This is a working method, not a product ranking. It is for paediatric trainees using Pastest for the MRCPCH Applied Knowledge in Practice (AKP) paper who want their AI Tutor mode to build reasoning rather than hand over answers. The workflow addresses the applied, data-heavy AKP items specifically. Its one limitation is honesty about grounding: because Pastest does not publish how Tutor mode sources its answers, every output has to be verified against a named UK reference before you keep it.
What Pastest offers for MRCPCH Applied Knowledge in Practice right now
A compact, vendor-reported snapshot from the AKP product page, accurate on 19 July 2026 — verify before you buy.
| Item | Current state (vendor-reported, 19 July 2026) |
|---|---|
| AKP coverage | Dedicated MRCPCH AKP bank (FOP, TAS and Clinical sold separately) |
| Questions | Over 2,000 exam-style SBAs, plus 3 past papers (vendor-reported) |
| AI feature | "Tutor mode" — on-demand AI help while you answer; side panel; resets each question (vendor-reported) |
| Access & price | 3 / 6 / 12 months at approx. £89.99 / £149.99 / £179.99, plus a 48-hour free trial (vendor-reported — verify on the product page) |
Tutor mode is a real AI assistant, but with no disclosed model and no stated grounding, it is best treated as a fast explainer whose claims you check — which is exactly what this workflow builds in.
The AKP exam anchor
AKP is the second written MRCPCH stage: 120 single-best-answer questions across two 60-item papers, two hours 30 minutes each, both on the same day, computer-based, with no negative marking. It is deliberately applied — case scenarios, images, growth and observation charts, and data interpretation — at core-training level, and the RCPCH publishes the syllabus and official example questions that define it. Your budget is about two and a half minutes an item, much of it spent reading data before you decide. A workflow for AKP therefore has to train interpretation and next-step decisions, not just recall.
The prompt taxonomy: six jobs for the tutor
Give the tutor a defined job each time you open it. Six cover almost every missed AKP item: mechanism (why the finding occurs), discriminating features (what separates look-alikes), option elimination (why each distractor dies), guideline verification (which UK source and date), counterfactuals (what change flips the answer), and retrieval testing (turn the item into a question you must re-answer later). Never ask "what's the answer?" — that is the one prompt that destroys the item.
Twenty-five prompts, grouped by the six commonest error types
Keep these reusable and generic; fill the brackets with your own item, and never paste proprietary question text into a tool.
Error type 1 — Knowledge gap (you did not know the underlying fact)
- "State the single mechanism that explains why [finding] occurs in [condition], in two sentences."
- "Give the one-line definition and the key diagnostic criterion for [condition] as used in UK paediatric practice."
- "List the three commonest UK causes of [presentation] by age band, most likely first."
- "What is the physiological reason [parameter] is [high/low] in this scenario?"
Error type 2 — Discriminator miss (you could not separate two look-alikes) 5. "What single feature best discriminates [diagnosis A] from [diagnosis B] in this age group?" 6. "Give the two findings that would move me from [A] to [B], and the two that would move me back." 7. "Rank these options from most to least likely and justify each in one clause." 8. "Which one investigation result would most change my differential here, and in which direction?"
Error type 3 — Guideline or currency error (outdated or wrong-jurisdiction guidance) 9. "Name the specific UK guideline (NICE, CKS, SIGN or RCPCH) that governs this, and its publication or update date." 10. "Has the recommended management of [condition] changed in the last three years? If so, what changed?" 11. "Is this threshold or schedule the current UK one, or a US or historical value? State which." 12. "For [medicine], what does the SmPC/eMC give as the paediatric indication and dose range? Do not guess."
Error type 4 — Data or interpretation error (misread investigation, image or chart) 13. "Walk me through reading this [blood gas / growth chart / ECG] step by step, then state the one abnormality that matters." 14. "Which two numbers in this data set are decisive, and why?" 15. "What normal range should I use for [parameter] at this age, and is the value inside or outside it?" 16. "If I had only the image or chart and no text, what single diagnosis would it point to?"
Error type 5 — Management or next-step error (right diagnosis, wrong action) 17. "Given the diagnosis, what is the single best next step now, as opposed to the definitive treatment later?" 18. "What is the most urgent action here, and what would I do first if the child deteriorated?" 19. "Which option is safe but wrong-order, and why is it not the best next step?" 20. "What is the disposition decision — admit, refer, or discharge with safety-netting — and the trigger for each?"
Error type 6 — Careless or process error (misread stem, time pressure, poor elimination) 21. "Re-read this stem and tell me which single word or number determines the answer." 22. "Eliminate the three clearly wrong options and state the rule that kills each." 23. "What is the examiner testing here — recall, discrimination or management?" 24. "If I have 60 seconds left, what is the fastest safe way to choose between the final two?" 25. "Restate the question in one sentence so I can confirm I understood what is being asked."
The anti-answer-leak rule
Before Tutor mode opens, commit — in writing — to your chosen option and a one-line reason. This is the rule that keeps a question diagnostic. Reveal the key first and you convert a test into a read; your Pastest percentage rises, your readiness does not, and you have quietly proven why a bank percentage is not your exam score. No commitment, no tutor.
The jurisdiction-and-date check
Paste this reusable prompt on any management, dosing or threshold answer: "State whether this answer reflects current UK paediatric practice for the MRCPCH, name the specific NICE, CKS, SIGN, SmPC/eMC or RCPCH source, and give its date. If you cannot, say so." An AI that cannot name a dated UK source has given you a hypothesis, and you should verify it before trusting it — especially for immunisation schedules, resuscitation steps and drug doses, where non-UK or superseded values are the classic trap.
Create a misconception record
Every genuine miss earns four lines: the incorrect rule you were carrying (one sentence), the corrected rule with its source (one sentence), a transfer question that tests the same idea in a different scenario, and a review date three to seven days out. This converts a wrong answer into a scheduled, checkable piece of learning instead of a vague memory that "safeguarding is a weak spot".
Weekly verification sample
Once a week, take a small random sample — five to ten — of the tutor's answers you accepted and independently check them against official or primary sources. Log any discrepancy: what the tutor said, what the source says, and the corrected note. Two or three real discrepancies in a fortnight is your signal to tighten verification and lower default trust; zero over a month, and you can relax the sample size a little. This is the same disciplined calibration we describe for AI-graded feedback.
Worked example: a seven-day loop around clinical work
Pastest does the teaching; iatroX supplies unseen, timed measurement. No claims are made about either platform's algorithm.
- Monday (30 min): 20 Pastest AKP items, tutor off; flag misses.
- Tuesday (40 min): rework misses using the prompt taxonomy; write a misconception record for each; run the jurisdiction-and-date check on every management item.
- Wednesday (25 min): one Pastest textbook topic on your weakest domain; draft three retrieval questions.
- Thursday (40 min): fresh 30-item Pastest set in your two weakest areas.
- Friday (30 min): 20 unseen, timed iatroX items, no assistance; first-attempt accuracy only.
- Saturday (20 min): weekly verification sample; re-test due misconception records.
- Sunday: rest, or a timed 60-item mock near the exam.
Keeping Pastest and iatroX on separate jobs is the two-Q-bank rule; because iatroX's UK-core banks are free, they measure transfer without duplicating your Pastest drilling.
Decision checklist: continue, supplement, switch or stop
- Continue while unseen first-attempt accuracy rises and misconception records close on schedule.
- Supplement with an independent measurement bank if your Pastest percentage outruns your unseen score.
- Switch only for a measured, recurring coverage gap — never novelty.
- Stop anything that has not moved your unseen score in two honest weeks.
Frequently asked questions
Is Pastest enough for MRCPCH Applied Knowledge in Practice on its own? As a primary bank, yes for most candidates — a 2,000-plus AKP-specific pool with explanations and three past papers (vendor-reported, 19 July 2026) covers the blueprint. This workflow adds what a bank alone cannot: verified rework and unseen, timed measurement so you know the learning transfers.
Which MRCPCH Applied Knowledge in Practice component does Pastest not reproduce well? The untimed Tutor mode cannot reproduce the timed, mixed pressure of two 60-item, data-heavy papers in a day. Drill and learn in Pastest; rehearse the timed, unassisted paper separately.
How should I verify Pastest AI answers for MRCPCH Applied Knowledge in Practice? Commit to an answer first, run the jurisdiction-and-date prompt, then confirm against NICE, CKS, SIGN, the SmPC/eMC or the RCPCH example set. Treat any unsourced answer as provisional, and record real discrepancies in your weekly verification sample.
When should I stop using Pastest and move to mixed mocks? When fresh-set accuracy plateaus and misses stop clustering, shift weight to full-length, timed, mixed mocks — usually the final three to four weeks before the diet.
How should I combine Pastest with iatroX without duplicating practice? Different jobs, not repeated questions: Pastest for drilling and AI rework, iatroX for unseen, timed, unassisted measurement read on first-attempt accuracy. That is the two-Q-bank rule, and it keeps your percentage honest.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026; vendor-reported figures (question counts, prices, access periods and the Tutor mode description) were accurate to the Pastest AKP product page on that date and change without notice — verify the current figure on the product page. Disclosure: iatroX operates a competing UK question bank and clinical-knowledge platform; this workflow confines iatroX's role to unseen readiness measurement and Socratic rework of missed items — jobs Pastest does not itself claim — and does not offer iatroX as a replacement for a primary content bank. Corrections are welcome via the feedback route on iatrox.com.
References: RCPCH — Theory exams: structure and syllabi (rcpch.ac.uk/education-careers/examinations/theory/structure-syllabi) and the official AKP example questions; Pastest MRCPCH AKP (pastest.com/products/mrcpch-akp); iatroX — "Your Q-Bank Percentage Is Not Your Exam Score"; "Calibrating AI feedback"; the two-Q-bank rule; and the iatroX comparison hub.
Try a fresh MRCPCH AKP item in iatroX and run the verification prompt first →