This is a decision framework, not a league table. The right MRCEM SBA (Intermediate) resource depends on who you are: first attempt or retake, how many weeks you have, what you can spend, and whether your real gap is knowledge, pacing or blueprint coverage. No single bank is best for everyone. Below is a decision tree that routes each candidate profile to a minimum stack, a platform-to-job matrix and three worked plans. Measure where you are first; then follow the branch.
The exam you are actually preparing for
MRCEM SBA (Intermediate) is 180 single-best-answer questions delivered as two papers of 90, each lasting two hours, with an hour's break between them. That is roughly 80 seconds per item — fast enough that pacing is a skill in its own right. The blueprint is mapped to the 2021 RCEM Emergency Medicine curriculum (the Year 1–3 Specialty Learning Outcomes, SLO1 and SLO3–7), and the item weighting is public:
| Domain | Approx. items (of 180) |
|---|---|
| Complex/Stable (non-acute) patient | 55 |
| Resuscitation | 40 |
| Injured patient | 30 |
| Paediatric emergency medicine | 25 |
| Procedural skills | 20 |
| Complex/challenging situations | 10 |
You must hold MRCEM Primary first, and the OSCE follows later. There is no negative marking, and the pass mark is set by a modified Angoff method (with a standard-error adjustment). That weighting is the map every resource in this article is judged against: a bank that is superb on trauma but thin on the complex stable patient is under-serving the largest single slice of the paper.
Start with a baseline, not a purchase
The first input to every branch below is a measured domain profile, not a product. Before you buy anything, sit a short timed mixed set — the official RCEM sample questions plus a small block of unseen items — under the clock, then score it by domain and lay the result against the weighting above. That tells you three things a sales page cannot: which domains are already at floor, where your accuracy collapses under time, and whether your problem is knowledge or pace. Your percentage on a bank you have half-memorised will not tell you any of that; an unseen, timed, domain-scored baseline will.
Segment yourself: six learner profiles
| Profile | Principal need |
|---|---|
| First attempt, solid clinical exposure | Breadth and calibration to question style |
| Retake after a near miss | Targeted domain repair and unseen re-measurement |
| Busy trainee revising around shifts | High-yield micro-blocks and spaced error review |
| Weak foundations / early in training | Teaching or reference first, then bank breadth |
| Strong knowledge, poor pacing | Timed mixed blocks, not more content |
| Strong MCQ technique, weak clinical judgement | Application and management items, case reasoning |
Most candidates are a blend, but one need usually dominates. Name yours before you shop.
The minimum stack — and why more is usually worse
For most candidates the effective stack is small:
- One primary Q-bank for breadth and explanations.
- Official calibration material — the RCEM sample SBAs and the regulations and information pack — as the format gold-standard.
- One teaching or reference source, and only for a domain that is genuinely weak.
- One modality tool where relevant (ECGs, imaging, paediatric dosing) if that is your specific gap.
Adding a second large bank early tends to hurt rather than help: you dilute your time, you cannot finish either bank, and — most importantly — you contaminate your own readiness signal, because a rising percentage across two overlapping banks measures recognition of seen items, not transfer to unseen ones. The disciplined way to add a second resource is set out in the two-Q-bank rule: keep one bank for learning and reserve a second, unseen bank purely for measurement. That is the role iatroX's free UK-core content is built to play here — the neutral measurement layer, not a third pile of content to grind.
Budget bands (verify every price on the day you buy)
| Band | What it looks like |
|---|---|
| Free / low-cost | RCEM official sample material + a free or starter bank + iatroX free UK-core blocks for unseen measurement |
| One premium resource | A single paid bank chosen because its best job matches your dominant gap — nothing else until you have a measured reason |
| Comprehensive stack | Primary paid bank + a reserved measurement bank + one targeted teaching source for a proven weak domain |
Prices and access periods change; several MRCEM products quote "from" pricing that depends on the access window. Confirm the current figure and the length of access on the product page before you pay — do not budget from a number in an article, including this one.
Time bands: what to run, and what to omit
The commonest mistake is loading every plan with everything. Each band below states what to leave out.
- Under four weeks (triage). Run mixed, timed blocks from day one, review errors by type, and calibrate against the official sample paper. Omit new topic teaching and do not start a 2,600-item bank you cannot finish; a half-worked giant bank is worse than a fully-worked small one.
- Four to twelve weeks (the standard build). First pass through one primary bank, ordered by your weak domains, then convert to mixed timed blocks in the last third. Reserve an unseen pool from the start. Omit second banks until the reserved pool shows a plateau.
- More than twelve weeks (foundations then breadth). Front-load teaching or reference only where a domain is genuinely weak, then bank breadth, then mixed simulation. Omit the temptation to finish everything — coverage of the weighting matters more than completion of the catalogue.
Platform-to-job matrix
Figures below are vendor-reported or aggregator-listed and were last checked on 19 July 2026; verify the live count, price and access period on each product page before relying on them.
| Resource | Best job for MRCEM SBA | Covers Intermediate SBA? | Notes |
|---|---|---|---|
| RCEM official sample material | Format calibration (gold-standard) | Yes (official) | Free sample SBAs + regulations pack; finite, so calibration not volume |
| Pastest | Analytics + integrated tutor/anatomy | Yes | ~1,500 SBAs reported; 3D anatomy, past papers, a "Tutor Mode"; established brand |
| MRCEM Success | Volume + illustrated explanations | Yes | ~1,800+ SBAs reported; guideline links |
| MRCEM Exam Prep | Volume + written tutorials | Yes | 2,600+ SBAs and 310+ tutorials reported; from £35; see the dedicated child audit |
| FRCEMtutor | Teaching notes + timed practice | Yes | ~1,250 questions reported; timed mode, analytics |
| Bromley Emergency (course) | Taught teaching for weak domains | Yes (course) | Two-day course + practice questions |
| BMJ OnExamination | Primary basic sciences only | No — Primary, not Intermediate SBA | ~590+ Primary-style items reported (anatomy/physiology/pharmacology, etc.); see flag below |
| iatroX | Unseen measurement + Socratic Tutor | Yes (measurement layer) | Free UK-core blocks; used to re-measure transfer, not as your only content |
Map your dominant need (left of the profile table) to the best-job column — that is the branch, not the biggest catalogue.
Honest coverage flag: not every "MRCEM" product is an SBA product
One trap is worth stating plainly. BMJ OnExamination's emergency-medicine product is built for the MRCEM/FRCEM Primary examination — the basic-sciences layer (anatomy, physiology, pharmacology, microbiology, pathology, statistics and evidence-based medicine) — not for the Intermediate SBA. It is a strong option for Primary, but if you buy it expecting Intermediate SBA clinical coverage you will have bought the wrong stage. Always confirm the exam component a product targets, in writing, before paying; the word "MRCEM" in a product name does not tell you which of the two written stages it covers.
Cannibalisation guardrail
This article is the hub for the "which resource" decision. It deliberately does not reproduce the long product write-ups — for the detailed fidelity and coverage evidence on a specific bank, follow the dedicated child audits (for example, the MRCEM Exam Prep question-style audit and workflow) and the comparison hub. Choose on the job to be done and the measured gap, never on catalogue size or on money already spent.
Three worked plans
Profile A — retake, eight weeks, near miss with weak Resuscitation and Paediatric EM. Weeks 1–2: domain-filtered blocks in Resus and Paeds EM on the primary bank; error-code every miss with one corrective action. Weeks 3–5: widen to all domains in weighted proportion; begin timed mixed blocks twice weekly; keep a reserved unseen pool untouched. Weeks 6–8: mixed timed papers only; re-measure the two weak domains on unseen items. Exit criterion: both weak domains at or above your overall accuracy on unseen, timed blocks, and pace at or under 80 seconds per item — not a percentage on the primary bank.
Profile B — busy trainee, six weeks around shifts. Micro-blocks of 15–20 items on shift days, one longer mixed block on days off. Error review capped at three high-value misses per session. Exit criterion: every domain touched to floor, unseen accuracy stable across two consecutive weekends.
Profile C — strong knowledge, poor pacing, four weeks. Almost no new content: mixed timed blocks daily, deliberately practising the 80-second decision. Review only timing failures and high-confidence errors. Exit criterion: completes a 90-item block within time with accuracy holding to the last ten items.
None of these predicts a pass; they define measurable exit criteria that make "ready" an observation rather than a feeling.
Evidence hierarchy (how to weigh what you read)
Rank your sources: official RCEM material first for format and blueprint; primary UK guidance (NICE, CKS, SIGN, and the Summary of Product Characteristics via the eMC) for clinical content; vendor pages for product facts (counts, price, access), always dated and labelled vendor-reported; and independent user reports last, for user experience only. When a vendor claim and an official document disagree, the official document wins.
Reading your baseline: three common misreads
A domain-scored baseline only helps if you read it correctly, and three misreads are common. The first is treating a high overall score as readiness when it is carried by two strong domains and hides a floor in the largest one; always read by domain against the weighting, never as a single number. The second is mistaking a slow but accurate baseline for a knowledge problem when it is a pacing problem — if your accuracy is fine but you ran out of time, more content is the wrong prescription. The third is reading a single weak block as a trend; one timed set is noisy, so confirm a weak domain on a second unseen block before you rebuild your plan around it.
When a second bank is actually justified
Adding a second bank is justified in exactly one situation: your reserved unseen pool has plateaued, a specific domain remains below floor on those unseen items, and your primary bank has no untested questions left in that domain. That is a measured trigger, not a feeling of insufficiency. If you cannot state which domain, on which unseen evidence, at what accuracy, you are buying reassurance rather than coverage — and you will pay for it in corrupted calibration, because a second overlapping bank tends to inflate your percentage faster than it fills a genuine gap.
A twenty-minute way to use this tree
If you have twenty minutes and a decision to make, do this. Spend ten minutes on a short timed unseen set and score it by domain. Spend five minutes naming your dominant profile from the six above and your time band from the three. Spend the last five mapping that to a minimum stack and a single best-job resource from the matrix, and writing one exit criterion you will hold yourself to. That is the whole method: measure, classify, choose the smallest sufficient stack, and define what "done" looks like before you start — then revisit only when the unseen numbers, not your nerves, tell you to.
Bottom line
There is no universal best MRCEM SBA bank; there is a best next action for your profile, budget and calendar. Measure a domain baseline, pick the smallest stack that closes your dominant gap, reserve an unseen pool for honest re-measurement, and switch to timed mixed blocks once your domain floors are met. Use the matrix to match need to job, and confirm every count and price on the day.
FAQ
How do I know whether I have covered the full MRCEM SBA blueprint? Coverage is not "questions finished" — it is evidence, per domain, that you can answer unseen items at the pace the paper demands. Build a six-row table using the official weighting (Complex/Stable 55, Resuscitation 40, Injured 30, Paediatric EM 25, Procedural 20, Complex/Challenging 10) and, for each domain, record items attempted, first-attempt accuracy on unseen blocks, the date last reviewed and your confidence. A domain is covered only when unseen, timed accuracy sits at or above your target and has held on re-measurement — not when the progress bar reads 100%.
Can one question bank be enough for MRCEM SBA? For many first-attempt candidates with solid clinical exposure, yes — one well-matched primary bank plus the official RCEM sample material can be sufficient, provided you also measure yourself on items you have not seen. The failure mode of a single bank is not missing content; it is that your rising percentage measures recognition of that bank's items rather than transfer to new ones. Reserve an unseen pool (iatroX's free UK-core blocks are built for exactly this) so your readiness signal stays honest.
What should I measure instead of my overall Q-bank percentage for MRCEM SBA? Measure four things: unseen, timed accuracy by domain against the weighting; pace (are you holding roughly 80 seconds per item to the end of a 90-item block?); your high-confidence error rate (items you were sure of and got wrong — the most dangerous category); and retention on spaced re-tests. A single headline percentage hides all four. There is a longer treatment of why the percentage misleads in the linked article on Q-bank scores.
When should I stop doing new MRCEM SBA questions? Stop adding new items when every domain is at floor on unseen timed blocks, your pace is stable, your high-confidence error rate is low, and re-tested topics are holding. At that point new questions have diminishing returns and your remaining time is better spent on full timed simulation and rest. Doing new questions past this point tends to buy reassurance, not marks.
Which MRCEM SBA resource should I use for my weakest component? Match the resource to the failure type, not the domain label. If the weakness is knowledge, use a teaching or reference source, or a bank with strong written explanations for that domain. If it is pacing, use timed mixed blocks — more content will not help. If it is interpretation (ECGs, imaging, paediatric dosing), use a modality-specific tool. Then re-measure that component on unseen items to confirm the repair transferred.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026. Product figures (question counts, prices, features) are vendor-reported or aggregator-listed as at that date and change frequently; verify each on the product page before relying on it. Disclosure: iatroX operates an MRCEM SBA question bank and Socratic Tutor and therefore competes with several products named here; this article confines iatroX's role to unseen measurement and to jobs the audited products do not claim, and names honestly where another resource is the better fit — including where a product covers Primary rather than Intermediate SBA. Corrections are welcome via the feedback route on iatrox.com.
References: Royal College of Emergency Medicine — MRCEM exams and the MRCEM SBA Regulations and Information pack with sample questions (rcem.ac.uk); vendor product pages (Pastest, MRCEM Success, MRCEM Exam Prep, FRCEMtutor, BMJ OnExamination); iatroX MRCEM SBA bank and comparison hub; "Your Q-Bank Percentage Is Not Your Exam Score" and the completion-is-not-coverage blueprint-matrix pillar.