the knowledge platform

akt evidence & data interpretation: the 2026 playbook

a ruthless method for akt charts, confidence intervals, risk tools and ebm traps—so you stop losing easy marks on “data questions”.

The Bottom Line

  • AKT EBM/data questions are often <strong>method questions</strong>, not medicine questions: your job is to interpret what’s shown (denominator, axis, rate vs count).
  • Use a <strong>repeatable 6-step scan</strong> for every chart/table before you even look at the answer options.
  • Most lost marks come from: <strong>rates vs totals</strong>, <strong>relative vs absolute risk</strong>, and misreading <strong>confidence intervals</strong>.
In the AKT, evidence-based practice questions are designed to test your interpretation of data presentations (graphs, tables, audit-style summaries) and your ability to reason about risk and uncertainty. The fastest improvement comes from using a fixed method—like a pilot checklist—rather than “hoping the graph makes sense”.

The mental model

Treat every EBM/data question as: <strong>What is being measured?</strong> → <strong>In whom?</strong> → <strong>Over what time?</strong> → <strong>Compared to what?</strong> → <strong>How certain is the estimate?</strong>
1

Step 1 — Identify the unit of meaning (10 seconds)

Before reading options, state the unit out loud: <strong>rate per 1,000</strong>, <strong>percentage</strong>, <strong>incidence</strong>, <strong>odds ratio</strong>, <strong>hazard ratio</strong>, <strong>absolute number</strong>. Many “trick” questions are just denominator traps.
2

Step 2 — Read the axes like a lawyer (15 seconds)

Confirm: axis labels, time window, population, whether the plot is per 1,000 / per 100,000, whether it is <strong>log scale</strong>, whether there’s a second axis, and whether the plotted values are <strong>adjusted</strong> or raw.
3

Step 3 — Convert the figure into one sentence (10 seconds)

Write (mentally): “In population X, outcome Y is higher/lower in group A vs B over time T.” If you cannot summarise in one sentence, you are not ready to answer.
4

Step 4 — Separate association from causation (10 seconds)

If it’s observational: default to <strong>association</strong>, consider confounding and bias. If it’s RCT/meta-analysis: consider randomisation, effect size, and certainty.
5

Step 5 — Use the CI rule properly (10 seconds)

For ratios (RR/OR/HR): CI crossing <strong>1</strong> → not statistically significant at that level. For differences (mean difference): CI crossing <strong>0</strong> → not statistically significant. Then ask: even if significant, is it clinically meaningful?
6

Step 6 — Translate relative risk into absolute impact (30 seconds)

If you can estimate baseline risk, convert to <strong>absolute risk reduction</strong> and then <strong>NNT</strong> (NNT ≈ 1 / ARR). The AKT likes testing whether you can spot when a “big relative risk” is actually small absolute benefit.

The AKT Data Question “Trap List” (memorise this)

1

2

3

4

5

6

What to do when you feel unsure

Do not “guess from vibes”. Apply the method: unit → axes → one-sentence summary → comparison → uncertainty. Your score improves because your process is stable, not because you learn more facts.
1

The 20-minute weekly drill (high ROI)

Once weekly, do 10 AKT-style data questions timed. For each wrong answer, write a <strong>one-line error label</strong> (e.g., “denominator trap”, “CI crossing 1”, “association/cause confusion”). After 4 weeks, your pattern will be obvious—and fixable.
2

Build a micro-cheat-sheet (one page only)

Your one page should contain: CI rule, RR→ARR→NNT conversion, sensitivity vs PPV, and 5 most common graph traps. If it becomes 3 pages, you have lost the point.
SourceRCGP: AKT Example Questions (with answers) — Nov 2025 (official PDF)
Open Link
SourceRCGP: Evidence & Data Interpretation in the AKT (official PDF)
Open Link
SourceRCGP: Preparing for the AKT (official page)
Open Link