A single-best-answer bank can make you fast and accurate at recognising the correct option. It cannot, on its own, prepare you to interpret a live arterial trace at the bedside, to talk an examiner through your reasoning under time pressure, to justify an ethical decision aloud, or to handle equipment in front of a marker. Those are the exact skills the Final FFICM tests outside the written paper — and where most well-drilled candidates lose marks. This is the exam-level hub; the platform-specific workflows link here.
The official format map
The Final FFICM has three components, and you must pass the MCQ before you can attempt the other two; that pass is valid for three years.
| Component | Format (verified on ficm.ac.uk, 20 July 2026) | What it primarily tests |
|---|---|---|
| MCQ | 130 single-best-answer questions, 3 hours; 80 short one-mark and 50 long two-mark items (max 180). All-SBA since June 2022 | Breadth of factual and applied knowledge |
| OSCE | 13 seven-minute stations (12 scored, 1 test), domains: Data, Equipment, Professionalism, Resuscitation | Applied performance: data, imaging, equipment, communication, resuscitation |
| SOE | 4 fourteen-minute stations built from 8 questions, each scored by two examiners | Structured oral reasoning under time pressure |
The MCQ is the knowledge gate. The OSCE and SOE are performance assessments, and no volume of MCQs converts directly into performance. iatroX and every other bank named in the child articles sit at the MCQ knowledge layer; they measure and build knowledge, and they explicitly do not replace the OSCE or SOE.
Separate knowledge from performance
A correct selected answer proves you can recognise the right option when it is written down in front of you, with four distractors to react against, and no one watching the clock or your face. It does not prove you can generate that answer from a blank page, defend it when an examiner pushes back, interpret the raw data that the MCQ has already pre-digested into a stem, or perform a task with your hands. The gap between recognition and generation is precisely the gap between the MCQ and the OSCE/SOE. Most candidates who "know the material" but under-perform have trained recognition to a high level and generation barely at all.
A worked example: the strong-MCQ candidate who stalls
Consider a trainee who scores consistently in the high 80s on her question bank and has passed the MCQ comfortably. In an OSCE data station she is handed a raw arterial blood gas, a lactate trend and a set of ventilator numbers and asked to talk the examiner through them in seven minutes. She freezes — not because she cannot define a mixed acid–base disturbance, but because the bank always told her that was what it was. Here she must find it, sequence it against the ventilator data and commit to a plan out loud while the clock runs. She reaches the right answer at five minutes, too late to act on it, and her narration is a list of observations rather than a prioritised argument.
Her SOE goes the same way: asked to structure an approach to the crashing post-operative patient, she produces accurate facts in no particular order, and when the examiner probes she repeats rather than adapts. None of this is a knowledge deficit. It is a generation-and-performance deficit, invisible to her bank percentage, and it is closed only by the modality ladder below — timed raw-data narration, rehearsed viva structure and rubric-scored feedback from a human. Her MCQ score was necessary and genuinely earned; it simply measured a different thing from what the OSCE and SOE reward.
The under-tested skills, one by one
Data interpretation. In the MCQ, the numbers arrive interpreted: the stem tells you the lactate is rising and the pressures are climbing. In the OSCE data station, you are handed the raw gas, the raw haemodynamics, the raw trend, and you must extract the abnormality, sequence it and act — out loud, at pace. The skill is not knowing what a high lactate means; it is finding it, prioritising it and narrating a plan while the clock runs.
Imaging. A bank can show you a chest film with the answer keyed. The OSCE asks you to describe a film or an ultrasound systematically, state what you would do next, and defend it. Systematic description under time — not pattern-matching to a labelled image — is the trainable skill, and it is invisible to a multiple-choice format.
Ethics and professionalism. MCQ ethics collapses to choosing the "most appropriate" option. Live, the OSCE and SOE ask you to reason through capacity, consent, resource limitation, end-of-life decisions and communication with a distressed relative — coherently, humanely and defensibly, without a menu of options to anchor to. The reasoning has to be generated and voiced, not selected.
Structured oral reasoning. The SOE is the purest example. Eight questions, four stations, two examiners each, fourteen minutes apiece: you must structure an answer, commit to a position, and adapt as the examiner probes. This is a performance skill with its own grammar — signpost, prioritise, justify, safety-net — and it is trained by rehearsal and feedback, not by reading.
Equipment and resuscitation. The OSCE examines Equipment and Resuscitation as explicit domains, and both are pure performance. Recognising a device in a photograph is an MCQ skill; assembling it, checking it, troubleshooting it, or leading a resuscitation sequence with the right priorities and clear closed-loop communication is not. These stations reward rehearsed, physical, spoken competence built by doing — in a skills room, against a checklist, with someone watching — and reading about them moves the needle very little.
For each skill: behaviour, task, feedback and exit standard
| Skill | Observable behaviour | Deliberate-practice task | Feedback source | Exit standard |
|---|---|---|---|---|
| Data interpretation | Finds and prioritises the abnormality aloud, then acts | Timed raw-data drills narrated to a colleague | Colleague/consultant against FICM sample answers | Consistent correct sequence within station time |
| Imaging | Describes systematically, states next step | Two-minute structured image descriptions | Radiology-aware colleague | Full systematic description, no omissions, on time |
| Ethics/professionalism | Reasons through principles, communicates humanely | Scripted scenarios role-played with a partner | Consultant/examiner-style rubric | Defensible reasoning without a prompt menu |
| Structured oral reasoning | Signposts, commits, adapts under probing | Rehearsed viva pairs against SOE-style stems | Trained examiner or study group | Coherent, safety-netted answers at SOE pace |
A four-week modality ladder
Do not jump from reading to mock exam. Climb the ladder so difficulty rises deliberately.
- Week 1 — isolated skill. Drill one modality at a time: a set of raw gases, a set of films, a set of ethical stems. Slow, un-timed, focused on method.
- Week 2 — coached case. Take an integrated case with a colleague who interrupts and probes. Still forgiving on time; the aim is structure and out-loud reasoning.
- Week 3 — timed integrated case. Run stations at true station length — seven minutes for OSCE-style, fourteen for SOE-style — with a partner marking to a rubric.
- Week 4 — unseen simulation. Full mock conditions on material you have not seen, ideally in a group or on a course, scored against the official rubric.
Your MCQ knowledge — built and measured on a bank such as iatroX or the platforms in the child articles — feeds every rung, but the climbing itself is performance practice.
When AI feedback helps, when it does not
Automated feedback is useful for structured, rule-bound tasks: whether your systematic image description included every heading, whether your data narration reached the abnormality, whether your answer omitted a safety-net. It is unreliable for judgement-heavy, contextual marking — the humanity of a breaking-bad-news station, the defensibility of a resource-limitation decision, the subtle prioritisation an examiner rewards. And it cannot replace a human where the mark depends on live interaction and professional judgement, which is most of the OSCE and all of the SOE. Calibrate any automated score against a human and the official rubric before you trust it; treat a machine grade as a prompt for a human check, never as the verdict.
Reading your performance scores
A rubric-scored rating on a timed, unseen station is worth far more than a self-graded impression, but it still needs reading with care. A single strong station no more proves competence than one weak station proves failure — look at the pattern across several unseen, timed attempts, and weight the ones scored by someone other than yourself. Watch the trend by domain: rising data-station scores alongside static ethics scores tell you exactly where the next fortnight belongs. And keep the two currencies separate — your MCQ accuracy and your station ratings answer different questions, and averaging them into a single feeling of "readiness" hides the very component that will cost you marks. As with the written paper, a score is a prompt for a decision, not a verdict.
Build a balanced case matrix
Candidates gravitate to the scenarios they already like. Counter this by building a case matrix crossing the OSCE domains (Data, Equipment, Professionalism, Resuscitation) and the SOE's clinical-science reasoning against acuity and system, and deliberately practising the empty cells. If your log shows ten resuscitation rehearsals and no ethics stations, the matrix has told you where the next fortnight goes. Coverage of the performance blueprint matters exactly as much as coverage of the knowledge blueprint.
Why the official material is your calibration standard
FICM publishes the curriculum, the examination regulations and sample MCQ, OSCE and oral questions. These are not merely more practice material; they are the only calibrated statement of what "good enough" looks like, and every rehearsal you run should ultimately be marked against them rather than against your own sense of a strong answer. Third-party banks and courses are useful precisely to the extent that they move you toward that official standard, and misleading to the extent that they drift from it — an off-format question, an outdated marking scheme, or a viva rubric someone invented all quietly recalibrate you to the wrong target. Before a mock, read the relevant FICM sample answers; after it, mark against them; when a colleague scores your data station or viva, hand them the official rubric first. Let the exam body, not the loudest resource, define the standard you are climbing toward.
Red flags in your preparation
- Memorised scripts. If your viva answers are recited rather than reasoned, an examiner's first unexpected probe will expose it.
- Repeated cases. Rehearsing the same three stations feels productive and measures nothing new.
- Generic feedback. "That was good" is not feedback; feedback names the missed heading, the late prioritisation, the unsafe omission.
- Uncalibrated scoring. A self-assigned or machine mark that has never been checked against the official rubric is a number, not a standard.
- No official-rubric check. If you have not compared your performance to the FICM sample OSCE and oral questions, you are marking to your own imagination.
The bottom line
A question bank is the right tool for one job in FFICM — building and measuring the knowledge that the MCQ gates — and the wrong tool, used alone, for everything the OSCE and SOE test. The candidates who struggle are rarely short of facts; they are short of generation and performance, trained late and measured never. Keep your bank work honest and unseen, cap it when transfer is stable, and spend the reclaimed time climbing the modality ladder against the official rubric. Recognition gets you through the gate; generation gets you through the exam.
Frequently asked questions
How do I know whether I have covered the full FFICM blueprint? Build a coverage matrix that maps both blueprints — the MCQ knowledge domains and the OSCE/SOE performance domains — and mark every cell you have genuinely practised, not merely read. Coverage means you have produced the skill under representative conditions and checked it against the FICM curriculum and sample material, not that you have seen a topic go past. Empty cells are your plan; a bank's completion percentage tells you nothing about the performance cells.
Can one question bank be enough for FFICM? One bank can be enough for a large share of the MCQ if its coverage and volume are adequate, but it cannot be enough for the exam, because the OSCE and SOE assess generation and performance that no multiple-choice format touches. Use a bank for the knowledge gate and measurement, and train the performance components separately with people.
What should I measure instead of my overall Q-bank percentage for FFICM? Measure first-attempt accuracy on unseen items by domain, pacing against 80 seconds per mark, retention across a spaced gap, and — for the performance components — rubric-scored ratings on timed, unseen stations. Your overall bank percentage blends seen and unseen items and rewards familiarity; it is not your exam score, and the child workflows lean on the same point.
When should I stop doing new FFICM questions? Stop adding new questions when your unseen first-attempt accuracy has stabilised across domains, your pacing is on target, and your retention holds across a gap — and redirect that time into the OSCE and SOE performance ladder, which is usually the higher-yield place to be in the final weeks. New questions past the point of stable transfer add reassurance, not readiness.
Which FFICM resource should I use for my weakest component? Match the resource to the deficit: for MCQ knowledge, an unseen bank with good blueprint coverage; for OSCE data and imaging, raw-data and image drills with a colleague and, where useful, a platform with a data-interpretation library; for the SOE, rehearsed viva pairs and a course with examiner-style feedback. If the weak component is a performance skill, the answer is almost never "more MCQs" — it is structured, human-fed practice against the official rubric.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 20 July 2026. Exam-format details are drawn from the Faculty of Intensive Care Medicine's own pages and are correct to that date; FICM periodically revises its examinations, so confirm the current format, station counts and sample material on ficm.ac.uk before you plan. Disclosure: iatroX operates an FFICM MCQ question bank and therefore competes at the knowledge layer; this article deliberately limits the iatroX role to building and measuring MCQ knowledge and states plainly that it does not replace the OSCE or SOE, which require human, performance-based practice. Corrections are welcome via the feedback route on iatrox.com. References: the Faculty of Intensive Care Medicine (ficm.ac.uk) FFICM Final MCQ, OSCE and SOE pages, curriculum, examination regulations and official sample MCQ, OSCE and oral questions; iatroX's FFICM bank; the comparison hub; the guide on why your Q-bank percentage is not your exam score; the pillar on building a blueprint-coverage matrix; and the framework on calibrating AI-graded SAQs and OSCEs.