Article summary
What should you know about The AKP Stem Is Not a Story?
Learn to separate decisive evidence from background noise in long AKP stems using a practical six-step reasoning framework.
How to separate decisive evidence from background noise
A long AKP stem can create the impression that every sentence deserves equal attention.
It does not.
The child may have a chronic diagnosis, an impressive family history, three abnormal blood results, a previous admission, an unusual medication and a worried parent. Yet the question may ultimately turn on one quiet detail: the child is hypotensive, the urine is concentrated, the weakness is ascending, the symptoms began after a new drug, or the lungs are clear.
The skill AKP rewards is not simply noticing abnormalities. It is deciding which abnormality matters for the decision in front of you.
That distinction matters because the RCPCH describes AKP as testing knowledge, understanding and clinical decision-making. The current examination contains 120 single-best-answer questions across two 150-minute papers. More importantly, the College explicitly explains that in a well-written SBA, several options may contain something true, but one is superior in the circumstances given.
That is why approaching an AKP stem as though it were a conventional textbook case can go wrong.
The stem is not there to tell you a beautiful clinical story. It is there to force a decision.
The first question is not "What is the diagnosis?"
It is tempting to begin every vignette by hunting for a disease.
Sometimes that is exactly what the question requires. Often it is not.
A candidate may correctly recognise meningitis and still lose the mark because the question asks for the next investigation. They may recognise nephrotic syndrome but choose the definitive treatment when the question asks what must happen immediately. They may identify an adverse drug reaction correctly but fail because the lead-in asks which drug should now be stopped.
RCPCH's own examples of SBA lead-ins include the most likely diagnosis, the investigation most likely to establish it, the best next step, the best advice to parents, the pathogenesis and the most likely cause.
These are different cognitive tasks.
So before solving the child, identify the task.
If the lead-in says "What is the most appropriate next step?", your answer must survive a timing test. Something can be medically correct eventually and still be wrong now.
That single distinction explains a surprising number of AKP mistakes.
Compress the case before you interpret it
Long stems become difficult when every detail is held in working memory at the same level of importance.
Instead, compress the case.
Try to reduce it mentally to:
Age + tempo + dominant problem + severity + decisive clue
A paragraph about a previously well 8-year-old who had an upper respiratory infection ten days ago, then developed progressive leg weakness, difficulty climbing stairs and absent lower-limb reflexes becomes:
School-aged child + acute progressive ascending weakness + post-infectious + areflexia.
The birthday, school performance and minor laboratory abnormalities may still be true. They simply do not deserve equal weight.
This kind of compression is closely related to the broader "key features" approach used in medical assessment: clinical problems contain critical steps where a candidate's decision determines whether the case is resolved correctly.
The objective is not to throw information away. It is to establish hierarchy.
Find the pivot
Most difficult AKP questions contain a pivot: the feature that changes the ranking of the answers.
Imagine an adolescent with asthma who remains markedly tachypnoeic after repeated bronchodilator treatment. "Severe asthma" is an obvious explanation. But suppose air entry has improved substantially, wheeze has diminished and lactate has risen after intensive salbutamol therapy.
The background diagnosis is still asthma.
The new problem, however, may no longer be worsening bronchoconstriction.
The candidate who keeps responding to the headline diagnosis risks escalating the wrong treatment. The candidate who notices the chronology and treatment response recognises that the decision has changed.
Pivots often hide in chronology, physiological instability, treatment response, trends, thresholds and negative findings.
Negative findings deserve particular respect. "No hepatosplenomegaly", "clear lung fields", "normal blood pressure", "no response to bronchodilator" or "no ketones" can appear less impressive than a dramatic positive finding, yet may exist specifically to defeat the nearest alternative.
Medical reasoning research has demonstrated how salient but irrelevant clinical features can pull clinicians towards an incorrect diagnosis. More recent work on anchoring similarly shows how an early, compelling feature can make it harder to respond appropriately to later contradictory evidence.
In AKP terms: the loudest clue is not necessarily the best clue.
Do not ask whether your answer fits. Ask whether it beats its rival.
This is perhaps the most useful change in SBA technique.
Candidates often choose an answer and then look for evidence supporting it.
But difficult SBAs are rarely difficult because the correct answer does not fit. They are difficult because another answer also fits.
The RCPCH makes this unusually explicit: the strongest SBA options can all be partly correct, while one remains better than the others.
So once you have a provisional answer, identify the option most likely to beat it.
Then ask:
Why is my answer better than that one in this specific child?
Suppose two diagnoses can cause metabolic acidosis. Do not accumulate features supporting both. Find the discriminator.
Suppose two investigations would eventually be appropriate. Which one changes immediate management?
Suppose two treatments belong in the guideline. Which belongs at this stage of the pathway?
Suppose both diseases fit the history. Which one also explains the examination and investigations?
The correct answer often emerges from the difference between the two strongest options, not from repeatedly rereading all five.
Five common ways good candidates get distracted
The same reasoning failures recur often enough to deserve names.
The diagnostic anchor appears when an early feature produces a convincing diagnosis and everything afterwards is forced to fit it.
The chronic-disease magnet appears when every new symptom in a child with sickle cell disease, cerebral palsy, cystic fibrosis or epilepsy is automatically attributed to that condition.
The dramatic-number trap appears when the largest abnormal laboratory value attracts attention even though a different result actually determines management.
The rare-association trap appears when a candidate remembers an obscure association and allows one matching clue to outweigh the overall phenotype.
And the right-later trap appears when an answer is good medicine, but belongs after stabilisation, after confirmation, or after the action the question is asking about now.
These errors are not necessarily failures of knowledge. They are failures of weighting knowledge.
A practical AKP routine
The following is a PaedVault reasoning framework rather than an official RCPCH algorithm, but it maps closely onto the decision-making demands of the SBA format:
- Task - What exactly am I being asked: diagnosis, investigation, interpretation, immediate management or longer-term treatment?
- Compress - Reduce the stem to age, tempo, principal problem, severity and discriminating findings.
- Pivot - Which detail changes the answer?
- Rival - What is the strongest competing option?
- Defeat - What specific finding makes my answer better than that rival?
- Check - Have I answered what should happen now, rather than what might happen eventually?
There is an interesting parallel in the College's own advice for image questions. RCPCH recommends reading the question stem carefully before examining the image and only then looking at the answer options, specifically to reduce confirmation bias. The College notes that the same principle applies more broadly to examination questions.
That is a useful discipline even when there is no image.
Understand the problem before the options start suggesting diagnoses to you.
Review your mistakes differently
After a wrong question, "I didn't know this" is sometimes the correct diagnosis of the problem.
But not always.
You may have known every fact required and still answered incorrectly because you ignored the chronology, missed a negative finding or chose the right management at the wrong point in the sequence.
When reviewing questions, therefore, do not record only the disease.
Write:
The decisive feature was ________.
Then add:
The closest alternative would have been correct if ________.
That second sentence is particularly valuable. It teaches the boundary between two plausible answers.
And those boundaries are where AKP becomes difficult.
Read less. Weight better.
The candidate who improves at AKP does not necessarily become faster at recognising diseases.
They become better at determining what deserves attention.
A long stem may contain ten facts. Perhaps six support the general diagnosis. Two create realism. One points towards a tempting distractor.
And one changes the answer.
Your job is to find that one.
The AKP stem is not a story to be remembered from beginning to end. It is a decision problem whose information has unequal value.
Read the whole case.
Then compress it.
Find the task. Find the pivot. Challenge the strongest rival.
Only then choose the answer.