This is for UK final-year students and foundation doctors deciding whether BMJ OnExamination is enough for the Prescribing Safety Assessment (PSA). The honest headline first: despite how these products are often described, BMJ OnExamination's PSA offering does not, on its own product page as checked on 19 July 2026, document an adaptive difficulty engine. It is a compact, well-explained question bank with mock tests, leaderboards and peer comparison. Its analytics optimise your ranking against other users and your familiarity with recurring themes — not, by themselves, your coverage of the PSA blueprint or your transfer to unseen prescribing.
Current-state box (checked 19 July 2026)
The figures below are vendor-reported from the BMJ OnExamination PSA product page and should be re-verified before purchase.
| Item | What the product page states (19 July 2026) |
|---|---|
| Question count | Around 110 PSA questions (vendor-reported) |
| Access periods and price | 3 months £6.99; 6 months £9.99; 9 months £12.99; 12 months £14.99 (vendor-reported) |
| Free trial | Ten free questions per day before subscribing |
| Adaptive engine | Not documented on the PSA product page. Difficulty is chosen through a "Select Questions" function, i.e. user-selected, not an automatic adaptive algorithm |
| Mock tests | Curated by the BMJ editorial team, described as reflecting recent exam themes |
| Analytics | Peer comparison and progress tracking; ability to recap questions found hard |
| Extras | Offline app access and daily leaderboards ("Group Learning") |
The value proposition is clear and, at this price, reasonable: BMJ editorial quality, sensible explanations, and mock papers themed to what recent candidates report. The limitation is equally clear: a bank of roughly 110 items is small relative to the breadth the PSA samples, and "personalised revision" here appears to mean daily-refreshing free questions and manual difficulty selection rather than an algorithm that reshapes your feed. Treat the "adaptive" framing as a marketing label until the product page says otherwise.
Exam anchor: the PSA blueprint and official practice papers
The PSA is a two-hour, computer-based assessment of 60 items carrying 80 marks (the item types are not equally weighted). It samples eight question types — Prescribing, Prescription Review, Planning Management, Communicating Information, Calculation Skills, Adverse Drug Reactions, Drug Monitoring and Data Interpretation — across seven clinical domains: Medicine, Surgery, Elderly Care, Paediatrics, Psychiatry, Obstetrics and Gynaecology, and General Practice. Prescribing and Prescription Review carry the largest share of the marks, which is why a bank that under-samples them can leave you exposed even with a flattering overall percentage. The pass mark is set by modified Angoff standard-setting, and a pass is valid for two years. The exam is sat by final-year students and some foundation-year-one doctors.
Two things follow. First, the official practice papers published through the PSA programme are the gold standard for calibration, because they are written to the real blueprint and mark scheme — no third-party bank, BMJ OnExamination included, substitutes for working through them. Second, for UK medicines the reference standard is the Summary of Product Characteristics via the electronic medicines compendium (SmPC/eMC), alongside NICE and CKS; make sure any bank's explanations reconcile with those sources rather than with a house style.
Define every metric before you trust it
Analytics only help if you know what each number means and whether the product actually shows it.
- First-attempt accuracy — your score the first time you see an item, before any repeat. This is the only accuracy figure that reflects genuine recall; guard it jealously.
- Repeat accuracy — your score on items you have already seen. It inflates fast and mostly measures memory of the bank, not prescribing competence.
- Percentile / peer comparison — where you rank against other users. BMJ OnExamination surfaces this; it tells you about the population using the product, not about the pass standard, which is Angoff-set independently of the crowd.
- Predicted score — a modelled pass likelihood. Verify whether the PSA product offers one at all; if it does, treat it with heavy scepticism on a bank of this size.
- Coverage — the share of the blueprint your attempted items span, by question type and clinical domain. This is the number that actually governs readiness, and it is the one home screens rarely foreground.
- Difficulty — a per-item index of how hard others found a question. Useful for spotting where you are weak, but not a substitute for coverage.
- Time per item — your pace. The PSA is not a sprint, but calculation and data-interpretation items eat minutes, so pace by item type matters.
If a metric is not on the product page, do not assume it exists. On 19 July 2026 the PSA product clearly documents peer comparison and progress tracking; a formal difficulty index, per-item timing and a predicted score are not clearly documented — verify on the product page before you build a plan around them.
Selection bias: why the numbers can flatter you
Even without an adaptive engine, self-directed practice introduces selection bias. When you choose your own questions, you drift toward the types you find satisfying — often Prescribing and Adverse Drug Reactions — and away from Calculation Skills and Data Interpretation. Your overall percentage then rises because it is dominated by your comfortable types, while the mark-heavy or fiddly types stay thin. If a product did run an adaptive feed that over-samples your weak areas, the opposite distortion appears: your raw percentage looks worse than your true standing because you are being fed harder items. Either way, a headline percentage from a filtered or self-selected set is not comparable with your score on a mixed, unseen block — which is the only distribution that resembles the exam.
Blueprint audit: compare your distribution, not your average
Once a fortnight, ignore the home-screen average and audit your distribution. Export or tally how many items you have attempted in each of the eight question types and, as far as the product allows, each of the seven clinical domains. Put that beside the official weighting. You are looking for two failure modes: a question type you have barely touched (Calculation Skills and Drug Monitoring are the usual suspects), and a clinical domain you have avoided (Paediatrics and Obstetrics and Gynaecology are commonly under-practised). A bank of about 110 items cannot give deep coverage of all eight types across all seven domains, so expect to find gaps — and expect to fill them from official practice papers and a second, larger bank rather than by re-drilling the same 110.
Readiness test: five conditions for a credible signal
A score only predicts anything if it was earned under exam-like conditions. Before you believe you are ready, confirm all five:
- Unseen — items you have not attempted before, so you are testing recall, not memory of the bank.
- Timed — two hours for 60 items, with calculation and data items included so your pace is realistic.
- Mixed — all eight question types and a spread of the seven domains in one block, not a topic-filtered set.
- No assistance — no notes, no medicines reference open, nothing you will not have on the day.
- Adequate sample — enough items that the score is not noise; a single ten-item quiz tells you very little.
BMJ OnExamination's mocks can meet several of these, which is their strongest use. But a 110-item bank cannot supply many genuinely unseen full-length mixed blocks once you have worked through it — which is exactly where a second, larger bank of unseen items earns its place.
Algorithm override rules
Whether the "personalisation" here is a real feed or just your own selection, override it deliberately for the material that under-surfaces:
- Calculation Skills — force a fixed number every session; these are pure marks and pure practice.
- Data Interpretation and Drug Monitoring — schedule them explicitly; they are easy to avoid and heavily clinical.
- Prescription Review and Prescribing — the mark-heavy types; over-, not under-weight them.
- Adverse Drug Reactions and interactions — reconcile every explanation against the SmPC/eMC, not house notes.
- Under-sampled clinical domains — deliberately pull Paediatrics, Obstetrics and Gynaecology, and Elderly Care items even when the feed would not.
Worked dashboard example
Suppose after two weeks your BMJ OnExamination analytics show: first-attempt accuracy 74% overall; Calculation Skills attempted 6 items at 50%; Data Interpretation attempted 9 at 56%; Prescribing 40 items at 80%; peer percentile 62nd. Do not read "74% and 62nd percentile" as "on track." Read the distribution: two mark-relevant types are barely sampled and both are below 60%. Next week's quotas write themselves — 20 Calculation items, 20 Data Interpretation, and a single unseen mixed timed mock — while Prescribing, already strong and well-sampled, gets maintenance only. No pass prediction is made or needed; the quotas follow the gaps.
A seven-day plan: one job each
Use BMJ OnExamination for the job it does well and a second bank for unseen transfer. This pairing makes no claim to access any proprietary algorithm; it simply divides the labour.
- Day 1 — Sit an official PSA practice paper timed. This is your calibration anchor.
- Days 2–4 — BMJ OnExamination for one job only: work its themed mocks and its explanations to consolidate the recurring patterns, reviewing every miss against the SmPC/eMC.
- Days 5–6 — Switch banks for transfer: run fresh, unseen, mixed, timed blocks in iatroX's free PSA bank, which maps to the PSA blueprint and surfaces weak question types, so you are measuring recall on items you have never seen rather than re-scoring familiar ones. Log first-attempt accuracy by question type.
- Day 7 — Blueprint audit and a second official-style mock. Compare your two banks' distributions; fill whichever cells are still empty.
The principle is the two-Q-bank rule: one bank to learn a defined job, a second, larger unseen bank to measure transfer, with no duplicated items to corrupt your calibration.
Decision checklist: continue, supplement, switch or stop
- Continue BMJ OnExamination if you are still extracting new learning from its explanations and mocks and have not exhausted the bank.
- Supplement — the likeliest verdict — once your first-attempt accuracy is plateauing but your coverage audit shows thin question types or domains; add official papers and a larger unseen bank rather than re-drilling 110 items.
- Switch if you need volume and genuine adaptive weak-topic surfacing that this product does not document, and price is not the deciding factor.
- Stop adding new questions altogether when unseen mixed timed blocks are stable across all eight types and your errors are careless rather than knowledge-based — at that point, rest and light calculation practice beat more questions.
Base every branch on a measured gap, not on novelty or on the sunk cost of a subscription.
Frequently asked questions
Is BMJ OnExamination enough for Prescribing Safety Assessment on its own? For most candidates, no — not because it is poor, but because a bank of roughly 110 items (vendor-reported, 19 July 2026) cannot give deep, unseen coverage across eight question types and seven clinical domains, and the exam's mark weight sits in Prescribing and Prescription Review. It is a strong, inexpensive first structured pass with good explanations, but readiness needs the official practice papers plus a second, larger unseen bank for transfer. Treat it as one component of the stack, not the whole stack.
Which Prescribing Safety Assessment component does BMJ OnExamination not reproduce well? The components most vulnerable to under-sampling in any compact, self-selected bank are Calculation Skills, Data Interpretation and Drug Monitoring — the types candidates instinctively avoid and a 110-item bank cannot cover in depth. The product page does not document an adaptive engine that would force these to the surface (checked 19 July 2026), so you must schedule them deliberately. Reconcile every drug and monitoring answer against the SmPC/eMC, since house explanations are no substitute for the primary source.
How many BMJ OnExamination questions should I complete per day for Prescribing Safety Assessment? Volume matters less than coverage and conditions, but a workable pattern is 20 to 30 items on a normal study day, weighted toward your audited weak types rather than your comfortable ones, with the day's misses reviewed against the SmPC/eMC before you move on. Given the bank's size, you will exhaust genuinely new items within a couple of weeks of that pace — which is your cue to move to official papers and unseen mixed blocks rather than repeating the set for a rising, and increasingly meaningless, repeat-accuracy figure.
When should I stop using BMJ OnExamination and move to mixed mocks? Move to mixed mocks when you have worked through the bank's new items and its themed mocks and your first-attempt accuracy has plateaued — because beyond that point repeat accuracy climbs while real recall does not. The signal is not a calendar date; it is a stable first-attempt score combined with a coverage audit showing no untouched question types. At that stage, an unseen mixed timed block, ideally alternating an official practice paper with a fresh bank, tells you far more than another pass through familiar questions.
How should I combine BMJ OnExamination with iatroX without duplicating practice? Give each a distinct job. Use BMJ OnExamination to learn — its explanations and themed mocks — and use the free iatroX PSA bank to measure transfer on unseen, mixed, timed blocks mapped to the blueprint, logging first-attempt accuracy by question type. Never re-answer the same items across both, because duplicated questions inflate your scores and destroy the calibration that makes a second bank useful. If your two coverage audits show the same empty cell, fill it from official papers. This is the two-Q-bank rule applied to the PSA, and the comparison hub shows which tool owns which job.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026; question counts, prices and features are vendor-reported and change without notice, so verify the current figures on the BMJ OnExamination PSA product page before purchasing. Where this article notes an absence — such as no documented adaptive engine or predicted score — it reflects what the product page states on the checked date, not a claim about internal design. Disclosure: iatroX operates a competing, free UK PSA question bank; this article confines iatroX's role to the unseen-measurement and adaptive weak-topic jobs that the audited BMJ OnExamination PSA product does not claim, and directs calibration to the official PSA practice papers. Corrections are welcome via the feedback route on iatrox.com.
References: Prescribing Safety Assessment programme — blueprint and official practice papers (prescribingsafetyassessment.ac.uk and the BPSA learner portal); BMJ OnExamination PSA product page (onexamination.com/products/psa); electronic medicines compendium (SmPC/eMC); iatroX PSA landing and comparison hub (iatrox.com/psa, iatrox.com/compare); iatroX — "Your Q-Bank Percentage Is Not Your Exam Score" and the two-Q-bank rule pillar.