Ordinary multiple-choice practice trains one thing well: recognising the single best answer to a written stem, at pace, without help. It cannot tell you whether you can take an accurate history from a frightened parent, examine a distressed toddler, judge a child's development against milestones in real time, prioritise problems out loud, or explain a management plan to a family. Those are the skills the RACP Divisional Clinical Examination assesses, and no question bank reproduces them — iatroX included. This article maps what the multiple-choice format proves, what it does not, and how to train the gap before the clinical exam, not after you have failed it.
This is the exam-level hub for that single question. The platform-specific articles in this series link up to it; here we stay on the skills, not the products.
The official format: two exams testing two different things
The RACP pathway for Paediatrics & Child Health has a written barrier and a clinical barrier, and they are not interchangeable. You must pass the Divisional Written Examination (DWE) before you sit the Divisional Clinical Examination (DCE).
The Divisional Written Examination is a single-day, computer-marked test delivered as two papers. Based on the current RACP and Pearson VUE examination information (verify the live figures on racp.edu.au before you plan), it comprises a Clinical Applications paper of 100 questions — reported as 92 A-type single-best-answer MCQs plus 8 extended matching questions, over three hours with ten minutes' additional reading time — and a Medical Sciences paper of 70 questions — 66 A-type MCQs plus four extended matching questions, over two hours with ten minutes' reading. That is roughly 170 questions in total, predominantly single-best-answer, with no negative marking, and a break between papers. The Paediatrics & Child Health and Adult Medicine divisions sit their own versions on the same day.
The Divisional Clinical Examination is a different test of a different skill set. In Paediatrics & Child Health it comprises two long cases and four short cases. Each long case gives you 60 minutes with the patient, 10 minutes to prepare, and 25 minutes of discussion with two examiners; each short case gives you 2 minutes to read a stem and 15 minutes with the patient. Long cases are marked on the accuracy of your history and examination, your synthesis and prioritisation of the child's problems, your understanding of the illness's impact on the child and family, and the quality of your management plan. Short cases are marked on your interaction with the child and family, your examination technique and its accuracy, your interpretation of physical findings, and your investigation and management. Each case is scored out of six against predetermined criteria, not on a curve.
Read those two descriptions again. The written exam rewards fast pattern-recognition of a printed vignette. The clinical exam rewards performance in front of a real child, a real family and two examiners who are watching your hands, your questions and your reasoning. A question bank is built for the first and structurally cannot deliver the second.
Knowledge and performance are not the same thing
A correct selected answer proves something narrow and real: that, given a curated stem with the salient features already extracted for you, you can recognise the most likely diagnosis or the next best step. That is genuine knowledge and it is necessary. It is not sufficient.
A correct answer does not prove that you could have elicited those salient features yourself from a rambling, anxious or incomplete history. It does not prove you can perform the cardiovascular or neurological examination that generated the sign in the stem. It does not prove you can weigh two competing problems in a medically complex child and say which matters most this week. It does not prove you can deliver a diagnosis of a lifelong condition to parents without either frightening them or falsely reassuring them. The written exam hands you the data; the clinical exam makes you generate the data and then defend what you did with it. Confusing the two is the single most common reason strong written candidates are surprised by the clinical exam. For the underlying trap — a high bank percentage that does not transfer — see Your Q-Bank Percentage Is Not Your Exam Score.
The skills MCQs under-test, and how to train each one
Three clusters of skill fall outside what any MCQ bank can assess: paediatric long and short cases, developmental assessment, and communication with children and families. For each, you need an observable behaviour to practise, a deliberate-practice task, a credible feedback source and an explicit exit standard. Generic revision does none of these; deliberate practice targets one behaviour, gets specific feedback, and repeats.
| Skill cluster | Observable behaviour to train | Deliberate-practice task | Feedback source | Exit standard |
|---|---|---|---|---|
| Long case | Structured history + focused exam in 60 min, then a prioritised, family-aware presentation | Weekly timed long case on a real ward child; present to a consultant against the RACP marking domains | Supervising paediatrician using the official criteria | Two consecutive long cases scored at pass standard by different assessors |
| Short case | Slick, gentle, accurate system examination in 15 min with a running commentary | Daily 15-minute focused examination (cardiac, gait/neuro, abdomen, dysmorphology) observed at the bedside | Registrar or consultant observing technique and findings | Correct signs elicited and interpreted in three different systems, unassisted |
| Developmental assessment | Placing a child's gross motor, fine motor, language and social milestones against age and flagging deviation | Assess development in clinic children across age bands; state the developmental age before checking the record | Community/developmental paediatrician; the child's actual record | Age estimate within the expected band and red-flag deviations named, across five children |
| Communication | Explaining diagnosis, uncertainty and a plan to a family; checking understanding | Role-play breaking a new diagnosis and negotiating a plan; be observed | Clinician or trained simulated parent, against a communication rubric | Family (or simulator) confirms understanding; examiner rates domain at pass |
Notice that the feedback source in every row is a clinician or a real patient encounter, never a screen. That is not an accident. These behaviours produce no machine-checkable single answer, so the only valid signal comes from a human watching you do them.
A four-week modality ladder
Skills like these are not learned by reading about them; they are built in stages, each adding a constraint the previous stage lacked. Run this ladder for whichever cluster is weakest, then repeat for the next.
Week 1 — isolated skill. Strip the task to its smallest unit and drill it slowly, without a clock. Practise a single system examination until the sequence is automatic; rehearse a developmental screen for one age band; script the opening two minutes of breaking bad news. Accuracy before speed.
Week 2 — coached case. Add a supervisor. Perform the whole component — a full short case, a full developmental assessment, a full family conversation — with a clinician watching and marking against the official domains. The point of this week is calibration: you find out where your self-assessment and an examiner's assessment diverge.
Week 3 — timed integrated case. Add the clock and the exam structure. Run a full long case to the real timings (60/10/25) or four back-to-back short cases at 15 minutes each. You are now training stamina and time discipline, not just technique.
Week 4 — unseen simulation. Add unfamiliarity. Do a mock on a child you have never met, ideally in an unfamiliar setting, marked by an assessor who does not know your habits. This is the closest safe proxy to exam day and it is where memorised scripts break down.
Only the very last rung has any use for a simulator or an AI tool, and even then only as an add-on. The rungs that matter — coached, timed, unseen — need real children and real examiners.
When AI feedback helps, when it misleads, and when only a clinician will do
AI feedback has a genuine but bounded role in this preparation, and being precise about the boundary saves you from trusting it where it is weakest.
AI feedback is useful for the knowledge scaffolding beneath the clinical exam: generating unseen practice questions on the paediatric blueprint, drilling the differential for a presentation, quizzing you on developmental red flags, or producing a structured model answer you can compare your own reasoning against. It is fast, available at 2am, and tireless. Used this way it strengthens the written-knowledge layer that every clinical performance stands on.
AI feedback is unreliable the moment it is asked to grade a performance it cannot observe. It cannot see whether your hands were gentle, whether the toddler cooperated, whether the parent understood, or whether the sign you claim you elicited was actually there. A tool scoring a transcript of a case is scoring your description of what happened, not what happened. Automated scores on open performance drift, reward fluent wording over correct substance, and give the same confident number to a safe answer and an unsafe one. Before you trust any automated score on a spoken or written response, calibrate it against a human marker first, as set out in calibrating AI-graded SAQs and OSCEs, and audit the tool itself for grounding and hallucination as in how to audit an AI medical exam tutor.
Only a clinician will do for the actual DCE competencies: examination technique, developmental judgement, synthesis under questioning, and family communication. These are the marked domains, and the examiners marking them are experienced paediatricians. Substitute a human for a screen here without apology.
A balanced case and task matrix
Left to our own preferences, we practise what we are already good at and on the children who happen to be on our ward. That produces a candidate who is fluent on cardiology and helpless on the neurodisability long case. Build a matrix and tick every cell before exam day, so coverage is deliberate rather than accidental.
| System / domain | Short case | Long case | Developmental | Communication |
|---|---|---|---|---|
| Cardiac / respiratory | ☐ | ☐ | — | ☐ |
| Abdominal / nutrition / growth | ☐ | ☐ | — | ☐ |
| Neurology / neurodisability | ☐ | ☐ | ☐ | ☐ |
| Development & behaviour | — | ☐ | ☐ | ☐ |
| Neonatal / genetic / dysmorphology | ☐ | ☐ | ☐ | ☐ |
| Chronic / complex / oncology | ☐ | ☐ | — | ☐ |
Aim for breadth first, then depth on the cells you fear. A child with a chronic, multi-system condition is the classic long case precisely because it forces prioritisation and family-impact reasoning — do not leave it to exam day. For turning a blueprint into a coverage plan rather than a completion count, use the blueprint-coverage matrix method.
Red flags that you are training the wrong thing
Watch for these patterns; each one feels like progress and is not.
- Memorised scripts. Reciting a fixed opening to every long case. Examiners probe, and a script cannot answer an unexpected question.
- Repeated cases. Practising the same three cooperative children over and over. Familiarity inflates your sense of readiness without building transfer.
- Generic feedback. "That was good" is not feedback. Insist on domain-referenced comments tied to the official marking criteria.
- Uncalibrated scoring. Trusting a number — human or automated — that has never been checked against a second assessor.
- No official-rubric check. Practising without ever having read the RACP marking domains, so you are optimising for your own idea of a good case rather than the college's.
Frequently asked questions
How do I know whether I have covered the full RACP Paediatrics blueprint? You cannot know it from a completion percentage or from how many questions you have answered, because completion measures consumption, not coverage. Build a matrix from the RACP Paediatrics & Child Health curriculum domains and the written exam's Medical Sciences and Clinical Applications structure, then log your first-attempt performance in each cell, marking any cell you have never tested as an explicit blind spot. Coverage means every domain has been sampled with unseen questions and every clinical modality — long case, short case, developmental, communication — has been practised under observation, not that a bank says you are "100% complete".
Can one question bank be enough for RACP Paediatrics? One bank can be enough for the written-knowledge component if it is broad, current and used with spaced retrieval rather than passive reading, but it is structurally insufficient for the exam as a whole, because the DCE tests skills no bank contains. Treat any single bank as covering one job — measuring and building recall — and plan separate, human-supervised practice for the clinical exam. If you add a second bank, add it to reach unseen questions and different phrasing, not to duplicate content; the two-Q-bank rule explains how to do that without destroying your calibration.
What should I measure instead of my overall Q-bank percentage for RACP Paediatrics? Measure first-attempt performance on unseen questions, broken down by blueprint domain, and track the trend over time rather than a single headline figure. Your overall percentage is inflated by repeated questions you now remember and tells you nothing about the domains you have avoided. For the clinical components, measure the number of long and short cases scored at pass standard by different assessors, and the number of developmental and communication encounters rated at pass — those counts predict clinical readiness far better than any written score.
When should I stop doing new RACP Paediatrics questions? Stop adding new questions when your first-attempt, unseen performance has plateaued across domains at a comfortable margin above the standard and your remaining errors are careless rather than knowledge gaps — at that point additional questions add little and displace clinical practice you need more. The stopping signal is not "I have finished the bank"; it is "new questions are no longer changing my domain profile". Redirect the freed time to supervised long and short cases, because that is where your marginal hour now buys the most.
Which RACP Paediatrics resource should I use for my weakest component? Match the resource to the modality of the weakness rather than reaching for whichever tool is most convenient. If the weakness is written knowledge in a domain, a broad question bank with unseen items and spaced retrieval — including a live iatroX RACP Paediatrics block — is the right tool. If the weakness is examination technique, developmental judgement or family communication, no bank helps: book supervised bedside practice, a developmental clinic and observed communication role-play with a clinician who will mark you against the official criteria. The decision-tree companion to this article walks that choice through by profile.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 20 July 2026. Exam-format figures are drawn from the RACP and Pearson VUE examination information current at that date and should be re-verified on racp.edu.au before you rely on them, because sitting-to-sitting details change. iatroX operates a question bank that competes with the products mentioned in this series; its role here is confined to the written-knowledge and unseen-MCQ measurement layer, and it is explicitly not a substitute for the clinical examination or for supervised bedside practice. Corrections are welcome via the feedback route on iatrox.com.
References: RACP Divisional Written Examination (racp.edu.au/trainees/examinations/divisional-written-examination); RACP Paediatrics & Child Health Divisional Clinical Examination (racp.edu.au/trainees/examinations/divisional-clinical-examination/paediatrics-child-health-dce); RACP certification testing with Pearson VUE (pearsonvue.com); iatroX Australia RACP Paediatrics bank (iatrox.com/australia); iatroX comparison hub (iatrox.com/compare).