Critical-appraisal revision becomes more useful when each wrong answer identifies a particular misunderstanding. Reopening an entire statistics textbook after every disappointing mock makes it difficult to tell whether that misunderstanding has changed. Start with the decision the question required, reconstruct your reasoning, and choose a different problem that tests the same distinction.
The Royal College of Psychiatrists lists Paper B on 7 October 2026. Its preparation guidance, checked on 3 October 2026, allocates approximately one-third of the paper to critical review and says that the revised syllabus takes effect in January 2027, not during the remaining 2026 sittings. Use the applicable syllabus rather than reorganising your final review around next year's changes. See the College's examination dates and preparation guidance.
Find the decision hidden inside the question
Before choosing a formula, describe the task in ordinary language. Is the question asking how participants were selected, what could distort the comparison, how large an effect appears, how uncertain the estimate is, or whether the finding applies elsewhere? Those are different decisions even when the same abstract supplies the information.
A useful review begins by covering the options and completing the sentence: "To answer this, I need to establish..." Then identify which details in the stem establish it. This makes it harder to choose a familiar statistical term simply because it appears among the options.
For example, a question containing a confidence interval is not necessarily asking whether a result is statistically significant. It might ask about precision, the direction of the effect or the interpretation of a ratio. Locate the requested judgement before examining the calculation.
Separate study design from weaknesses in its execution
A study's label and the threats to its interpretation belong in separate columns. First establish how exposure, allocation, comparison and follow-up were organised. Then ask what went wrong, or might have gone wrong, within that design.
In a fictional service evaluation, people choose whether to join a new group programme. Outcomes are compared with those of people who decline. Calling the report a prospective study does not make allocation random. Conversely, identifying a randomised design does not settle whether outcome assessment, missing data or subsequent exclusions threaten its interpretation.
Write the mechanism of the possible bias, not just its name. Explain which comparison might be distorted and why. Where the direction is uncertain, say so rather than assuming that every weakness necessarily exaggerates benefit. This exercise is more demanding than matching a keyword to a label, but it gives the next question a clearer target.
Keep the denominator attached to the effect
Consider a constructed teaching example, not research findings: an outcome occurs in 18 of 100 people in one group and 12 of 100 in another, over the same follow-up period. The absolute difference is six percentage points. The relative reduction is one-third. Both describe the same fictional observations, but answer different questions.
During review, write the event, population, comparison and follow-up beside the result. A percentage without its denominator is easy to misremember. A favourable relative effect also does not establish the importance of the absolute benefit for a particular patient or service.
The Cochrane Handbook's chapter on effect measures, checked on 3 October 2026, distinguishes the measures used for different outcome types. Use the relevant explanation to repair the concept rather than reading the whole chapter regardless of the error.
Explain uncertainty without translating it into certainty
After checking an answer, explain what the information permits you to conclude and what it leaves unresolved. A result compatible with no difference is not automatically proof that two interventions are equivalent. A statistically significant difference does not, by itself, establish that the effect is important or that the design supports a causal interpretation.
A fictional candidate, Daniel, repeatedly chooses answers that describe inconclusive results as showing no effect. His correction is not another page of formulae. He writes two separate statements: the estimate observed in the study and the uncertainty around that estimate. He then practises deciding which claims overstate the evidence.
Keep this distinction visible when reviewing correct answers too. Daniel may select the right option because the alternatives are obviously implausible while still being unable to explain the result independently. That item deserves attention even though it did not lower his score.
Build a short final-review document
Use four fields: the decision required, the mistaken assumption, the correction and the next test. An entry might read: "Distinguish relative from absolute effect; treated a relative change as an absolute difference; identify both event rates first; retry with a different baseline risk."
Do not copy the whole stem or explanation. Record enough to reconstruct the distinction without turning the log into a substitute textbook. Give priority to recurring mistakes, misconceptions that affect several topics and errors you still cannot explain after reading feedback.
Retest with changed wording, a different denominator or another study design. Repeating an identical question immediately mostly tells you whether you remember its answer. A later unfamiliar item provides a more useful challenge to the correction you believe you have made.
Use guided discussion for one unresolved concept
In its September 2026 product information, iatroX describes a Socratic Tutor that opens on the attempted question and uses targeted follow-ups to explore the learner's misconception. For Daniel, the useful discussion is why "not demonstrated" and "demonstrated absent" are different, followed by an independent attempt at a new example. The Tutor is a learning aid, not an official Paper B examiner.
Ask-iatroX and free question access have no trial expiry or verification gate under the same product information. Ongoing Tutor, full question banks, study planner, simulations and CPD tools are included together in the paid subscription. In this final period, additional access should solve a named problem, not become another source of unreviewed material.
Finish each session with a statement you can defend without the explanation open. The target is not familiarity with every statistical phrase, but a clearer account of how the study supports, limits or fails to support the conclusion being offered.
Frequently asked questions
Should I revise every critical-appraisal topic in the final days?
Keep the applicable syllabus visible, but concentrate detailed review on recurring or consequential gaps. Another broad read is less useful when it displaces checking whether a specific misconception has changed.
Does getting the calculation right mean I understood the question?
No: the requested measure, denominator and interpretation must also be correct. Check that the number answers the question actually asked.
Does the new MRCPsych syllabus apply to October 2026 Paper B?
The College's preparation page checked on 3 October 2026 says the revised syllabus takes effect in January 2027. Follow the syllabus and instructions applicable to your booked sitting.