. The feedback source is a labelled reference set or a radiology/EM colleague, checked against how the finding actually reports. The exit standard is reliable recognition of the must-not-miss set (tension pneumothorax, subtle fractures, extradural haemorrhage, papilloedema) at speed. Banks under-serve image items because they are expensive to produce, so you must force volume.
ECGs. The observable behaviour is a systematic read — rate, rhythm, axis, intervals, ischaemia — delivered inside a fixed time, ending in a management decision. The deliberate-practice task is a timed ECG stack worked to the same system every time, mixing the high-stakes patterns (STEMI territories, complete heart block, hyperkalaemia, ventricular tachycardia, long QT) with normal-variant distractors. The feedback source is a cardiology or EM reference and a colleague who can catch a missed subtlety. The exit standard is that you never skip a step under pressure and you act on the life-threatening patterns without hesitation. An SBA can show an ECG, but it cannot make you produce the systematic read yourself — it lets you reverse-engineer from the options.
Time-critical management. The observable behaviour is sequencing the first ten minutes of a resuscitation — the order and timing of actions, not just their names. The deliberate-practice task is verbalised or simulated resus runs against a stopwatch: sepsis, major haemorrhage, anaphylaxis, the crashing asthmatic, the septic child. The feedback source is a resus-trained clinician marking against an accepted algorithm and a rubric. The exit standard is correct actions in the correct order within the window, repeatable when tired. This is the capability an MCQ is least able to touch: it can ask "what is the next step" but cannot test whether you deliver it in time.
A four-week modality ladder
Build each skill up the same ladder rather than jumping to full simulation cold.
- Week 1 — isolated skill. Drill each capability alone: image decks, ECG stacks, board-sorting lists, single-algorithm runs. High volume, low fidelity, fast feedback.
- Week 2 — coached case. Work integrated cases slowly with a senior stopping you to probe reasoning. Fidelity up, speed deliberately down.
- Week 3 — timed integrated case. Combine the skills under a clock: a case that needs a board decision, an ECG read and a management sequence together.
- Week 4 — unseen simulation. Full, unseen scenarios at real pace with no coaching, scored against a rubric — plus an unseen SBA block to keep the recognition layer honest.
Score the timed and simulated rungs against a written rubric, not a gut feeling. A workable rubric for a resus run marks four things out of two each: correct actions, correct order, correct timing, and safe communication with the team. A candidate who scores full marks on actions but loses points on timing has a specific, trainable deficit — and that is exactly the deficit an MCQ percentage cannot show.
A worked example: the candidate who scores 80% and freezes on the board
Consider a trainee — call her Dr A — who has worked several thousand bank questions and sits at 80% first-attempt accuracy, comfortably above most published pass thresholds. On paper she looks ready. Put her in front of a shop-floor board drill — a chest pain, a drowsy diabetic, a limping child, a wrist injury and a breathless older woman all arriving together — and she stalls. She can manage each patient correctly in isolation but cannot decide who to see first or what to delegate. The bank never asked her to. Her 80% measures the recognition step for single, clean stems; it is silent on triage, on reading the breathless woman's ECG in ten seconds, and on running the sepsis pathway to time. The fix is not more questions. It is the ladder above, aimed squarely at the board decision and the timed integrated case: two weeks of board drills and verbalised resus runs, marked by a senior against a rubric, move a capability her percentage could never see. The lesson generalises — a high bank score sitting on top of an untested performance layer is not readiness, it is an unmeasured risk.
When AI feedback helps, when it misleads, and when you need a clinician
AI feedback is genuinely useful for the isolated-skill rungs: generating explanations, quizzing you on ECG criteria, drilling image recognition, and structuring spaced review of facts. It is unreliable exactly where the exam gets hard — judging the timing and sequencing of resuscitation, weighing competing priorities on a live board, or confirming that a subtle image finding is real rather than plausible. For those, an automated score can be confidently wrong. Before you trust any automated grade, calibrate it against a known-good human mark, as we set out in calibrating AI-graded feedback. The rule of thumb: use AI to rehearse and to check facts; use a clinician or examiner to certify performance, prioritisation and timing.
A balanced case and task matrix
The commonest self-sabotage is practising only the scenarios you already enjoy. Build a matrix of the six domains against the four under-tested skills and make sure every cell is covered — paediatric ECGs and injured-patient images, not just adult resus. The grid below is a starting template; the point is that no cell should sit empty going into the exam.
| Domain (marks) | Prioritisation | Images | ECGs | Time-critical management |
|---|---|---|---|---|
| Complex/stable (55) | multi-patient clinic sort | rashes, abdominal films | AF, ischaemia, blocks | DKA, sepsis sequence |
| Resuscitation (40) | peri-arrest triage | CXR for tension or effusion | VT, VF, hyperkalaemia | ALS, anaphylaxis, major haemorrhage |
| Injured patient (30) | trauma primary-survey order | trauma radiographs, CT | trauma-related arrhythmia | haemorrhage control, RSI prep |
| Paediatric EM (25) | sick-versus-well child triage | paediatric films and rashes | paediatric rhythms, long QT | paediatric sepsis, fluid maths |
| Procedural (20) | when to intervene first | landmark and ultrasound images | pre-procedure rhythm check | procedural sedation safety |
| Challenging situations (10) | capacity and agitation priority | — | — | rapid tranquillisation timing |
Weight your volume towards the domains that carry the most marks — complex/stable and resuscitation together are 95 of 180 — without letting paediatric EM or the smaller domains fall to zero. The blueprint-coverage method behind this grid — a matrix of domain against attempted volume and measured accuracy — is the same one in completion is not coverage; here you simply apply it to skills as well as topics.
Red flags that you are training the wrong thing
- Memorised scripts — you can recite a resus sequence but stall when the scenario deviates.
- Repeated cases — your "mocks" are cases you have seen, so you are testing memory, not judgement.
- Generic feedback — comments that would fit any candidate and never name a specific error.
- Uncalibrated scoring — a number with no rubric and no human check behind it.
- No official-rubric check — you have never marked yourself against the RCEM curriculum standard, only against a vendor's.
Any two of these together mean your practice is drifting away from what the exam and the shop floor actually demand.
Where a bank still earns its place
None of this means abandon the MCQ bank — it means place it correctly. A good bank, including iatroX, is the right tool for the recognition layer and for measurement: an unseen, timed, mixed MRCEM SBA baseline tells you where your knowledge is thin so you can aim the deliberate-practice ladder at the real gaps. Position it as the underlying-knowledge and unseen-measurement layer beneath your case-based and simulated practice — not as a substitute for the board drills, ECG stacks and resus runs that train the four skills above. Compare banks on the iatroX comparison hub, then let the bank do the one job it does well.
Frequently asked questions
How do I know whether I have covered the full MRCEM SBA blueprint? Build a coverage matrix: list the six curriculum domains down one axis and, for each, record how many questions you have attempted and your first-attempt accuracy. Cross-reference against the published weighting — 55 complex/stable, 40 resuscitation, 30 injured, 25 paediatric, 20 procedural, 10 challenging situations. Any domain with low attempted volume or low accuracy is uncovered, regardless of what your overall percentage says. Coverage is a per-domain judgement, never a single number.
Can one question bank be enough for MRCEM SBA? For the written knowledge, a single high-quality bank can be enough to learn from, provided you measure transfer somewhere unseen. But no bank is enough for the exam as a whole, because prioritisation, image and ECG speed, and time-critical sequencing are performance skills an SBA cannot train. One bank plus deliberate skill practice and clinical exposure is the realistic minimum; one bank alone trains recognition and leaves the rest.
What should I measure instead of my overall Q-bank percentage for MRCEM SBA? Measure first-attempt accuracy on unseen, mixed, timed blocks, broken down by curriculum domain, and track it over time. Add pacing (are you inside ~80 seconds per item?) and skill-specific standards — seconds to an ECG diagnosis, correct see-first ordering on board drills, resus actions delivered in the right window. Those measures track competence; a rising overall percentage on repeated questions mostly tracks memory.
When should I stop doing new MRCEM SBA questions? Stop adding new items when your first-attempt accuracy on unseen mixed blocks has plateaued across all six domains and your remaining weaknesses are performance skills rather than knowledge gaps. At that point more MCQs give diminishing returns, and your time is better spent on timed integrated cases, ECG and image speed, and full resus simulations scored against a rubric.
Which MRCEM SBA resource should I use for my weakest component? Match the resource to the deficit. For thin knowledge, use a bank and structured notes and re-measure on unseen items. For slow image or ECG recognition, use timed labelled decks. For prioritisation and time-critical management, you need coached cases and simulation with a senior emergency clinician — no bank substitutes for that. Diagnose the weakness type first with an unseen MRCEM SBA baseline, then pick the tool that trains that specific skill.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026; exam format and weightings are drawn from the RCEM MRCEM SBA regulations and 2021 curriculum and should be verified in the current information pack, as counts and rules change between diets. Disclosure: iatroX operates an MRCEM SBA question bank; this article states plainly that a bank — including iatroX — trains the written recognition and measurement layer only and does not replace the simulation, coached cases and clinical exposure that train prioritisation, imaging, ECG speed and time-critical management. Corrections are welcome via the feedback route on iatrox.com.
References: Royal College of Emergency Medicine — MRCEM exams, SBA regulations and 2021 curriculum (rcem.ac.uk, rcemcurriculum.co.uk); iatroX MRCEM SBA bank; Your Q-Bank Percentage Is Not Your Exam Score; completion is not coverage; calibrating AI-graded feedback.