If you are revising for the MRCGP Simulated Consultation Assessment (SCA) with a single-best-answer question bank, you are training recall, not performance. The SCA marks how you gather data under time pressure, negotiate a shared plan with a simulated patient, and safety-net out loud — behaviours no multiple-choice item can score. This article names the exact skills a Q-bank cannot reach, and gives each one an observable behaviour, a deliberate-practice task, a feedback source and an exit standard.
The SCA format, in brief
The SCA is twelve simulated consultations, each lasting twelve minutes — 144 minutes of assessed contact time — taken in the ST3 year and conducted remotely from a local GP practice over the College's video platform, with a deliberate mix of video and telephone cases. The RCGP runs diets across roughly nine months of the year, and the current fee is £1,207 (RCGP, checked 19 July 2026). Each case is marked by an examiner against three domains — Data Gathering and Diagnosis, Clinical Management and Medical Complexity, and Relating to Others. Every domain is graded on a four-point scale: clear pass, pass, fail, clear fail. Your result is the aggregate mark across all twelve cases against the standard of a newly qualified, independent GP, not a separate pass or fail on each case in isolation. Crucially, the case mix is drawn across the breadth of general practice — different ages, presentations, ethnicities and clinical experience groups — so no candidate can pass on a narrow slice of medicine.
That format tells you something a percentage cannot: the SCA is an assessment of a performed consultation, sampled widely, and scored against descriptors rather than against a key. A question bank operates in a different universe.
What a correct answer proves — and what it does not
Separate knowledge from performance before you spend another evening on questions. A correct SBA selection proves that, when the diagnosis is already framed for you and the right answer sits among five options, you can recognise it. That is genuine and useful; the Clinical Management and Medical Complexity domain rewards safe, guideline-aware management, and you cannot manage what you do not know. But recognition is the floor, not the exam.
A correct answer does not prove that you can elicit the history that makes the diagnosis reachable when nobody has pre-sorted the stem for you. It does not prove that you surfaced the patient's actual concern rather than the one you assumed. It does not prove that you offered options the patient understood, checked they could act on the plan, or safety-netted with a specific, time-bound instruction. And it says nothing about whether you did all of that inside twelve minutes on a video call while a simulated patient behaved like a real one. The SCA scores the gap between knowing and doing, and MCQ percentages are silent on that gap.
The five skills a Q-bank cannot score
The table below names the five capabilities that ordinary question practice leaves untouched. Each has an observable behaviour (what an examiner can see), a deliberate-practice task, a feedback source that can actually judge it, and an exit standard that tells you when to stop drilling it.
| Skill | Observable behaviour | Deliberate-practice task | Feedback source | Exit standard |
|---|---|---|---|---|
| Voice and remote consultation | Clear opening, verbal signposting, managing silence and the absence of body language on telephone cases | Record three timed remote consultations; review for pace, interruptions and dead air | Peer or trainer review against a transcript; self-review of the recording | Consistent structure and pacing on both video and telephone modalities |
| Patient agenda (ideas, concerns, expectations) | Explicitly elicits and responds to the patient's ideas, concerns and expectations, not just symptoms | Role-play cases where the stated complaint hides the real concern | Simulated patient or trainer confirming the hidden agenda was reached | Reaches the true agenda in the first four to five minutes, most cases |
| Shared decision-making | Offers options, checks understanding, negotiates a plan the patient accepts | Practise presenting two management options and reaching an agreed plan | Simulated patient reporting whether they felt involved and understood | The plan is co-produced and the patient can restate it |
| Safety-netting | Gives specific, time-bound advice on what to watch for and when to return | Close every practice case with an explicit, worst-case-aware safety-net | Trainer checking specificity, not just presence, of the safety-net | Safety-net is specific, realistic and delivered in every case |
| Performing to the domain descriptors | Behaviour maps to Data Gathering, Clinical Management and Relating to Others | Score your own recorded cases against the published domain descriptors | Calibrated examiner or trained peer using the RCGP descriptors | Self-scores agree with a calibrated marker within one grade |
Reading down that table, notice how little of it a bank touches. Voice and remote consultation is a modality skill: telephone cases strip away the visual cues you have leaned on all through training, and candidates who never rehearse them arrive over-reliant on body language they cannot see. Patient agenda is the single most common reason strong knowledge produces a weak consultation — the candidate answers the presenting symptom expertly and never discovers that the patient came about something else. Shared decision-making is not "explaining the plan"; it is offering genuine options and reaching one the patient owns, and it is assessed under Relating to Others as much as under Clinical Management. Safety-netting is where good consultations are quietly won or lost: examiners look for specific, time-bound, worst-case-aware advice, not a reflexive "come back if you're worried." And performing to the domain descriptors is the meta-skill — the candidate who has internalised what a clear pass looks like in each domain can self-correct in real time, while the candidate who has only practised questions cannot see the rubric they are being marked against.
A four-week modality ladder
You cannot climb straight from a Q-bank to a mock exam; the rungs are too far apart. Build up in four stages, spending roughly a week on each and only moving up when the lower rung is secure.
Week one — isolated skill drills. Take one skill at a time. Spend a session on safety-netting alone: run five two-minute closes and make each one specific and time-bound. Spend another on agenda elicitation. The point is to overlearn each component so it costs no working memory in a full case.
Week two — coached single cases. Run whole twelve-minute cases with a trainer or study partner playing the patient, then debrief against the three domains. Coached cases let a person interrupt, model and correct in a way no automated tool can. Aim for four to six cases, spread across modalities.
Week three — timed integrated cases. Now add the clock and the constraint. Run cases back to back at exam pace, video and telephone mixed, with no debrief until the block ends. This trains the transition between cases — the reset that catches out candidates who ruminate on the last consultation into the next one.
Week four — unseen simulation. Sit a full or half-length simulated diet using cases you have never seen, ideally marked by someone calibrated to the descriptors. Unseen is the operative word: a case you have rehearsed measures memory, not competence.
Underneath all four rungs sits your knowledge base. This is the one job an MCQ bank does well, and it is where iatroX belongs in an SCA plan — as the unseen-question layer that keeps the clinical substrate for the Clinical Management domain fresh while you spend your consultation-practice time on the rungs above. More on that below.
When AI feedback helps, when it misleads, and when you need a person
Automated feedback has a real but narrow role in SCA preparation, and confusing its lanes is expensive.
AI feedback is genuinely useful for structure and surface features: whether you opened cleanly, whether you interrupted, whether you delivered a safety-net at all, whether you covered ideas, concerns and expectations as discrete items. A transcript-based tool can count these reliably.
It becomes unreliable the moment the judgement is qualitative and contextual — whether your safety-net was appropriate to this presentation, whether the shared decision genuinely reflected the patient's values, whether your empathy landed or merely appeared. Automated scorers reward the presence of a phrase and cannot judge its fit; a candidate can learn to trip the detector while consulting badly. Before you trust any automated score, calibrate it against a human marker on the same cases — the method in our pillar on calibrating AI-graded feedback applies directly here.
And a person is required for the things that decide the exam: nuanced domain-level judgement, reading whether a real interaction felt collaborative, and modelling how an experienced GP would have redirected a drifting consultation. No current tool replaces a calibrated examiner or an experienced trainer for SCA marking. Treat AI as a rehearsal metronome, not a judge.
A balanced case matrix so you do not rehearse only what you like
Candidates gravitate to the presentations they enjoy and quietly avoid the ones they dread. The SCA samples across the whole of general practice precisely to punish that habit. Build a matrix and force breadth across four axes at once:
| Axis | Deliberately sample across |
|---|---|
| Clinical experience group | Acute, chronic, mental health, women's and men's health, children and young people, older adults, cancer and palliative, health promotion |
| Modality | Video and telephone in roughly the proportion the exam uses |
| Patient factors | Age, ethnicity, health literacy, carers and third parties, interpreters |
| Domain stretch | Cases that stress Data Gathering, cases that stress Clinical Management, cases that stress Relating to Others |
Log every practice case against all four columns. If, after three weeks, your log is thick with acute video cases in articulate adults and thin on telephone mental-health cases, older-adult polypharmacy or consultations through a carer, you have found your blind spot before the examiner does.
Red flags that you are practising the wrong way
Five patterns reliably predict a disappointing SCA despite heavy revision. Audit yourself against them honestly.
- Memorised scripts. If you are reciting an opening or a safety-net verbatim, you will sound rehearsed and you will not adapt when the patient goes off-piste. Scripts are scaffolding to be removed, not lines to be performed.
- Repeated or leaked cases. Practising the same cases until you know the "answer" measures memory, not consultation skill, and any bank of cases circulating as the real thing is worthless and possibly a probity risk. Value unseen cases over familiar ones.
- Generic feedback. "That was good" trains nothing. Insist on feedback tied to a specific domain and a specific observable behaviour.
- Uncalibrated scoring. A marker who has never calibrated against the RCGP descriptors may be kind, harsh or simply inconsistent. Self-scores that no calibrated marker has ever checked are noise.
- No official-rubric check. If you have never scored a case against the published domain descriptors, you do not know what you are being marked against. Read the marking guidance before your next practice case, not the night before the exam.
Where iatroX fits — and where it explicitly does not
Be plain about this: iatroX is not a consultation or OSCE simulator, and it does not attempt to be. As of 19 July 2026 the iatroX SCA question bank is not yet live — the platform states openly that SCA-specific content is still in development and, in the meantime, points candidates to dedicated case-banks and role-player resources such as SimsBuddy and Bradford VTS, and to study-group practice, for the performance rungs of the ladder.
What iatroX does provide today is the knowledge substrate. Its live, free MRCGP AKT bank keeps the clinical and prescribing knowledge that feeds the Clinical Management and Medical Complexity domain sharp, and gives you an unseen, mixed, timed way to measure that knowledge rather than assume it. Use it for the substrate; use human-marked, unseen consultation practice for everything the SCA actually scores. If you want to see how a knowledge bank and a consultation resource divide the work without overlap, our two-Q-bank rule pillar sets out the same logic for any exam, and the comparison hub maps which tool covers which job.
Frequently asked questions
How do I know whether I have covered the full MRCGP SCA blueprint? You have covered it when your case log shows deliberate, human-marked practice across every clinical experience group, both modalities, and a spread of patient factors — and when your self-scores against the three domain descriptors agree with a calibrated marker. Coverage for the SCA is not a number of questions; it is a breadth of performed cases sampled the way the exam samples, with evidence that each of the three domains has been stretched, not just Data Gathering. If any cell of the case matrix above is empty, your blueprint is incomplete regardless of how high your knowledge-bank percentage is.
Can one question bank be enough for MRCGP SCA? No. A question bank cannot assess or train a single one of the five skills the SCA marks — voice consultation, patient agenda, shared decisions, safety-netting or performance to the domain descriptors — because all of them are behaviours in a live twelve-minute consultation, not recall of a best answer. A bank is a strong option for the knowledge that underpins the Clinical Management domain, and it belongs in your stack for exactly that job, but on its own it prepares you for an exam the SCA is not. Pair it with human-marked, unseen consultation practice.
What should I measure instead of my overall Q-bank percentage for MRCGP SCA? Measure domain-level performance on human-marked, unseen cases: the proportion of cases in which you reached the patient's true agenda inside five minutes, delivered a specific safety-net, and co-produced a plan the patient could restate. Track those as rates across a log, alongside the agreement between your self-scores and a calibrated marker. An overall percentage from a knowledge bank tells you about the substrate only — and as our standing caveat, your Q-bank percentage is not your exam score, puts it, that number was never designed to predict a performance exam.
When should I stop doing new MRCGP SCA questions? Stop adding new knowledge questions when unseen, timed, mixed blocks show your clinical management substrate is stable and your errors are no longer knowledge gaps but consultation-delivery gaps — because at that point more questions cannot move your score, and every hour spent on them is an hour not spent on human-marked cases. For SCA cases, you never really "stop" until the exam, but you should shift from isolated drills to unseen timed simulation once the lower rungs of the ladder are secure. The trigger to stop is a stable measured signal, not a target number reached.
Which MRCGP SCA resource should I use for my weakest component? Match the resource to the component. If the weakness is knowledge feeding Clinical Management, a mixed unseen MCQ bank such as the iatroX AKT bank is the efficient fix. If it is agenda, shared decisions, safety-netting or remote-consultation technique, no MCQ resource will help — use a dedicated case-bank with role-players and a calibrated human marker (for example SimsBuddy or Bradford VTS material with study-group or trainer marking). Diagnose the component first with human-marked cases, then choose; using a knowledge tool for a performance gap is the commonest wasted month in SCA preparation.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026; exam fees, diet counts and platform features change, so confirm the current figures on the RCGP and vendor pages before you rely on them. Vendor-reported figures are labelled as such and dated. Disclosure: iatroX operates a UK question bank and its own SCA content is in development and not yet live; this article confines iatroX's role to the knowledge-substrate and unseen-measurement jobs that a consultation simulator does not claim, and names independent case-bank and role-player resources for the performance work. Corrections are welcome via the feedback route on iatrox.com.
References: RCGP — Simulated Consultation Assessment, format and marking guidance (rcgp.org.uk/mrcgp-exams/simulated-consultation-assessment and /marking-and-results); iatroX MRCGP SCA and AKT landing pages (iatrox.com/mrcgp-sca, iatrox.com/mrcgp-akt); iatroX — "Your Q-Bank Percentage Is Not Your Exam Score"; iatroX framework pillars on calibrating AI-graded feedback and the two-Q-bank rule.