This audit is for GPs and doctors in women's health revising the DRCOG with BMJ OnExamination. It addresses the single written single-best-answer paper, the only component the diploma has. The principal limitation to state at the outset is that BMJ OnExamination markets "personalised revision" but does not publish a proprietary adaptive-difficulty engine, so the honest question is not "what is the algorithm doing" but "what does the visible personalisation optimise, and how do you read the numbers it produces."
What BMJ OnExamination offers for DRCOG right now
The table is vendor-reported and last checked 20 July 2026. Confirm everything on the product page before subscribing.
| Attribute | What BMJ OnExamination reports (vendor-reported, 20 July 2026) |
|---|---|
| Bank size | 350+ DRCOG questions written by O&G clinicians |
| Item types advertised | EMQs, best-of-five and MCQs, with content mapped to green-top guidelines |
| Personalisation | "Personalised revision," selectable difficulty and a recap of questions you found hard |
| Assessment | Editorially curated mock tests |
| Feedback | Performance feedback and daily leaderboards through Group Learning |
| Access and price | £29.99 for one month, £39.99 for two, £49.99 for three, £79.99 for six, £139.99 for twelve |
| Free trial | Ten free questions each day |
| Not documented | No published predicted-score model, percentile engine, per-item difficulty index or adaptive-difficulty algorithm on the product page |
Two findings to lead with. First, the item mix still advertises EMQs alongside best-of-five and MCQ, but the live DRCOG is now single-best-answer only. The best-of-five items are on-format, since best-of-five is a single-best-answer style; the EMQs are useful knowledge practice but no longer match the paper you will sit, so do not treat them as format rehearsal. Second, there is no evidence of a documented adaptive engine. The visible personalisation is a difficulty selector plus a recap of hard items, which is worth using but is not a black box optimising a hidden objective.
The exam you are actually sitting: the DRCOG
DRCOG is one computer-based paper: 120 single best answer questions in three hours, each SBA worth two marks for a total of 240, with the pass mark set by modified Angoff. The format is SBA-only; there is no EMQ section and no separate oral or practical component to prepare. The content spans obstetrics across the antenatal, intrapartum, postnatal and neonatal period and early pregnancy, gynaecology including menstrual disorders, gynae-oncology, urogynaecology and the menopause, sexual and reproductive health including contraception and abortion care, and the surrounding areas of women's health in primary care, safeguarding, consent and basic epidemiology. The RCOG sets the syllabus and does not publish a fixed percentage weighting per domain in the way some royal-college exams do, so audit your coverage against the breadth of the current syllabus on rcog.org.uk rather than an invented weighting table.
Define every metric before you trust it
A dashboard is only useful if you know exactly what each number means. First-attempt accuracy is your percentage correct the first time you ever see an item, and it is the only figure that reflects new learning. Repeat accuracy is your percentage on items you have seen before, and it rises with familiarity rather than knowledge, so it flatters you. A percentile, if shown, compares you with other users, whose sample and timing you do not control. A predicted score, if shown, is a model output, not a promise, and its assumptions are rarely published. Coverage is the proportion of the syllabus you have actually attempted, not the proportion of questions answered correctly. Difficulty is either the level you selected or a crowd-derived index, and the two are not interchangeable. Time per item tells you about pacing, which for DRCOG means roughly 90 seconds per question. Where BMJ OnExamination does not clearly document one of these, such as a predicted score or a per-item difficulty index, treat its absence as a reason to lean on your own first-attempt data rather than a headline average.
Selection bias: why the home-screen average misleads
Any feed that recaps your hard questions, or that you steer toward weak topics, will over-sample your weaknesses. That is pedagogically sensible, but it has a statistical consequence: your running percentage is now computed on a harder-than-average, non-representative set, so it is not comparable with a mixed unseen block and it is not comparable with the exam. The average on the home screen answers "how am I doing on the questions the feed chose to show me," which is a different question from "how would I do on a fair, blueprint-balanced paper." Never read the headline average as an exam prediction.
Blueprint audit: compare your distribution, not your average
Once a fortnight, ignore the average and look instead at the distribution of what you have attempted. List the syllabus areas down one column and your attempted-question count and first-attempt accuracy beside each. The pattern you are hunting for is a domain with a low attempted count and a low accuracy hiding behind a comfortable overall figure, typically the low-volume areas: contraindications in contraception, the specifics of antenatal screening pathways, gynae-oncology red flags, ethics and consent, and the numerate or data-interpretation items. The home-screen average cannot show you this; only the distribution can.
The readiness test: what a credible signal requires
A number only means "ready" if it was produced under exam-like conditions. That requires five things together: the questions were unseen, so no repeat-familiarity inflation; the block was timed at DRCOG pace; the items were mixed rather than topic-filtered, so you had to identify the problem type yourself; you used no assistance, notes or tutor mode; and the sample was large enough to be stable, which for a 120-question paper means blocks well into the dozens rather than a handful. Fail any one of these and the signal is soft, however good the percentage looks.
Algorithm override rules
Because the personalisation will not reliably surface everything, override it deliberately. Force low-volume domains onto the screen even when the recap feed would not, particularly contraception and sexual health detail, antenatal and intrapartum management, gynae-oncology and urogynaecology. Force any image or data-interpretation items the exam can include. Force ethics, consent and safeguarding scenarios, which are easy to under-practise. And force the numerate items so a calculation under time does not surprise you. The rule of thumb: if a topic is high-stakes but low-frequency, you must summon it manually, because neither a recap steer nor your own comfort will bring it to you often enough.
A worked dashboard example
Suppose after two weeks your dashboard reads: overall 71 per cent, first-attempt 63 per cent, repeat 88 per cent, and attempted counts that are heavy in general obstetrics and gynaecology but thin in sexual and reproductive health and in ethics. Do not read 71 per cent as a pass prediction; it is inflated by that 88 per cent repeat figure. The 63 per cent first-attempt on a weakness-weighted feed is your more honest floor. The action for next week is a set of quotas, not a target score: attempt a fixed block of unseen sexual-and-reproductive-health items, a fixed block of ethics and consent, and one timed mixed block of at least 40 unseen questions to get a clean first-attempt reading. No pass probability is produced, because none would be credible from this data.
A seven-day plan for a doctor in women's health
On day one, run a topic-filtered BMJ OnExamination block in your thinnest domain, fully reviewed. On day two, the same in a second thin domain. On day three, convert the worst misses into transfer practice on unseen items in iatroX, so you are testing the principle rather than re-recognising the stem. On day four, a light spaced review of earlier misses. On day five, a timed mixed block of unseen questions in iatroX for a clean first-attempt reading against your baseline. On day six, override the feed to force a low-frequency, high-stakes topic. On day seven, rest or brief consolidation. BMJ OnExamination is the learning bank; iatroX supplies the unseen measurement. Neither party's selection logic is claimed to be a proprietary algorithm here.
Decision checklist: continue, supplement, switch or stop
Continue if your unseen first-attempt accuracy is rising and your attempted-question distribution is filling the thin domains. Supplement with unseen timed blocks and the RCOG's own material if your headline average looks healthy but your distribution shows gaps, or if repeat accuracy is doing most of the work. Switch, or reweight your practice, given that some items are EMQ and off-format for the current SBA-only paper. Stop topic-filtered practice and move to full timed mixed mocks once the readiness conditions are met. Every branch is a measurable gap, not novelty or sunk cost.
The bottom line
BMJ OnExamination is a reasonable DRCOG learning bank, but two honest caveats shape how you use it: its personalisation is a recap-and-difficulty steer rather than a documented adaptive engine, and its advertised item mix still includes EMQs that no longer match the single-best-answer paper. Define every metric before you trust it, audit your attempted distribution rather than your average, and judge readiness only on unseen, timed, mixed blocks. Optimise for coverage and honest first-attempt performance, not for a comforting home-screen number.
Frequently asked questions
Is BMJ OnExamination enough for DRCOG on its own? It is a reasonable single bank, with more than 350 clinician-written questions on the vendor's figures, but note that some items are EMQ or older-format while the live DRCOG is single-best-answer only, so part of the bank does not rehearse the current paper. The calibrated approach is to pair it with the RCOG syllabus and material and a source of unseen questions for measurement.
Which DRCOG component does BMJ OnExamination not reproduce well? DRCOG is a single written SBA paper, so there is no OSCE or oral to miss. The real gap is format currency and measurement: the advertised EMQ items are off-format for the current SBA-only exam, and the platform does not clearly reproduce a full-length, 120-question timed mock unless you build one, so it under-serves stamina and pacing rehearsal.
How many BMJ OnExamination questions should I complete per day for DRCOG? There is no official number, and this is guidance rather than a vendor claim. Around 20 to 40 questions a day, fully reviewed and error-coded around clinical work, is a sustainable pattern. Because the paper runs at about 90 seconds per item, prioritise reviewing why you missed something over piling up an impressive daily count.
When should I stop using BMJ OnExamination and move to mixed mocks? Move to timed mixed mocks when your attempted-question distribution covers the syllabus, your first-attempt accuracy on unseen mixed blocks is stable, and you can hold roughly 90 seconds per item comfortably. That is a coverage-and-performance signal, not a completion signal, so do not wait for the bank to read one hundred per cent.
How should I combine BMJ OnExamination with iatroX without duplicating practice? Apply the two-Q-bank rule: use BMJ OnExamination as your review-heavy learning bank and iatroX as your measurement bank for unseen, timed, mixed single-best-answer blocks. The discipline is not to re-answer the same items in both tools, so iatroX stays genuinely unseen and gives you a clean first-attempt reading rather than a familiarity-inflated one.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 20 July 2026. All BMJ OnExamination figures, features and prices are vendor-reported and may change; confirm them on the product page, and verify the DRCOG format and syllabus on rcog.org.uk. Disclosure: iatroX operates a competing UK question bank, so its role here is confined to a job BMJ OnExamination does not claim to do well, namely providing unseen, timed, single-best-answer blocks for a clean readiness measurement. Corrections are welcome through the feedback route on iatrox.com.
References: RCOG DRCOG syllabus, format and sample questions (rcog.org.uk); BMJ OnExamination DRCOG product page (onexamination.com); iatroX, "Your Q-Bank Percentage Is Not Your Exam Score" (iatrox.com/blog/qbank-percentage-not-your-exam-score); the two-Q-bank rule and blueprint-coverage-matrix pillars on iatrox.com; and the iatroX comparison hub (iatrox.com/compare).