This is for emergency medicine trainees using Pastest's MRCEM Intermediate bank who want to turn its "Tutor mode" AI into structured learning rather than a hint button. It addresses the written SBA, and its principal limitation is worth stating plainly: the Tutor is a study aid trained on Pastest's own content, not a source of clinical truth. Emergency-medicine doses, thresholds and resuscitation algorithms change, so every safety-critical thing it tells you needs verifying against a current UK source before it goes into memory.
What Pastest offers for MRCEM SBA right now
| Item | Detail (vendor-reported, last checked 19 July 2026) |
|---|---|
| Bank | MRCEM Intermediate (SBA) — vendor-reported ~1,500 exam-style SBA questions mapped to the current RCEM syllabus; verify the live count on the product page. |
| AI feature | "Tutor mode" — vendor-reported on-demand AI, described as trained on Pastest's MRCEM content, that clarifies questions and explores options while you answer. |
| Mocks | A "tailored past paper" drawn from recent exams (vendor-reported). |
| Analytics | Self-configured performance insights; there is no machine-learning adaptive engine. |
| Price | Vendor-reported launch offer around £29.99 (3 months), £44.99 (6 months) and £59.99 (12 months); verify current pricing and access on the product page. |
| Components | Written SBA only. The MRCEM OSCE that follows is not covered. |
Pastest is an established provider with a large bank, and the Tutor is genuinely useful used correctly. Used carelessly — to be told the answer — it inflates your percentage and teaches little. The rest of this article is a discipline for using it well.
The exam you are actually preparing for
The MRCEM SBA is 180 single-best-answer items across two 120-minute papers of 90 questions — near 80 seconds per item — blueprinted to the 2021 RCEM curriculum (SLOs 1 and 3–7). The six domains carry indicative weights of: complex, stable patient 55; resuscitation and the unstable patient 40; injured patient 30; paediatric emergency medicine 25; procedural skills 20; complex and challenging situations 10. The complex-stable domain is the largest and most under-revised; resuscitation is over-practised; paediatrics and procedural skills are under-sampled by a self-selected feed. Confirm the current blueprint and official sample questions on the RCEM website.
The anti-answer-leak rule
The single most important rule: commit to an answer and a one-line rationale before you open the Tutor. Write your chosen option and your reason. Only then open Tutor mode — and even then do not ask "what is the answer?"; ask for the mechanism, the discriminating feature, or why a specific distractor is wrong. If the Tutor reveals the answer before you commit, you convert active recall into passive reading. The commitment step is what makes every missed question a learning event. This is the same answer-leakage risk described in the pillar on auditing an AI medical-exam tutor.
A prompt taxonomy for missed questions
Good prompts are reusable and generic — they never quote proprietary question text. Aim each at one of six purposes: mechanism (why the answer is correct at first principles); discriminating features (the detail separating best from runner-up); option elimination (why each distractor is plausible but wrong); guideline verification (what current UK practice says, and how current); counterfactuals (how the answer changes if one variable changes); and retrieval testing (a fresh question proving transfer).
Twenty-five reusable prompts, grouped by error type
Knowledge gap (mechanism and yield):
- "Explain in three sentences the pathophysiology that makes the correct answer correct, then name the one fact that separates it from the runner-up."
- "What first principle did I need to answer this without memorising the vignette?"
- "List the three highest-yield facts about this condition an MRCEM SBA is likely to test, and the type of UK source for each."
- "Summarise this drug's mechanism and its single most exam-relevant consequence for emergency management."
Misread stem (discriminators): 5. "Which exact words in a stem like this change the answer, and which detail do candidates most often skim?" 6. "Give me two near-identical stems where changing one detail flips the best answer." 7. "State the classic discriminating feature between these two conditions and how examiners phrase it." 8. "Rewrite the key line of a stem of this type as tersely as an examiner would."
Premature closure (option elimination): 9. "For a five-option SBA on this topic, explain in one line each why every distractor is plausible but wrong." 10. "Rank these five options most to least defensible and state the rule separating first-best from second-best." 11. "Which single distractor is designed to catch someone who anchors early, and why?" 12. "Teach me a general elimination heuristic for 'most appropriate next step' questions in emergency medicine."
Guideline error (currency): 13. "State the current UK guidance for this presentation, name the source type (NICE, CKS, Resuscitation Council UK, RCEM or SmPC/eMC) and give the date or version." 14. "Has the recommended drug, dose or threshold here changed recently in UK practice? If unsure, say so and tell me what to check." 15. "Give the current UK first-line drug, dose and route for this emergency as in the SmPC/eMC, and flag if it may be out of date." 16. "Confirm whether this reflects UK emergency-medicine practice rather than another country's, and cite the source type."
Calculation error: 17. "Walk through this weight-based paediatric dose calculation step by step, showing units at each stage." 18. "Give me three fresh calculation drills of this type — infusion rate, drug dose and unit conversion — and mark my answers." 19. "What is the most common arithmetic slip on this calculation, and how do I guard against it under time pressure?"
Time-pressure and retrieval testing: 20. "Ask me one fresh SBA-style question on the same concept but a different scenario, and wait for my answer before feedback." 21. "Quiz me on the discriminating feature we just discussed in a new context; do not reveal the answer until I respond." 22. "Give me three rapid recall prompts on this topic and mark my free-text answers against UK guidance." 23. "Set a delayed-recall check for three days' time on this concept and tell me what to test." 24. "Without repeating this question, test whether I can apply the rule to a paediatric version of the presentation." 25. "Give me a 60-second version of this decision so I can rehearse it at exam pace."
The jurisdiction and currency check (doses and algorithms)
Emergency medicine is where an out-of-date answer is most dangerous, because doses, defibrillation energies, sepsis thresholds, anticoagulation reversal and paediatric resuscitation figures all change. Make this reusable prompt mandatory for any safety-critical output:
"Before I accept this, state three things: whether it reflects current UK emergency-medicine practice for the MRCEM SBA; the source type (NICE, CKS, Resuscitation Council UK, RCEM or SmPC/eMC); and the date or version. If you cannot confirm currency, say 'verify against the primary source' and name which one."
Then actually open it. Verify doses against the SmPC/eMC, resuscitation and paediatric algorithms against Resuscitation Council UK, and pathways against NICE or CKS. The Tutor is trained on content of a certain vintage; it does not know what changed last month. Treating it as a prompt to check rather than a source to trust is the difference between safe revision and confidently learning something wrong. For how much to trust any automated explanation, see the pillar on calibrating automated feedback.
Build a misconception record
For every genuinely wrong answer, write a four-line record — short by design, so you actually do it: the incorrect rule you were carrying, in one sentence; the corrected rule, in one sentence, with its UK source and date; one transfer question in a fresh scenario, ideally unseen in iatroX; and one review date to re-test from memory. A running record beats any number of re-read explanations, because it isolates the exact rule that failed and schedules its correction.
Run a weekly verification sample
Once a week, take a small random sample of Tutor outputs — five to ten explanations — and check them independently against official or primary UK sources. Log any discrepancy: a dose that has moved, a threshold out of date, a step that is not current UK practice. If the discrepancy rate is trivial, keep trusting the Tutor for orientation while still verifying safety-critical facts. If it is material, downgrade it to a discussion partner and lean on primary sources for anything you will act on — the audit discipline the AI-tutor pillar recommends, on a weekly cadence.
Worked example: seven days around clinical work
Pastest for structured learning and Tutor work, iatroX for fresh, unseen measurement — no proprietary-algorithm claims.
| Day | Pastest (one job) | iatroX (unseen measurement) |
|---|---|---|
| Mon | 40 questions; commit-then-Tutor on every miss | — |
| Tue | Write misconception records; run currency checks | 1 fresh 20-item mixed block, timed |
| Wed | 40 questions weighted to paediatric + procedural | — |
| Thu | Weekly verification sample (5–10 Tutor outputs) | 1 fresh 20-item mixed block, timed |
| Fri | 30 questions; complex-stable focus | — |
| Sat | Tailored past paper, timed | — |
| Sun | Spaced recall of the week's records | Compare unseen accuracy vs baseline |
Three mistakes this workflow is designed to stop
First, answer-leak — opening the Tutor before committing, turning recall into reading. Second, trusting stale doses and algorithms — accepting an emergency-medicine figure without a currency check, the error most likely to be dangerous in real practice as well as costly in the exam. Third, confusing a Tutor-inflated percentage with readiness — measure on unseen questions instead, an application of the two-Q-bank rule, and remember a bank percentage is not an exam score.
Decision checklist: continue, supplement, switch or stop
- Continue Pastest where first-attempt accuracy on fresh questions is rising and your verification sample shows the Tutor reliable enough for orientation.
- Supplement with unseen measurement as soon as your percentage stops moving you — learn on Pastest, measure elsewhere.
- Switch or add a second bank only for a measurable coverage gap, such as thin paediatric sampling, not novelty.
- Stop expanding when coverage is even across the six domains, unseen timed scores are stable, pacing holds inside 80 seconds per item, and remaining errors are careless rather than conceptual.
Bottom line
Pastest's MRCEM Intermediate bank with Tutor mode is a strong option for the written SBA, provided you impose two disciplines it will not impose for you: commit before you open the Tutor, and verify every safety-critical dose and algorithm against a current UK source. Do those, keep a misconception record, run a weekly verification sample, and reserve unseen questions for measurement — and the Tutor becomes a teacher rather than a crutch.
Frequently asked questions
Is Pastest enough for MRCEM SBA on its own? For learning content, its MRCEM Intermediate bank can carry most of your written-SBA revision — vendor-reported at around 1,500 SBA questions mapped to the RCEM syllabus. For measuring readiness, no single bank is enough alone, because once you have reviewed its questions your percentage reflects recall rather than mastery. Keep a separate source of unseen questions.
Which MRCEM SBA component does Pastest not reproduce well? As a written single-best-answer resource it does not reproduce the MRCEM OSCE, and no MCQ bank rehearses live resuscitation, team working or hands-on procedures. Within the written paper, Tutor-assisted practice does not reproduce exam conditions unless you deliberately do timed, unassisted mixed blocks, because using the Tutor on every question removes the time pressure the real paper imposes.
How should I verify Pastest AI answers for MRCEM SBA? Make the jurisdiction-and-currency prompt mandatory for any safety-critical output, then open the primary source it names: doses against the SmPC/eMC, resuscitation and paediatric algorithms against Resuscitation Council UK, and pathways against NICE or CKS. Run a weekly sample of five to ten explanations checked against those sources and log discrepancies, so you know empirically how far to trust it.
When should I stop using Pastest and move to mixed mocks? Move from Tutor-assisted, topic-filtered practice to timed mixed mocks once every domain is sampled in roughly blueprint proportion, first-attempt accuracy is stable across domains, and pacing sits comfortably inside 80 seconds per item. The Tutor is a build-phase tool; the closing phase should look like the exam — mixed, timed and unassisted.
How should I combine Pastest with iatroX without duplicating practice? Use Pastest as your learning bank and iatroX only for fresh, unseen measurement, so you never test yourself on questions you have already worked through with the Tutor. Learn and review on Pastest, then run blueprint-sampled timed blocks in iatroX to see where you stand — the division of labour the two-Q-bank rule describes.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026. Exam format and blueprint figures reflect the published RCEM MRCEM SBA structure at that date; verify the current blueprint and official sample questions on the RCEM website before relying on any number. Pastest question counts, prices, the "Tutor mode" description and mock details are vendor-reported and should be confirmed on the product page, as launch pricing and features change.
Disclosure: iatroX operates its own question bank and Socratic Tutor, so it competes with the resources discussed here. This is written as an exam-level method rather than a product pitch, and iatroX is positioned only for the job it suits — fresh, unseen questions for measurement and a tutor for missed items — not as a replacement for your primary MRCEM bank or for the OSCE. Corrections are welcome through the feedback route on iatrox.com. Compare resources on the iatroX comparison hub.
References: Royal College of Emergency Medicine — MRCEM SBA format, 2021 curriculum, blueprint and official sample questions (rcem.ac.uk); Pastest MRCEM Intermediate product page (pastest.com); NICE, CKS, SIGN, SmPC/eMC, Resuscitation Council UK and NHS content for dose and algorithm currency; iatroX, why your Q-bank percentage is not your exam score and how to audit an AI medical-exam tutor.