This audit is for paediatric trainees who rely on BMJ OnExamination's analytics to judge how ready they are for MRCPCH Theory and Science — the science-and-mechanisms paper sat as the first written stage alongside Foundation of Practice. It explains what each metric does and does not mean, why a filtered feed makes raw percentages misleading, and what conditions a genuine readiness signal requires. The principal limitation to flag: BMJ OnExamination does not publish a proprietary predicted-score algorithm, so treat any single "readiness" number as a prompt to check the conditions below, not as a verdict.
What BMJ OnExamination offers for MRCPCH TAS right now
BMJ OnExamination lists a dedicated MRCPCH: TAS product and a combined FOP/TAS product. As of 19 July 2026 the standalone TAS bank advertises a vendor-reported 930+ questions mapped to the RCPCH curriculum, with the combined FOP/TAS bank advertising around 1,722 questions across both first-stage papers (vendor-reported, 19 July 2026); verify current counts on onexamination.com. Standalone TAS access runs from roughly £34.99 for one month to about £149.99 for twelve months, with a free trial of ten questions a day. The analytics-relevant features are per-topic performance, difficulty-level selection, a revision plan that targets weaker areas, peer comparison, daily question tracking and mock tests curated by the BMJ editorial team.
One honest qualification frames the whole audit: BMJ OnExamination does not advertise a proprietary adaptive-difficulty engine or a documented predicted-score model. Its personalisation is candidate-directed, and its headline analytics are per-topic accuracy and peer comparison rather than a black-box prediction. That is not a weakness for an analytics audit — it means the numbers are legible — but it does mean the interpretation is yours to get right. No proprietary-algorithm claims are made or needed.
MRCPCH Theory and Science: the official shape
TAS is one of the two first-stage written papers, sat alongside FOP and often on the same day, FOP in the morning and TAS in the afternoon. It is 100 single-best-answer questions in two hours, each worth one mark, no negative marking, computer-based at an exam centre or online — about seventy-two seconds per item. TAS is the science paper: physiology, pharmacology and therapeutics, genetics and dysmorphology, immunology, microbiology, biochemistry and metabolism, statistics and evidence-based practice, and the pathophysiology behind clinical paediatrics. The RCPCH publishes the TAS syllabus and official sample questions; those are the blueprint and the house style, and they outrank any vendor's tags or difficulty labels.
Define every metric before you trust it
Analytics dashboards use a shared vocabulary, and each metric answers a different question. Define them before you act on any of them:
- First-attempt accuracy — how often you got an item right the first time you ever saw it. This is the metric closest to real readiness, and the only one worth trending.
- Repeat accuracy — how often you get previously seen items right now. It climbs with exposure and mostly measures recognition, so never mistake it for readiness.
- Percentile or peer comparison — your standing against other users on the same items. It depends entirely on who else is in the pool and which items you have filtered to.
- Predicted score — a modelled estimate of exam performance. BMJ OnExamination does not document a proprietary predicted-score model for this bank, so if any such number appears, treat it with the percentage-article caveat below rather than as a forecast.
- Coverage — how much of the blueprint you have actually attempted, by domain. This is the metric candidates most often ignore and most need.
- Difficulty — the vendor's label for item hardness. Useful for building floors, but it is the vendor's interpretation, not the exam's.
- Time per item — your pace. At seventy-two seconds per question, this is a hard constraint, not a nicety.
Map each number on your BMJ OnExamination dashboard to one of these definitions before you let it change your plan. A metric you cannot define is a metric you cannot use.
Selection bias: why your percentage is not comparable
Here is the trap. The moment you filter to weak topics, retry flagged items, or let a revision plan steer you toward your gaps, your visible percentage is drawn from a biased, hardened sample and is no longer comparable with a mixed, unseen block. A feed weighted toward your weak areas depresses the number even as your true readiness rises; a feed of re-seen items inflates it even as your true readiness stalls. This is why two candidates with the same headline percentage can be miles apart, and why your own percentage last week and this week may not be comparable. The canonical caveat applies in full: your Q-bank percentage is not your exam score. Trend first-attempt accuracy on mixed, unseen items instead.
Blueprint audit: compare your attempted distribution with the domain weighting
Do not trust the home-screen average. Export or note your attempted-question distribution by domain and compare it with the TAS blueprint, because an average hides the shape underneath it. Use a matrix like this:
| TAS domain (illustrative) | Items attempted | First-attempt accuracy | Weekly floor |
|---|---|---|---|
| Pharmacology & therapeutics | — | — | set your own |
| Genetics & dysmorphology | — | — | set your own |
| Statistics & evidence-based practice | — | — | set your own |
| Immunology & microbiology | — | — | set your own |
| Physiology & biochemistry | — | — | set your own |
| Metabolic medicine | — | — | set your own |
The RCPCH does not publish fixed per-domain counts, so build the matrix from the syllabus; the reasoning is in our guide on why completion is not coverage. In TAS specifically, the quantitative domains — statistics and pharmacokinetics — are where a comfortable average most often hides a real gap.
Readiness test: the five conditions for a credible signal
A dashboard number is only a readiness signal when five conditions hold at once. The items must be unseen, so you are measuring transfer rather than recognition. The block must be timed at the real pace, so pacing is part of the test. The block must be mixed across domains, so interleaving is rehearsed. You must sit it with no assistance — no notes, no flags, no second guesses fed back in. And the sample must be large enough that the number is not noise; a forty-to-sixty-item mixed block is a reasonable floor. Miss any one of these and the number reverts to a study statistic, not a readiness estimate.
Algorithm override rules: force what the feed under-samples
Because the feed is candidate-directed and the revision plan steers toward your recent gaps, some material will be chronically under-sampled unless you force it. Override the feed deliberately for the low-volume, high-yield areas TAS candidates most often neglect: statistics and critical appraisal, calculations and pharmacokinetics, genetics and metabolic medicine, immunology, and any image or data interpretation. Schedule fixed weekly quotas in these areas regardless of what the dashboard nudges you toward, because an adaptive-feeling feed optimises for your recent behaviour, not for the blueprint. The override is how you stop a rising overall number from quietly hiding an untouched domain.
Worked dashboard example: from analytics to next week's quotas
Suppose your dashboard shows an overall first-attempt accuracy of 68%, but broken down: pharmacology 74%, physiology 71%, immunology 66%, genetics 58%, and statistics 49%, with statistics and genetics also your least-attempted domains and your average time per item running slightly over eighty seconds. Do not read "68%, nearly there." Read the shape: two quantitative domains are both weak and thin, and your pace is over budget. Next week's quotas write themselves — a domain floor of, say, forty unseen items each in statistics and genetics, a timed-block emphasis to pull your pace under seventy-two seconds, and a mixed unseen block at week's end to check whether first-attempt accuracy in the weak domains has moved. Note there is no pass prediction here, and there should not be: the dashboard sets quotas, it does not forecast outcomes.
A seven-day plan: BMJ OnExamination for one job, iatroX for transfer
Give BMJ OnExamination one job — legible per-topic practice and mock tests — and give iatroX a different job: unseen, timed transfer measurement on items it has never seen you attempt. No proprietary-algorithm claims are needed.
- Day 1 — baseline. Sit a mixed, timed, unseen block of forty to sixty items in the iatroX UK bank with no assistance; record first-attempt performance by domain.
- Days 2–4 — override and practise. Work your forced low-volume domains in BMJ OnExamination to their floors, tagging every miss by error type.
- Day 5 — calibration. Attempt the official RCPCH TAS sample questions under time and compare your reasoning with the model answers.
- Day 6 — fresh transfer. Sit a new timed iatroX block on different topics and trend your first-attempt accuracy, not your cumulative percentage.
- Day 7 — audit and plan. Re-run the blueprint matrix, check your pace, and set next week's quotas.
Continue, supplement, switch or stop
Decide on measurable gaps, not novelty or sunk cost. Continue with BMJ OnExamination if its legible analytics are moving your weak domains toward their floors and your first-attempt trend is rising on unseen items. Supplement with unseen iatroX measurement whenever your dashboard number and your transfer performance diverge, applying the two-Q-bank rule so you never repeat an item. Switch or add a second bank if a whole TAS domain stays thin and you cannot get above its floor. Stop trusting any single readiness number that fails the five-condition test above. Compare options on the iatroX comparison hub.
FAQ
Is BMJ OnExamination enough for MRCPCH Theory and Science on its own? It is a strong, legibly analysed single bank, but not a complete preparation on its own. TAS rewards secure coverage of the quantitative and mechanistic domains, so you also need the official RCPCH sample questions to calibrate house style and a separate pool of unseen, timed items to prove transfer rather than recognition. Treat its analytics as a planning tool, not a verdict, and verify its current question count on onexamination.com.
Which MRCPCH Theory and Science component does BMJ OnExamination not reproduce well? It does not reproduce the MRCPCH Clinical examination at all, and within the written stage its analytics reproduce readiness least well when the feed is filtered — a percentage drawn from re-seen or weak-topic items is not comparable with a mixed unseen block. The bank also cannot, by itself, force the low-volume statistics, genetics and calculation content unless you override the feed, so treat those as your responsibility rather than the dashboard's.
How many BMJ OnExamination questions should I complete per day for MRCPCH Theory and Science? Daily volume matters less than mix, override discipline and review, but a sustainable target around clinical work is roughly twenty to forty questions a day in timed blocks, weighted toward your forced low-volume domains, with every miss tagged and reviewed the same day. Chasing a high daily count of comfortable items inflates repeat accuracy while leaving your real gaps — usually the quantitative domains — untouched.
When should I stop using BMJ OnExamination and move to mixed mocks? Move to timed mixed blocks and full mocks once every TAS domain is above its coverage floor, your first-attempt accuracy in the quantitative domains has stabilised, and your pace holds near seventy-two seconds per item. At that point the per-topic analytics have given you what they can, and only mixed unseen blocks — sat under the five readiness conditions — can tell you whether the knowledge transfers under exam pressure, especially if you are sitting FOP and TAS the same day.
How should I combine BMJ OnExamination with iatroX without duplicating practice? Apply the two-Q-bank rule: give BMJ OnExamination the job of legible per-topic practice and give iatroX the separate job of unseen, timed measurement, and never run the same item through both. Because the dashboard is only as honest as the items behind it, keeping your iatroX block genuinely unseen is what lets it serve as the clean readiness check that a filtered feed cannot provide.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026; product figures described as vendor-reported were taken from public product listings on that date and change without notice — verify the current count, price and feature set on the vendor's own page. Disclosure: iatroX operates a competing UK question bank, so this article confines the iatroX role to unseen transfer measurement — a job the audited bank's analytics do not claim to replace. Corrections are welcome via the feedback route on iatrox.com. References: RCPCH examinations (rcpch.ac.uk/education-careers/examinations) and the official TAS sample questions; the BMJ OnExamination MRCPCH: TAS product page (onexamination.com); the iatroX comparison hub and the iatroX UK bank.
Run a fresh timed MRCPCH Theory and Science block in iatroX →