The DRCOG is a written, single-best-answer paper with no clinical or oral component, so it is fair to ask why a question bank is not the whole answer. It is because a correct option selection is a narrow proof. It shows you could pick the right answer from five, on that stem, on that day. It does not show you can apply current NICE and RCOG guidance to an ambiguous real case, interpret an image or data trend without the prompt of a tidy stem, or reason through a primary-care women's health consultation. This is the exam-level hub for that gap and how to close it.
The direct answer: three things ordinary MCQ practice cannot fully assess
Ordinary MCQ practice cannot fully assess three DRCOG-relevant skills. First, live guidance application — using the current NICE/CKS, RCOG and SIGN position when the case does not map neatly onto a remembered fact. Second, image and data interpretation from scratch — reading an ultrasound description, a CTG trace, a growth or laboratory trend, or a risk figure, when nothing flags that interpretation is required. Third, primary-care women's health reasoning — the GP's actual task of holding uncertainty, safety-netting and deciding what to do with an undifferentiated presentation. A bank trains recognition of the packaged version of each; the exam, and the job it certifies, ask for the unpackaged version.
Official format map
The DRCOG is one computer-based paper: 120 single best answer questions in three hours, each worth two marks (240 total), pass mark set by modified Angoff to the standard of a competent GP trainee, currently SBA-only. There is no EMQ paper and no separate clinical or oral station — a point worth stressing, because unlike the DCH or MRCPCH there is no OSCE to catch what the written paper misses. Everything the diploma certifies has to be signalled by the written paper alone, which is precisely why the transfer skills below matter: the SBA is the only gate, so its blind spots become yours unless you train them deliberately. The seven RCOG modules span fertility and sexual health, subfertility, early pregnancy, the pregnant woman, the peripartum period and neonate, and urgent and non-urgent gynaecology. The RCOG syllabus and sample questions are the format's primary reference.
Knowledge versus performance: what a correct answer proves
Separate two things that a rising bank percentage blurs. A correct answer proves knowledge: on this item, you recognised the right option. It does not prove performance: that you would generate the same decision without the five options in front of you, under time pressure, when the presentation is noisy. SBAs are recognition tasks with the answer present; real practice and the harder exam items are generation tasks with the answer absent. The gap between recognising and generating is where competent-looking candidates come unstuck, and no amount of additional recognition practice closes it. You close it by practising generation — producing the decision before you see the options, and defending it. A useful discipline is to read only the stem, cover the five options, commit to a plan in a sentence, and only then reveal the choices to check yourself; the items where your generated plan and the keyed answer diverge are the ones worth the most of your remaining time.
The three under-tested skills, trained
For each skill, fix an observable behaviour, a deliberate-practice task, a feedback source and an exit standard. Vague intentions do not train skills; specified behaviours do.
| Skill | Observable behaviour | Deliberate-practice task | Feedback source | Exit standard |
|---|---|---|---|---|
| Live guidance application | You state the current recommendation and why it fits this case | Take an ambiguous vignette, decide before reading options, cite the NICE/CKS or RCOG basis | The primary guidance itself; a clinician for edge cases | You reach the guideline-concordant plan on unfamiliar cases, not just familiar ones |
| Image and data interpretation | You describe the finding and its implication unprompted | Cover the lead-in; interpret the image/trace/trend first, then answer | Worked answer; a clinician for genuine images | You interpret correctly without the stem signposting it |
| Primary-care women's health reasoning | You state a plan with safety-net and follow-up | Rehearse undifferentiated presentations aloud as if in clinic; write the plan | A supervising GP; structured feedback | Your plan is safe and complete under time, on cases you have not seen |
Take each skill in turn with a concrete DRCOG example, because the difference between recognising and generating is easiest to see in a real case.
Live guidance application. A recognition item asks which contraceptive is most appropriate for a 39-year-old smoker with migraine with aura, and you pick the progestogen-only option because you recognise that the combined pill is contraindicated. The generation version is the woman in front of you who also has a raised BMI, is breastfeeding and wants something she can stop easily — now you must hold several current UKMEC and NICE/CKS positions at once and defend a plan, not eliminate one wrong option. Train it by deciding before you read the options and stating the basis out loud.
Image and data interpretation. A recognition item captions the ultrasound finding for you and asks the next step. The generation version hands you the raw measurements and expects you to say what they mean: an empty gestational sac above the discriminatory size, a beta-hCG that has not risen appropriately over 48 hours, a CTG description with reduced variability. Cover the lead-in, interpret first, then choose — the skill is naming the finding, not recognising a label someone else has written.
Primary-care women's health reasoning. A recognition item asks for the single next investigation in postmenopausal bleeding. The generation version is the undifferentiated woman in a ten-minute appointment: what you ask, what you examine, how you safety-net and when you review. That reasoning is observed and corrected by a supervising GP, not scored by a bank, which is exactly why it is the skill most often left untrained.
A four-week modality ladder
Build the skills in rungs, from isolated to integrated, so that each week's practice is closer to the real task than the last.
| Week | Rung | What you do |
|---|---|---|
| 1 | Isolated skill | Drill one sub-skill at a time — a set of guideline-application items; a set of pure image/data reads; a set of "state the plan" prompts |
| 2 | Coached case | Work integrated cases with a supervising clinician correcting your reasoning aloud, not just your answer |
| 3 | Timed integrated case | Timed mixed items that embed guidance and data together, at exam pace, with generation before options |
| 4 | Unseen simulation | A full unseen timed block, calibrated against the RCOG official sample, reviewed by coded error |
The ladder matters because jumping straight to unseen simulation without the isolated and coached rungs just re-measures the gap; climbing it closes the gap first.
When AI feedback helps, when it misleads, and when a clinician is required
Automated feedback earns its place on the knowledge layer: it can mark a selected answer, surface an explanation instantly, and drive spaced retrieval across large volumes — useful, scalable and immediate. It becomes unreliable at the edges that matter most here: nuanced guidance where recommendations conflict or have recently changed, genuine image interpretation, and any judgement about safety-netting, safeguarding or communication, where a plausible-sounding explanation can be confidently wrong. A clinician or examiner is required precisely where performance rather than knowledge is being judged — the consultation reasoning and real image interpretation that the ladder's coached rung depends on. Before you trust any automated score, calibrate it against a worked answer or a clinician, and audit the tool for grounding and answer leakage rather than assuming it is right. iatroX sits deliberately on the knowledge and unseen-MCQ layer; it does not replace a supervising GP for the consultation-reasoning work, and this article does not suggest it can.
A balanced case matrix
Candidates practise what they enjoy and avoid what they fear, so their scenario set skews. Force balance with a matrix: for each module, ensure you have practised the knowledge item, the guidance-application task, the data-interpretation task and the consultation-reasoning task at least once.
| RCOG module | Knowledge item | Guidance task | Data/image task | Consultation task |
|---|---|---|---|---|
| Fertility & sexual health | ☐ | ☐ | ☐ | ☐ |
| Subfertility | ☐ | ☐ | ☐ | ☐ |
| Early pregnancy | ☐ | ☐ | ☐ | ☐ |
| The pregnant woman | ☐ | ☐ | ☐ | ☐ |
| Peripartum & neonate | ☐ | ☐ | ☐ | ☐ |
| Urgent gynaecology | ☐ | ☐ | ☐ | ☐ |
| Non-urgent gynaecology | ☐ | ☐ | ☐ | ☐ |
An empty column is a modality you are avoiding across the whole blueprint — usually data interpretation or consultation reasoning, rarely knowledge.
Red flags that you are training recognition, not competence
- Memorised scripts — you can recite the answer to a familiar case but stall when the presentation shifts.
- Repeated cases — your practice set is small enough that you recognise items rather than solve them.
- Generic feedback — the explanations you rely on are vague or unsourced, and you cannot trace them to primary guidance.
- Uncalibrated scoring — you trust a percentage that has never been checked against official material or a clinician.
- No official-rubric check — you have never compared your reasoning to the RCOG sample answers or a supervisor's standard.
Any two of these together mean your rising scores are measuring familiarity, and the transfer skills above are the corrective.
A worked example: the 82% candidate who is not ready
Consider a candidate three weeks out who sits at 82% on her bank and feels ready. Her case matrix tells a different story. The knowledge column is full across all seven modules; the data-interpretation column is empty for early pregnancy, the pregnant woman and the peripartum period; and the consultation-reasoning column is empty almost everywhere, because she has practised only by selecting options, never by generating a plan.
Put her in front of three generation tasks. First, a woman with a borderline beta-hCG and an ultrasound that does not show an intrauterine pregnancy: asked what this means and what she would do, she hesitates, because she has only ever seen the interpretation handed to her. Second, a 44-year-old with heavy menstrual bleeding and a raised BMI who wants to avoid surgery: asked for a plan with a safety-net and review, she produces a list of facts rather than a decision. Third, a guidance-application item on hypertension in pregnancy where the threshold has recently moved: she recalls the older figure.
None of this shows in her 82%, because the percentage is a recognition average and recognition is the one thing she has trained. This is the same warning as the canonical caveat that your Q-bank percentage is not your exam score, applied to the modality gap rather than the coverage gap. The corrective is not more questions; it is two weeks of generation practice — decide before the options, interpret data with the lead-in covered — and two or three supervised consultations on the presentations she avoids. Do that, and the same 82% will finally mean what she assumed it meant all along.
FAQ
How do I know whether I have covered the full DRCOG blueprint? Coverage in the sense this article cares about is not "attempted a question in every module" but "practised each module across all four modalities — knowledge, guidance application, data interpretation and consultation reasoning" — so use the case matrix, and treat an empty column as an uncovered modality even if your knowledge score in that module is high. A full knowledge column with empty performance columns is partial coverage wearing a complete disguise.
Can one question bank be enough for DRCOG? For the knowledge layer a single strong bank can do most of the work, but no bank trains the generation and consultation-reasoning skills the written paper only indirectly samples, so "enough" depends on whether you supplement recognition practice with generation practice and clinician-observed reasoning; as a knowledge engine one bank can suffice, as your entire preparation it cannot.
What should I measure instead of my overall Q-bank percentage for DRCOG? Measure your performance on generation tasks — can you state the guideline-concordant plan before seeing the options — your unprompted image and data interpretation, and your reasoning as judged against official sample answers or a supervisor, alongside unseen timed accuracy and pace; the overall percentage is a recognition average and, as the canonical caveat notes, your Q-bank percentage is not your exam score.
When should I stop doing new DRCOG questions? Stop adding new recognition questions once every module clears your knowledge floor and your case matrix is full across all four modalities; past that point the return on more MCQs is small, and the higher-value activities are timed unseen simulation, generation practice, and clinician-observed reasoning on the presentations you find hardest.
Which DRCOG resource should I use for my weakest component? Match the resource to the modality of the weakness: for a knowledge gap, targeted unseen questions such as a fresh iatroX block plus the primary NICE/CKS or RCOG guidance; for a data-interpretation gap, image and data items practised with the lead-in covered; and for a consultation-reasoning gap, supervised case discussion with a GP, which no bank and no automated tool can replace.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 20 July 2026. Exam-format facts are drawn from the RCOG DRCOG format and syllabus pages; any product figures are vendor-reported as of that date and should be re-verified before you rely on them. Disclosure: iatroX operates a competing UK question bank; it is positioned here only as the knowledge and unseen-MCQ layer, and this article states plainly that it does not replace clinician-observed consultation practice or real image interpretation. Corrections are welcome via the feedback route on iatrox.com.
References: RCOG, DRCOG format, pass mark and syllabus (rcog.org.uk); RCOG DRCOG sample questions (rcog.org.uk); NICE and CKS (nice.org.uk); iatroX, "Your Q-Bank Percentage Is Not Your Exam Score" (https://www.iatrox.com/blog/qbank-percentage-not-your-exam-score); iatroX, "Calibrating AI-graded feedback" (https://www.iatrox.com/blog/ai-graded-saqs-and-osces-how-to-calibrate-automated-feedback-before-you-trust-the-score); iatroX, "How to audit an AI medical exam tutor" (https://www.iatrox.com/blog/how-to-audit-an-ai-medical-exam-tutor-grounding-answer-leakage-hallucinations-and-retention); iatroX DRCOG content-gap checklist (https://www.iatrox.com/blog/the-drcog-q-bank-content-gap-checklist-what-to-verify-before-you-stop-doing-new-questions).