This is for candidates working through a question bank for MRCP(UK) Part 1 who want to know what their completion percentage is not telling them. Part 1 is entirely best-of-five multiple choice, so the gap is not a missing modality — it is what a headline percentage conceals inside the MCQ format: whether you can sustain accuracy across the whole blueprint, interpret raw data cold, read unfamiliar images, and hold your pace across two three-hour papers. Those are the skills to train and measure directly.
What a completion percentage shows for MRCP Part 1 — and what it hides
As of 19 July 2026, the established Part 1 banks (Pastest, PassMedicine and StudyPRN among them) each hold several thousand best-of-five items with explanations and analytics; treat any specific count or price as vendor-reported and confirm it on the product page. What every one of them reports back to you is a completion percentage and a running accuracy figure. Those two numbers answer one question honestly — how much of this bank you have seen and how often you were right on it — and they quietly imply a second answer they cannot actually give: whether you are ready.
A percentage cannot see whether your correct clicks are spread across the whole blueprint or bunched in the specialties you enjoy. It cannot see whether you reached the answer by interpreting the data in the stem or by recognising a question you have met before. It cannot see how your accuracy behaves in the last twenty minutes of a three-hour paper. Part 1 is a knowledge exam, but it is also a performance under constraints, and the constraints — breadth, unfamiliar data, unfamiliar images, and time — are exactly what the completion metric averages away. The rest of this article names those constraints and gives each one an observable behaviour, a practice task, a feedback source and an exit standard.
The official MRCP Part 1 format, mapped
Part 1 is two three-hour papers, each of 100 best-of-five questions, with no negative marking — so every item must be answered, and a blank scores the same as a wrong click. That gives a working pace of about 1.8 minutes per question, less once you allow for the data-heavy stems that carry laboratory panels, traces and images. Across the two papers (200 questions in total), the MRCP(UK) blueprint samples the whole of general medicine in roughly these proportions:
| Blueprint area | Approx. items (of 200) | Blueprint area | Approx. items (of 200) |
|---|---|---|---|
| Clinical sciences | 25 | Endocrinology | 14 |
| Clinical pharmacology & therapeutics | 15 | Haematology | 10 |
| Cardiology | 14 | Psychiatry | 9 |
| Infectious diseases | 14 | Dermatology | 8 |
| Neurology | 14 | Geriatric medicine | 8 |
| Nephrology | 14 | Oncology | 5 |
| Respiratory | 14 | Medical ophthalmology | 4 |
| Rheumatology | 14 | Palliative medicine | 4 |
| Gastroenterology | 14 |
The shape matters more than any single number. Clinical sciences is the single largest block and the one candidates most often under-rehearse; the long tail — oncology, medical ophthalmology, palliative medicine — is small per topic but collectively decisive at the margin. A bank percentage of 80% tells you nothing about whether those small, high-consequence cells are covered. Confirm the current blueprint and work the official sample questions on the MRCP(UK) site, because the sample items are the closest guide to house style and stem length.
Knowledge versus performance: what a correct answer proves
A correct answer in untimed, single-topic revision proves something real but narrow: that in that context, with that priming, you could distinguish the right option from four distractors. It does not prove that you can retrieve the same fact cold, thirty questions into a mixed paper, when the topic has switched three times since you last thought about it. It does not prove that you built the answer from the sodium, the film or the trace in front of you rather than recognising a stem you have already seen twice. And it does not prove that you can do any of this at 1.8 minutes an item without your accuracy sagging in the final block.
This is the separation the rest of the plan protects. Knowledge is necessary and a bank builds it well. Performance — retrieval under breadth, unfamiliarity and time — is a different capability, and the only way to measure it is on questions you have not met, sat under the real constraints. That is why the canonical caveat, your Q-bank percentage is not your exam score, is the first thing to internalise: the number you watch during revision is a learning metric, not a readiness metric.
The four things a percentage cannot see
The title names three; the honest list is four, because images are their own trap. For each, here is the behaviour to watch, the deliberate-practice task, where trustworthy feedback comes from, and the standard that tells you to move on.
| Skill | Observable behaviour | Deliberate-practice task | Feedback source | Exit standard |
|---|---|---|---|---|
| Broad blueprint coverage | Accuracy holds in low-volume specialties, not just favourites | Blueprint-proportioned blocks; a coverage matrix with a floor per cell | Per-specialty accuracy on unseen items | No cell below its floor for two weeks |
| Data interpretation | You reach the answer from the raw values, not the label | Stripped-stem drills (ABG, iron studies, LFTs, ECG, CSF) with the diagnosis hidden | Reference ranges + a senior sanity-check on reasoning | Correct interpretation cold, within pace |
| Image reading | You read a fresh image, not a remembered one | Rotating sets of unseen films, fundi, ECGs, slides | Reference atlas / clinician verification | Reliable fresh-image calls under time |
| Time pressure | Accuracy in the final block matches the first | Full-length timed unseen papers, both sittings in a day | First-pass accuracy by block; pace log | No decay across a simulated two-paper day |
Broad blueprint coverage. The commonest failure is not weakness but imbalance: a candidate at 82% overall who has done 90% of the cardiology and respiratory items and barely touched clinical sciences, medical ophthalmology or palliative care. The fix is structural — build a blueprint-coverage matrix, set a floor of attempted-and-reviewed items per cell in proportion to the table above, and drive the neglected cells up. Our guide to why completion is not coverage sets out how to build that matrix for any exam.
Data interpretation. The real paper rarely hands you a labelled syndrome; it hands you a set of numbers and asks what they mean. Practise with the label removed — an arterial blood gas with no diagnosis, iron studies with no anaemia named, a sodium and osmolality with no scenario — and force the interpretation before you read on. This is a skill a bank trains only incidentally, because its explanations give you the answer the moment you reveal them.
Image reading. A bank that reuses the same fundus photograph or blood film trains recognition of that image, not interpretation of a new one. Rotate your image sources so the pictures stay unfamiliar, and treat a correct call on a repeated image as no evidence at all.
Time pressure. With no negative marking, the discipline is to answer everything and flag rather than freeze. The only honest rehearsal is a full-length, timed, unseen paper — ideally both papers in a day at least once — with your accuracy logged by block so you can see whether the last twenty questions cost you marks a shorter set would have hidden.
A four-week modality ladder
Skills are built by climbing from isolated drill to unseen simulation, not by repeating whole mixed papers from the start. A four-week block might run:
| Week | Rung | What you do | What it proves |
|---|---|---|---|
| 1 | Isolated skill | Untimed then timed drills on one weakness (e.g. data interpretation) | The component works in isolation |
| 2 | Coached case | Mixed items with full reasoning written before reveal; misses classified | You can reason, not just recognise |
| 3 | Timed integrated | Full 100-item timed blocks under exam conditions | The skill survives breadth and pace |
| 4 | Unseen simulation | A fresh, blueprint-proportioned, timed paper you have never seen | Readiness on novel material |
The rungs matter in order. Jumping straight to timed mixed papers on a familiar bank measures your memory of that bank; climbing the ladder measures the capability underneath. Keep a reserve of unseen questions from a separate source for week four so the simulation is genuinely cold.
When AI feedback helps, when it misleads, and when you need a clinician
AI feedback is genuinely useful at Part 1 for three narrow jobs: explaining why a distractor is wrong when a bank's note is thin, generating variant stems on the same decision so you test transfer rather than recall, and — through a citation-first tool such as Ask iatroX — checking that a management point still matches current UK guidance rather than an older habit. It is unreliable for anything requiring calibrated judgement: it will confidently narrate an ECG or a film it cannot actually see well, it cannot tell you your pacing is safe, and it can produce fluent, wrong explanations that read as authoritative. Before you trust any automated explanation, read how to calibrate AI feedback. Where a human is still required is verification: a senior who can confirm your data interpretation and image reading are examiner-standard, because a wrong rule learned confidently from an AI explanation will fail across a whole class of questions.
A balanced task matrix
To stop yourself practising only familiar scenarios, plan blocks against a simple matrix rather than a topic list. Cross the blueprint areas with the task the item demands — recall, data interpretation, image reading, next-step management — and make sure the demanding cells are populated.
| Blueprint area | Recall item | Data interpretation | Image reading | Next-step management |
|---|---|---|---|---|
| Cardiology | ✓ | ECG / echo values | ECG, angiogram | ✓ |
| Nephrology | ✓ | U&E, ABG, urine | Biopsy image | ✓ |
| Haematology | ✓ | FBC, film indices | Blood film | ✓ |
| Endocrinology | ✓ | Dynamic tests | — | ✓ |
| Clinical sciences | ✓ | Statistics, genetics | — | — |
The point is to force the right-hand columns. Most banks over-supply the recall column and under-supply data and image work, which is precisely the imbalance the real paper exploits.
A worked example: turning a dashboard into quotas
Take a candidate six weeks out at 79% overall completion and 74% accuracy — comfortable-looking numbers. Broken down, the dashboard shows 91% completion in cardiology and respiratory, 88% in gastroenterology, but 34% in clinical sciences, 20% in medical ophthalmology and 15% in palliative medicine, and it shows a first-pass accuracy of 76% untimed against 58% on the one timed mixed block they attempted. The blended figure hid three separate problems: the largest blueprint block is barely touched, the long tail is empty, and accuracy falls almost twenty points under time.
The framework converts that into quotas rather than reassurance. Clinical sciences gets a floor of 25 reviewed items and a daily allocation until it clears; medical ophthalmology and palliative medicine each get a small fixed quota so no cell is blank; a stripped-stem data-interpretation drill runs three times a week; and every second block is timed and logged by block so the pace problem becomes visible and trainable. Six weeks later the honest readiness signal is not the 79% — it is the first-pass accuracy on a fresh, timed, blueprint-proportioned paper, and whether it now holds from the first block to the last. No number here predicts a pass; the exercise is to remove the three specific weaknesses the average concealed.
Three mistakes this framework is designed to stop
The first is treating a rising percentage as rising readiness. On a familiar bank, the last ten points are usually recognition of items you have already met, not new capability — which is why a preserved unseen reserve and the two-Q-bank rule exist: keep one bank for learning and a second, unseen source purely for measurement. The second is single-topic comfort: doing forty cardiology questions in a row builds priming the real mixed paper strips away, so your session accuracy overstates your exam accuracy. The third is skipping the data and image work because the bank's explanations make it feel done — the explanation hands you the interpretation you were supposed to perform, so you never actually practise performing it. Each mistake produces a flattering number and a real gap, and each is fixed by measuring on unseen, mixed, timed material.
Continue, supplement, switch or stop
Continue with your main bank while your unseen mixed accuracy is still climbing and your coverage matrix is still filling — it is doing its job. Supplement the moment your completion percentage rises faster than your performance on novel questions, because that divergence is recognition, not readiness; add unseen timed blocks and targeted data-interpretation drills. Switch, or add a second bank, only for a measurable gap — thin clinical-sciences coverage, no image work, no timed mode — not for novelty or because a competitor looks busier. Stop grinding the bank in the final fortnight once coverage holds and pace is on target; the last weeks belong to unseen simulation and error review, not a third pass through familiar items.
The bottom line
MRCP Part 1 is an MCQ exam, so no simulator is required — but that is exactly why the completion percentage is so misleading. It measures the one thing that is easy to measure, how much of a familiar bank you have finished, and stays silent on the four things that decide the paper: breadth across the whole blueprint, interpretation of raw data, reading of unfamiliar images, and accuracy sustained under time. Build the coverage matrix, drill the data and images with the answer hidden, and measure yourself on unseen, timed, blueprint-proportioned material. The percentage is a learning metric; readiness is what survives when the questions are new.
Frequently asked questions
How do I know whether I have covered the full MRCP Part 1 blueprint? Map your attempted-and-reviewed items to the MRCP(UK) blueprint categories and their proportions, then set a floor per cell — clinical sciences at roughly 25 of 200, the specialty blocks at about 14 each, the long tail (oncology, medical ophthalmology, palliative medicine) small but never zero. Coverage is a matrix with no empty high-consequence cells, not a single completion figure; the completion-is-not-coverage guide shows how to build it.
Can one question bank be enough for MRCP Part 1? A single strong bank can be the backbone of your learning, but one bank cannot honestly measure readiness, because once you have seen its items your accuracy on them is recognition. Keep a second, unseen source purely for measurement under the two-Q-bank rule, and let your readiness signal come only from questions the learning bank never showed you.
What should I measure instead of my overall Q-bank percentage for MRCP Part 1? Measure first-pass accuracy on unseen, timed, blueprint-proportioned blocks; per-specialty accuracy so imbalance is visible; data-interpretation accuracy on stripped stems; and pace, logged by block so end-of-paper decay shows up. These track performance under the exam's real constraints, which the blended percentage averages away.
When should I stop doing new MRCP Part 1 questions? Stop when unseen timed blocks are stable across every blueprint area and new questions have stopped revealing gaps — not at a fixed number and not when the bank reads "complete". Finishing a familiar bank is not readiness; stable transfer to cold, timed material is, and once you have it, the marginal question adds little.
Which MRCP Part 1 resource should I use for my weakest component? Match the resource to the miss. For breadth, a broad bank governed by a coverage matrix; for data interpretation, stripped-stem drills and dynamic-test practice; for images, rotating unseen film and ECG sets; for pace, full-length timed unseen papers. There is no single resource that fixes all four, and choosing by component rather than brand is the point of the framework.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026; any bank question counts or prices mentioned are vendor-reported — verify them on the relevant product page. MRCP Part 1 format and blueprint are per MRCP(UK). Disclosure: iatroX operates a competing MRCP Part 1 bank, and its role here is confined to jobs a single learning bank does not claim — unseen, timed measurement and citation-first guideline checks. Corrections via the feedback route on iatrox.com. References: MRCP(UK) Part 1 format, blueprint and sample questions (mrcpuk.org); the iatroX comparison hub; related reading: why your Q-bank percentage is not your exam score and completion is not coverage.