If you are preparing for the Prescribing Safety Assessment (PSA), ordinary multiple-choice practice will carry you a long way but stops short in four places: navigating the online formulary under time pressure, constructing a complete and legally safe prescription from a blank field, getting dose, route and frequency exactly right rather than approximately right, and holding all of that together at roughly two minutes per item. Recognising the correct drug in a five-option list is a different act from writing it correctly when nothing is prompting you. This article maps those gaps to the current PSA blueprint and gives you a trainable, observable standard for each.
This is the exam-level hub for that single question. Individual platform workflows link up here; the aim is to give you the framework once, in full, so you can decide what to practise and how to know when you have practised enough.
The skills a correct selected answer does not prove
A multiple-choice item measures recognition and discrimination: given a stem and options, can you identify the safest answer. That is genuinely necessary, and several PSA question types — Prescription Review, Planning Management, Adverse Drug Reactions, Drug Monitoring and Data Interpretation — are delivered in recognition or short-selection formats that a good bank reproduces well. But the PSA is not a pure recognition test, and its highest-weighted section is deliberately built to defeat recognition.
Four things a correct selected answer does not prove:
- That you can find the fact under time pressure. In the exam you have access to an online formulary hosted on the NICE website and to MedicinesComplete; the real skill is locating a dose, an interaction or a monitoring parameter in seconds, not knowing it cold.
- That you can produce a prescription rather than pick one. The Prescribing question type gives you a blank prescription and expects a complete, unambiguous, safe entry — drug, dose, route, frequency, formulation, and any necessary instruction.
- That your dose, route and frequency are exact. A bank marks a whole item right or wrong. The PSA can penalise an otherwise reasonable choice for the wrong strength, an inappropriate route, or a frequency that is unsafe in this patient.
- That you can sustain accuracy at pace. Sixty items in two hours, with the Prescribing section demanding the most work per mark, is an endurance and sequencing problem that untimed practice hides.
Separating knowledge from performance is the whole point. Your bank percentage measures how often you recognise the right answer with unlimited time and no construction demand. The exam measures whether you can perform the prescribing act, correctly and quickly, when nothing cues you. Those are correlated but not identical, which is why we treat the percentage as a coverage signal, not a readiness signal — see Your Q-Bank Percentage Is Not Your Exam Score.
The official format map
Work from the current blueprint, not from memory of an older paper. The PSA is a two-hour online assessment of 60 items totalling 200 marks, standard-set by modified Angoff, delivered by the British Pharmacological Society and MSC Assessment. Final-year medical students sit it (summatively or formatively depending on the school) and Foundation Year 1 doctors who have not previously passed must do so before the end of F1. A pass is valid for two years. The eight question types and their published mark weightings (verified on the PSA blueprint, 20 July 2026) are:
| Question type | Items | Marks each | Section marks | Share |
|---|---|---|---|---|
| Prescribing | 8 | 10 | 80 | 40% |
| Prescription Review | 8 | 4 | 32 | 16% |
| Planning Management | 8 | 2 | 16 | 8% |
| Providing Information | 6 | 2 | 12 | 6% |
| Calculation Skills | 8 | 2 | 16 | 8% |
| Adverse Drug Reactions | 8 | 2 | 16 | 8% |
| Drug Monitoring | 8 | 2 | 16 | 8% |
| Data Interpretation | 6 | 2 | 12 | 6% |
| Total | 60 | — | 200 | 100% |
Two structural facts drive the whole revision plan. First, Prescribing is worth 40% of the marks in eight items — roughly ten times the mark density of a Planning Management item — so a marginal candidate who cannot construct prescriptions cannot pass by being strong everywhere else. Second, items are set across seven clinical domains (Medicine, Surgery, Elderly Care, Paediatrics, Psychiatry, Obstetrics & Gynaecology, and General Practice), so a domain you avoid is a domain that can appear at the highest mark weight. The official practice papers on the PSA website are the calibration gold standard for both facts; a third-party bank supplies the volume the finite official set cannot. For the coverage logic behind this, see the blueprint-coverage matrix method.
The four under-trained skills, broken down
Formulary navigation
The PSA is open-book by design, which is the opposite of a memory test. Candidates who fail here are usually not ignorant of the drug; they are slow to find the right entry, or they read the wrong sub-section. The observable behaviour is search-to-fact latency: how many seconds from "I need the maximum dose of this in renal impairment" to the correct figure on screen. A bank does not train this because the answer is already on the page.
Prescription construction
Given a blank prescription, a safe entry has to specify the approved drug name, the dose and its units, the route, the frequency or timing, the formulation where it matters, and any critical instruction (for example, "with food", "stop on day 5", or a stat versus regular distinction). The failure mode is omission: a correct drug at the wrong strength, a missing route, or an ambiguous frequency. Recognition practice never surfaces omission because the option already contains all the parts.
Dose, route and frequency accuracy
This is where approximate knowledge is punished. Enoxaparin for treatment versus prophylaxis, gentamicin by an appropriate regimen and route, insulin written as units without abbreviation, a controlled drug written in full — these are exactness problems, not concept problems. The observable behaviour is the proportion of your constructed prescriptions that are safe as written, marked against the criteria the official papers use, not "close enough".
Time pressure and sequencing
At roughly two minutes an item and with the Prescribing section front-loaded for effort, pacing is a skill in its own right. The failure mode is spending nine minutes rescuing one difficult prescription and leaving four two-mark items unread. The observable behaviour is completion: did you attempt all 60 items, and did the mark-dense section get the time it deserves.
A trainable standard for each skill
For each under-trained skill, fix an observable behaviour, a deliberate-practice task, a feedback source and an exit standard. Train the skill in isolation before you integrate it, exactly as you would drill a component before a full mock.
| Skill | Observable behaviour | Deliberate-practice task | Feedback source | Exit standard |
|---|---|---|---|---|
| Formulary navigation | Seconds from question to correct on-screen fact | Timed "find the fact" drills using the NICE-hosted online formulary and MedicinesComplete | Stopwatch + SmPC/eMC to confirm the figure | Under ~30 seconds for a routine dose or interaction lookup |
| Prescription construction | Proportion of blank-field prescriptions that are complete and safe as written | Write 10 prescriptions/day from a clinical stem, no options | Official practice-paper criteria; a prescriber's review | 9/10 complete with no safety-critical omission |
| Dose/route/frequency accuracy | Proportion correct on strength, route and frequency specifically | Focused sets on high-risk drugs (anticoagulants, insulin, opioids, gentamicin) | SmPC/eMC and CKS; a clinician for edge cases | Zero wrong-route or wrong-frequency errors on high-risk drugs |
| Time pressure / sequencing | All 60 items attempted; mark-dense section protected | Full timed papers with a per-section clock | Your own timing log | Finish with time to review, Prescribing section not rushed |
A four-week modality ladder
Skills transfer badly when you only ever practise them integrated. The ladder moves from isolated drilling to unseen simulation, so each layer is measurable before you add the next. This is a pattern, not a proprietary schedule — adjust it to your weeks available.
- Week 1 — isolated skill. Drill formulary navigation and calculation speed on their own. Write single prescriptions from stems without options. The aim is competence in the component, not a realistic exam.
- Week 2 — coached case. Take integrated cases but allow yourself to pause, check the formulary and reason aloud. Score each constructed prescription against the official criteria and record every omission as a coded error. Coaching feedback — from a prescriber, a study partner, or a well-grounded tool used carefully — matters most here.
- Week 3 — timed integrated case. Remove the pauses. Do mixed sets under a per-section clock so pacing and construction are stressed together. Now you are training performance, not knowledge.
- Week 4 — unseen simulation. Sit full, unseen, timed papers you have never worked before, including the official practice papers as the calibration anchor and unseen third-party items for volume. The readiness signal is a stable score on material you have not seen, at pace, with no assistance — not a rising percentage on a bank you have half-memorised.
When AI feedback helps, when it does not, and when you need a human
Automated feedback is genuinely useful for the recognition-format question types and for explaining why a distractor is wrong; it is fast, patient and available at 2 a.m. It is also useful as a first-pass check on a calculation. It is unreliable, and sometimes confidently wrong, when it grades a constructed free-text prescription: whether an instruction is unambiguous, whether a route is appropriate for this patient, and whether an omission is safety-critical are judgements that a general model can get wrong without flagging its uncertainty. Before you trust any automated score on a constructed answer, calibrate it against the official criteria — the method is set out in calibrating automated feedback. You need a human prescriber — a clinical pharmacist, an educational supervisor, or a peer who prescribes — for the final sign-off on prescription construction and for any high-risk drug where the exam's tolerance is narrow. Reserve the clinician for the judgement calls; let the software carry the volume.
A balanced case matrix
The commonest self-sabotage in PSA revision is practising the comfortable scenario repeatedly — the same three cardiology prescriptions — while whole domains go untouched. Build a matrix and sample it deliberately. Rows are the seven clinical domains; columns are the eight question types. You do not need every cell, but you should not have an empty row or an empty high-weight column.
| Domain | Prescribing | Review | Calc | ADR | Monitoring | Other types |
|---|---|---|---|---|---|---|
| Medicine | ✓ | ✓ | ✓ | ✓ | ✓ | sample |
| Surgery | ✓ | ✓ | ✓ | ✓ | sample | sample |
| Elderly Care | ✓ | ✓ | sample | ✓ | ✓ | sample |
| Paediatrics | ✓ | sample | ✓ | ✓ | sample | sample |
| Psychiatry | ✓ | ✓ | sample | ✓ | ✓ | sample |
| O&G | ✓ | sample | ✓ | sample | sample | sample |
| General Practice | ✓ | ✓ | ✓ | ✓ | ✓ | sample |
Prioritise the Prescribing column in every domain, because that is where the marks and the construction demand concentrate. Paediatric and elderly prescribing deserve specific attention: weight-based dosing and renal dose adjustment are exactly the places where approximate knowledge fails.
Red flags in your own preparation
Stop and change something if you notice any of these: memorised scripts — you can only write the prescriptions you have already seen, and a novel stem freezes you; repeated cases — your rising percentage comes from re-attempting familiar items, not unseen ones; generic feedback — your review notes say "revise anticoagulation" rather than "I omitted the review date on a DOAC prescription"; uncalibrated scoring — you are trusting an automated mark on a constructed prescription without ever checking it against the official criteria; and no official-rubric check — you have never scored yourself against the PSA's own practice-paper standards, so your internal pass mark is a guess. Each of these produces a confident candidate with a hidden gap.
Frequently asked questions
How do I know whether I have covered the full Prescribing Safety Assessment blueprint? Coverage is a matrix, not a total. Map the eight question types against the seven clinical domains and mark which cells you have genuinely practised on unseen items, with the Prescribing row and the high-risk-drug entries weighted most heavily because they carry the most marks and the most construction demand. You have covered the blueprint when no domain is empty, no high-weight question type is untouched, and you have confirmed dose accuracy on the recurring high-risk drugs — anticoagulants, insulin, opioids, aminoglycosides. A completed bank with an empty O&G or paediatric row is not covered, however high the percentage.
Can one question bank be enough for Prescribing Safety Assessment? One good bank can be sufficient for the recognition-format types and for volume, but no single bank should be your only preparation, because the exam also tests construction, formulary navigation and pacing that a bank under-trains. The minimum stack is the official practice papers as your calibration anchor, one primary bank for volume, and deliberate prescription-construction practice marked against the official criteria. If you use a second bank, add it to measure transfer on unseen items rather than to repeat content — the discipline for that is the two-Q-bank rule.
What should I measure instead of my overall Q-bank percentage for Prescribing Safety Assessment? Measure four things the percentage hides: your unseen, timed score on material you have never worked; your prescription-construction accuracy, scored against the official criteria and reported as "safe as written, no omission" rather than "roughly right"; your dose, route and frequency error rate on high-risk drugs specifically; and your completion — did you attempt all 60 items with the Prescribing section adequately timed. A rising overall percentage on a re-seen bank can coexist with a falling unseen score, which is the pattern that fails candidates.
When should I stop doing new Prescribing Safety Assessment questions? Stop adding new questions when new items stop changing your error profile — when a fresh unseen block produces the same small set of coded errors you have already logged, rather than new categories of mistake. At that point the marginal question teaches nothing, and your time is better spent on timed full papers, prescription-construction reps on your weak domains, and calibrating against the official practice papers. Volume for its own sake past this point inflates your percentage without improving readiness.
Which Prescribing Safety Assessment resource should I use for my weakest component? Match the tool to the deficit rather than defaulting to more questions. If the weak component is calculation, drill focused calculation sets and use the SmPC/eMC to check every figure. If it is prescription construction, you need blank-field writing practice marked by a prescriber or against the official criteria, not more recognition items. If it is formulary navigation, do timed "find the fact" drills in the actual on-screen resources. If it is pacing, only timed full papers will fix it. Our companion decision-tree article maps each learner profile and weakness to a recommended resource in detail.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 20 July 2026. The PSA format and mark weightings are taken from the official PSA blueprint and practice materials; the eight-question-type mark table was verified on the PSA/British Pharmacological Society learner portal on 20 July 2026. Exam bodies revise formats — confirm the current blueprint on the official PSA website before you sit. UK medicines facts in this article should be checked against the SmPC/eMC, with NICE and CKS for practice guidance.
Disclosure: iatroX operates a PSA question bank and clinical-knowledge tools, so it competes with the resources discussed here. We have confined iatroX's role to the jobs a recognition bank does not claim — unseen, timed measurement and knowledge grounding — and named plainly where a human prescriber or the official material is required instead. Corrections are welcome via the feedback route on iatrox.com.
References: PSA blueprint, question-type weightings and practice papers (prescribingsafetyassessment.ac.uk and the BPS Assessment learner portal); MSC Assessment PSA guidance; NICE-hosted online formulary and MedicinesComplete (the resources available in the exam); SmPC/eMC and NICE/CKS for medicines content. Internal: the iatroX PSA bank, the comparison hub, and Your Q-Bank Percentage Is Not Your Exam Score.
Complete a fresh Prescribing Safety Assessment baseline in iatroX →