You are ready to stop doing new MRCEM SBA questions only when you can produce evidence — not a feeling and not a headline percentage — against every line of the checklist below. The minimum evidence is this: documented coverage across all six curriculum domains at their blueprint weighting; stable first-attempt accuracy on unseen, timed, mixed blocks; deliberate practice on the interpretation and time-critical formats the written paper leans on; a recency check on guidance-sensitive topics; and calibration against RCEM's own sample questions. Miss any one and you have a content gap, however complete the bank looks.
Current exam snapshot
The MRCEM SBA (Intermediate) is 180 single-best-answer questions delivered as two computer-based papers of 90 questions, each lasting two hours, with a break between them — roughly five hours of attendance and a pace near 80 seconds per item (RCEM, checked 19 July 2026). It is machine-marked with no negative marking, and the pass mark is set per sitting by modified Angoff standard-setting, so there is no fixed percentage to aim at. It maps to the 2021 RCEM curriculum, sampling the specialty learning outcomes, and it sits between MRCEM Primary (its prerequisite) and the MRCEM OSCE (which follows). The authoritative sources are the RCEM curriculum and RCEM's official sample questions; treat those as your calibration standard and everything else as practice volume.
The blueprint weighting across the six adult and paediatric domains, expressed as approximate questions per 180-item exam, is shown below. RCEM does not publish an exact per-sitting count for each domain, so treat these as blueprint proportions and confirm against the current curriculum rather than as a guarantee for your specific paper.
The blueprint coverage table
This is the core instrument. Fill one row per domain, every fortnight, from your bank's analytics. If you cannot complete a cell, that is itself a finding.
| Domain (approx. weight per 180) | Questions attempted | First-attempt accuracy | Last reviewed | Confidence (H/M/L) |
|---|---|---|---|---|
| Complex or Common Stable Patient (~55) | ||||
| Resuscitation (~40) | ||||
| Injured Patient (~30) | ||||
| Paediatric Emergency Medicine (~25) | ||||
| Procedural Skills (~20) | ||||
| Complex or Challenging Situations (~10) |
The column that matters most is the second, read against the weighting in the first. Because the Complex or Common Stable Patient domain is the largest single slice, banks and candidates both over-invest there, and it flatters an overall percentage. Resuscitation and the Injured Patient carry disproportionate clinical and exam weight for their size. If your attempted-question distribution does not roughly track the blueprint proportions, your headline accuracy is measuring the wrong thing — it is measuring your best domain, loudly.
Ten domain-level blind spots self-selected practice hides
These are the areas most likely to stay hidden when you choose your own questions. Require an exam-specific clinician review of each before you conclude you have covered it — a colleague or trainer who works in the ED and knows the RCEM blueprint.
- Paediatric resuscitation and weight-based drug dosing under time pressure.
- Toxicology and the specific antidotes and their thresholds.
- Environmental emergencies — hypothermia, drowning, electrical and heat injury.
- Ophthalmological and ENT emergencies, which are small in volume but examinable.
- Major trauma decision-making, including massive transfusion and damage-control priorities.
- Safeguarding — children and vulnerable adults — and the associated legal duties.
- Mental health presentations in the ED, including capacity and the relevant legislation.
- Obstetric and gynaecological emergencies presenting to the ED.
- Elderly-care emergencies, frailty and the atypical presentation.
- End-of-life and palliative decisions in the emergency setting.
Any of these can be under-represented in a self-selected feed while your overall score climbs. Tick a blind spot off only when you have attempted unseen items in it and an ED clinician has confirmed the depth is examination-appropriate.
Format checklist: the ways the paper tests, not just the topics
The SBA is a written paper, but it leans hard on emergency-specific reasoning that flashcard-style recall does not cover. Verify deliberate practice on:
- ED prioritisation — items that force you to choose the next action when several are reasonable, under an implicit clock. This is the signature of emergency-medicine questions.
- Image interpretation — the paper uses pictorial material; confirm you have practised on images, not just text stems.
- ECGs — rhythm and ischaemia recognition against time-critical management, not in isolation.
- Time-critical management — sequencing of resuscitation and stabilisation, where the correct answer is about order and timing as much as about the drug or manoeuvre.
If your bank practice has been overwhelmingly text-only single facts, you have a format gap even if your topic coverage looks complete.
Interpretation checklist
Separately from topics and formats, confirm you have deliberately practised each interpretation modality the exam can draw on:
- Images — rashes, wounds, foreign bodies, clinical photographs.
- ECGs — the full range from blocks to ischaemia to electrolyte effects.
- Radiographs — plain films for fractures, chest and abdominal emergencies.
- Laboratory trends — reading a changing picture, not a single snapshot: gases, renal and liver trends, sepsis markers.
- Calculations — paediatric doses, infusion rates, correction formulae.
- Ethics, capacity and relevant statute, and basic statistics where the item calls for it.
Tick each only when you have unseen practice in it. An untested modality is a silent gap.
Recency checklist
Emergency medicine is guidance-sensitive, and a bank written two years ago can teach a superseded answer. For every guidance-dependent topic, record the date and jurisdiction of the source you are trusting:
- Resuscitation algorithms — confirm against the current Resuscitation Council UK guidance.
- Sepsis, major haemorrhage and trauma pathways — confirm against current national guidance.
- Toxicology thresholds and antidote regimens — confirm against current UK toxicology advice.
- Any topic where NICE or a specialty society has updated within your revision window.
Beside each, write the source and its date. If you cannot name the date and jurisdiction, treat the answer as unverified until you can.
Performance checklist
Finally, verify the performance conditions that make a score meaningful:
- Unseen, timed, mixed blocks — you have measured yourself on items you had not seen, at ~80 seconds each, with all domains mixed, not topic-filtered.
- Speed — you finish within time without a late-paper collapse in accuracy.
- High-confidence errors — you have specifically reviewed the questions you got wrong while sure you were right; these are the dangerous ones, and they do not show up in an average.
- Retention — items you learned weeks ago are still correct now, tested by spaced re-exposure, not re-drilled the same day.
- Official-material calibration — your bank accuracy is sanity-checked against RCEM's own sample questions.
Stop or continue: a decision tree
Run your completed checklist through this tree rather than through your mood:
- If a coverage cell is empty or below your accuracy floor → continue new questions, targeted at that domain or modality.
- If coverage is complete but first-attempt accuracy on unseen blocks is unstable → consolidate: fewer new questions, more spaced review of misses and short source reads.
- If coverage and accuracy are stable but pace or high-confidence errors are the problem → simulate: full-length timed mixed papers under exam conditions.
- If a specific domain resists improvement despite practice → seek teaching from an ED clinician rather than grinding more items.
- If every line is green and calibration against official material holds → stop adding new questions and rest; more volume now adds fatigue, not marks.
Worked example (invented data)
A candidate three weeks out shows: Complex or Common Stable Patient 220 items at 78%; Resuscitation 60 at 71%; Injured Patient 45 at 74%; Paediatric EM 20 at 58%; Procedural Skills 18 at 69%; Complex or Challenging Situations 8 at 63%. Overall accuracy is a comfortable-looking 74%. The checklist tells a different story: two domains near their blueprint weight (Paediatric EM, Complex or Challenging Situations) are barely sampled and below the accuracy floor, and image or ECG practice has not been logged at all. The decision is not "stop" — it is continue, targeted at paediatric EM and the challenging-situations domain, add ECG and image blocks, then re-audit before any talk of stopping. The 74% was hiding the gap.
Your one-page checklist
Copy this and mark each line pass or fail before you decide to stop:
- All six domains attempted in rough proportion to the blueprint weighting.
- First-attempt accuracy logged per domain on unseen items.
- Each of the ten blind spots attempted and clinician-reviewed.
- ED prioritisation, images, ECGs and time-critical sequencing all practised.
- Every interpretation modality (images, ECGs, radiographs, lab trends, calculations, ethics/stats) tested.
- Recency and jurisdiction recorded for every guidance-sensitive topic.
- Unseen, timed, mixed blocks completed; high-confidence errors reviewed.
- Retention confirmed by spaced re-testing.
- Calibrated against RCEM official sample questions.
Three mistakes this checklist is designed to stop
The first mistake is treating bank completion as coverage. A bank that reports "100% complete" tells you that you have seen its questions, not that you have covered the RCEM blueprint — the two are different objects, and a bank can be finished while whole domains such as toxicology or environmental emergencies remain thinly examined. The coverage table exists precisely to divorce "I have finished the bank" from "I have evidence across the blueprint," and it is the reason the checklist is built around attempted questions per domain rather than a completion badge.
The second mistake is trusting an overall percentage that hides a weak, blueprint-heavy domain. Because the Complex or Common Stable Patient domain is the largest slice, strong performance there can carry an average well above your true standing while Resuscitation, the Injured Patient or Paediatric EM sit dangerously low. An examiner does not mark your average; the paper samples every domain, and a candidate who is 82% in their best domain and 55% in a heavily weighted one is not the 74% their home screen shows. Reading the distribution beneath the average is the single most valuable habit this checklist enforces.
The third mistake is stopping on a streak of familiar questions. A run of correct answers on items you have already reviewed measures memory of the bank, not readiness, and it is the most seductive false signal in the final fortnight. The exit criteria therefore turn on unseen, timed, mixed blocks and calibration against RCEM's own sample questions — never on a repeat-accuracy streak. If your last strong session was on questions you had seen before, it does not count toward the decision to stop.
Frequently asked questions
How do I know whether I have covered the full MRCEM SBA blueprint? You know when your coverage table is complete across all six domains in rough proportion to the blueprint weighting, each of the ten blind spots is attempted and clinician-reviewed, and every interpretation modality has unseen practice logged against it — not when a bank reports "100% complete." Bank completion measures the bank; blueprint coverage measures you against the RCEM curriculum. The completion-is-not-coverage distinction is the whole point of a coverage matrix, set out in our blueprint-coverage pillar; build the matrix and the gaps become visible.
Can one question bank be enough for MRCEM SBA? One bank can be enough to learn from if it is large, current and blueprint-mapped, but no single bank should be your only measurement, because you cannot test transfer on items you have already seen. The safer pattern is one bank to build knowledge and a second, unseen bank to verify it under exam conditions. If you use only one, at minimum ring-fence a block of its items you never review during learning, so you retain something genuinely unseen for a readiness check, and calibrate against RCEM's official samples.
What should I measure instead of my overall Q-bank percentage for MRCEM SBA? Measure first-attempt accuracy per domain on unseen, timed, mixed blocks, read against the blueprint weighting; your rate of high-confidence errors; your pace across a full 90-item paper; and your retention of older material. An overall percentage blends your strong and weak domains into one flattering number and, because the pass mark is Angoff-set per sitting, it does not even map onto a fixed target. As our standing caveat puts it, your Q-bank percentage is not your exam score — the distribution beneath the average is what tells you whether to stop.
When should I stop doing new MRCEM SBA questions? Stop when the checklist is green: coverage complete in blueprint proportion, first-attempt accuracy stable on unseen mixed blocks, all interpretation modalities practised, recency recorded, retention holding and calibration against official samples secure. At that point additional questions add fatigue rather than marks, and your time is better spent on spaced review, full-length simulation and rest. If any line is red, you are not stopping — you are targeting the red line. The trigger is measured completeness, never a round number of questions answered.
Which MRCEM SBA resource should I use for my weakest component? Diagnose the weakness first from your coverage table, then match the resource to it. If it is a knowledge or coverage gap in a specific domain, use a large, blueprint-mapped bank with mixed unseen blocks — the free iatroX MRCEM SBA bank maps to the 2021 RCEM curriculum and lets you filter to a weak domain and then test it unseen. If it is an interpretation gap (ECGs, images, radiographs), use targeted image and ECG resources plus RCEM sample material. If a domain resists all self-study, book teaching with an ED clinician; some gaps are conceptual, not volume-related, and more questions will not close them.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026; exam structure and curriculum weighting can change, and RCEM does not publish an exact per-sitting domain count, so treat the weightings here as blueprint proportions and confirm against the current RCEM curriculum and sample questions. Vendor and platform figures are labelled and dated where used. Disclosure: iatroX operates a free UK MRCEM SBA question bank that competes with the other banks in this market; this article is an exam-level coverage framework rather than a product pitch, and confines iatroX to the unseen-measurement and blueprint-audit jobs the checklist describes. This piece is intended as the exam-level hub for MRCEM SBA coverage; child platform articles should link here. Corrections are welcome via the feedback route on iatrox.com.
References: RCEM — MRCEM exams and the 2021 curriculum with official sample questions (rcem.ac.uk/mrcem-exams; RCEMLearning MRCEM SBA); iatroX MRCEM SBA landing page (iatrox.com/mrcem-sba); iatroX — "Your Q-Bank Percentage Is Not Your Exam Score" and the completion-is-not-coverage / blueprint-matrix pillar.