This audit is for geriatric medicine higher specialty trainees, usually ST4 and above, using BMJ OnExamination for the SCE and trying to read its analytics honestly. BMJ OnExamination does publish a Geriatric Medicine SCE bank, and its dashboard can guide revision. The principal limitation is that a personalised, weakness-weighted feed optimises for your recent errors, not for the official blueprint, so its home-screen average can drift out of step with a mixed, unseen paper. This article defines every metric, exposes the selection bias, and audits the feed against the Federation blueprint. iatroX is used only as the unseen-measurement layer.
First, a word on "adaptive". BMJ OnExamination markets revision plans, a difficulty filter and peer comparison, but it does not publish an adaptive-difficulty algorithm, so this audit treats the engine as weakness-weighted personalisation rather than a black box, and makes no proprietary-algorithm claims. Understanding what that engine is optimising, your improvement on weak topics, rather than blueprint coverage or exam-day calibration, is the whole point of reading its numbers correctly.
What BMJ OnExamination offers for SCE Geriatric Medicine right now
Vendor-reported, checked 20 July 2026; verify on the product page before purchase.
| Item | BMJ OnExamination SCE Geriatric Medicine (vendor-reported, 20 July 2026) |
|---|---|
| Question count | 245+ questions |
| Price and access | From £69.99 for one month up to £194.99 for twelve months |
| Personalisation | "Select Questions" difficulty filter, revision plans that target weaker topics |
| Assessment | Mock tests curated by BMJ's editorial doctors |
| Social and feedback | Group learning with daily leaderboards, peer comparison |
| Adaptive-difficulty algorithm | Not published; treat the feed as weakness-weighted |
One honest observation up front: a bank of around 245 questions is modest relative to a 200-question exam drawn from a broad blueprint, so per-domain depth is limited and a weakness-weighted feed has little spare material to reach every small domain. That makes the coverage audit below more important here than for a larger bank, and it strengthens the case for an unseen second layer.
The exam you are actually sitting
The SCE in Geriatric Medicine is two papers of 100 best-of-five questions, 200 in total, three hours per paper, one day, computer-based on Surpass, one mark per correct answer, no negative marking. The pass mark is set per diet by criterion-referenced standard-setting, not a fixed percentage.
The Federation blueprint is unusually broad because geriatric medicine is a generalist specialty. The largest syndrome domains are falls and poor mobility at about 16 questions and stroke care at about 15, followed by rehabilitation and transfers of care at about 14, dementia at about 12, and continence, palliative care and orthogeriatrics with osteoporosis at about 10 each. Intermediate and long-term care sits near 9, delirium near 8, old age psychiatry near 7, and tissue viability near 6, with geriatric assessment, nutrition and surgical liaison contributing smaller shares. On top of these syndrome domains, the blueprint also samples acute illness and chronic disease across the medical systems, cardiovascular, neurology, infection and respiratory each carrying several questions. This spread is exactly what a weakness-weighted feed can distort.
Define every metric before you trust it
The dashboard shows several numbers, and each answers a different question.
- First-attempt accuracy: the share of questions correct the first time you saw them. This is the honest signal of current knowledge and the one to watch.
- Repeat accuracy: accuracy on questions you have already attempted. This is inflated by memory of the item and should never be read as readiness.
- Percentile or peer comparison: your rank against other users of the same bank, a self-selected group that is not the exam cohort, so treat it as directional only.
- Coverage or completion: how much of the bank you have attempted, which is not the same as blueprint coverage and does not measure learning.
- Difficulty tag: the bank's own label, useful for filtering but not standardised to the exam's difficulty.
- Time per item: your pace, which must be reconciled with roughly 1.8 minutes per question in the real paper.
- Predicted score, if shown: any such figure is a vendor estimate built on that bank's population and item pool, not a probability of passing, and should be read with caution rather than as a forecast.
Selection bias: why the feed and the paper disagree
A weakness-weighted feed deliberately over-samples your weak areas. That is good for learning but it makes your running percentage incomparable with a mixed, unseen block, because you are answering a harder-for-you, non-representative sample. Two candidates with identical knowledge can show very different feed percentages depending on how much the engine has pushed them into their weak topics. This is the central reason a rising home-screen average can coexist with a flat performance on a balanced mock, and why you should anchor readiness to unseen mixed blocks rather than the feed. The iatroX explainer on why your Q-bank percentage is not your exam score makes the same point in general terms.
Blueprint audit: compare your attempts with the official weighting
Once a week, export or note your attempted-question distribution by topic and lay it beside the Federation blueprint weighting. You are looking for domains where your share of attempts is far below the blueprint share. Given a 245-question bank, it is easy to under-attempt tissue viability, continence, nutrition or old age psychiatry simply because there are few items and the feed keeps steering you elsewhere. Where a domain is under-represented, override the feed. The structured method for this comparison is the iatroX blueprint-coverage matrix.
Readiness test: the conditions for a credible signal
A number only means something under the right conditions. A credible readiness signal in geriatric medicine requires questions that are unseen, timed at exam pace, mixed across the blueprint rather than filtered, answered with no external assistance, and drawn from a sample large enough to be stable, at least 40 to 50 items, ideally more. A high repeat-accuracy figure on a filtered block fails every one of those conditions. Run the readiness check in iatroX or in a fresh BMJ mock you have not seen, so it stays independent of the items you have drilled.
Algorithm override rules: force what the feed under-serves
Because the feed optimises for your recent errors, you must manually force the material it under-serves: the low-volume syndrome domains such as tissue viability, continence and nutrition; ethics and medicolegal content around capacity, consent and deprivation of liberty, which is central to geriatric practice but thin in many banks; any calculation or data-interpretation items; and image or data-based questions. Set a weekly quota for these and complete it regardless of what the revision plan proposes. When the feed and the blueprint disagree, the blueprint wins.
Worked dashboard example: turning analytics into next week's quotas
Suppose your dashboard shows first-attempt accuracy of 74 per cent overall, but by domain: falls 80 per cent, stroke 78 per cent, dementia 76 per cent, delirium 70 per cent, continence 58 per cent, tissue viability 52 per cent, old age psychiatry 60 per cent, and only 5 attempted items in nutrition. Coverage stands at 61 per cent of the bank, peer percentile at the 55th, and pace at 2.1 minutes per item, slower than exam pace.
Translate that into quotas, not a pass prediction. Next week: 20 questions on tissue viability and 20 on continence to lift the two weakest domains and their coverage; a nutrition block to clear the near-empty domain; a 40-item timed mixed block held to 1.8 minutes per item to attack pace; and a 15-item ethics and capacity set forced regardless of the feed. Leave falls, stroke and dementia to spaced maintenance. Nothing here forecasts a result; it converts the numbers into actions.
A seven-day plan around clinical work
A realistic week using BMJ OnExamination for the personalised content job and iatroX for unseen measurement. No proprietary-algorithm claims are made.
- Monday: the two weakest domains from your audit, tagged by error type.
- Tuesday: a short source read on the worst misses, then 15 unseen iatroX items on the same principles.
- Wednesday: your forced quota, tissue viability, continence, nutrition, ethics and capacity.
- Thursday: a 40-item timed mixed block or BMJ mock section at exam pace; review by error type.
- Friday: stroke, falls and dementia maintenance, kept light.
- Saturday: an unseen, timed iatroX readiness block; log first-attempt accuracy and pace.
- Sunday: rest or a light spaced review.
Decision checklist: continue, supplement, switch or stop
| Situation | Action |
|---|---|
| Attempt distribution far from blueprint | Continue, but override the feed to force under-served domains |
| Home-screen average high, unseen mixed accuracy low | Supplement with iatroX unseen blocks; trust the mixed signal |
| Bank nearing exhaustion at 245 items | Supplement with a second bank or iatroX for fresh volume |
| Coverage, unseen accuracy, pace and retention all stable | Stop adding material; rehearse full mocks and rest |
Frequently asked questions
Is BMJ OnExamination enough for SCE Geriatric Medicine on its own? Its vendor-reported bank of around 245 questions is a useful revision aid, but at that size it is thin for a broad, generalist blueprint, so for most candidates it is better treated as one component than as a sole resource. Pair it with a larger specialty bank or with unseen iatroX blocks so that coverage of the low-volume domains and honest measurement are both covered.
Which SCE Geriatric Medicine component does BMJ OnExamination not reproduce well? A weakness-weighted bank tends to under-serve the low-volume syndrome domains and the ethics, capacity and multidisciplinary-care material that defines geriatric practice, and no MCQ bank reproduces the integrated, multimorbidity reasoning the exam probes at full density. It also does not reproduce the exact Surpass interface and two-paper stamina, which you should rehearse in timed mocks.
How many BMJ OnExamination questions should I complete per day for SCE Geriatric Medicine? Around clinical work, roughly 30 to 40 questions per day with full review is sustainable, though a 245-question bank will be exhausted quickly at that rate, so plan to move to unseen material for measurement once you have worked through it. As always, thorough review of fewer questions beats skimming a larger block.
When should I stop using BMJ OnExamination and move to mixed mocks? Move to mixed, timed blocks once your attempted-question distribution matches the blueprint reasonably well and your first-attempt accuracy on filtered blocks is stable, then use the final one to two weeks for full-length, timed mocks. With a smaller bank you will often reach this point relatively early, which is a signal to bring in fresh, unseen material.
How should I combine BMJ OnExamination with iatroX without duplicating practice? Give each a single role: BMJ OnExamination is your personalised content and analytics engine, and iatroX is your unseen, timed measurement and spaced-retrieval layer. Do not re-test seen BMJ items and call it readiness; use fresh iatroX items for the credible signal. The two-Q-bank rule explains how to keep the measurement layer uncontaminated, and the pillar on auditing an AI medical exam tutor applies the same scepticism to any automated feedback.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 20 July 2026; BMJ OnExamination's question count, prices and feature descriptions are vendor-reported and were correct at the time of checking, so verify current details on onexamination.com before purchase. No adaptive-difficulty algorithm is claimed on the vendor's or iatroX's behalf; the feed is described as weakness-weighted personalisation. Disclosure: iatroX operates a UK question bank and competes with the products discussed; iatroX is not a geriatric-medicine-specific SCE bank, and its role here is confined to cross-specialty knowledge, unseen measurement and spaced retrieval. Corrections are welcome via the feedback route on iatrox.com. References: the Federation SCE Geriatric Medicine page and blueprint (thefederation.uk); BMJ OnExamination SCE Geriatric Medicine product page (onexamination.com); the iatroX comparison hub; the pillar on building a blueprint-coverage matrix; and the note on why your Q-bank percentage is not your exam score.