Most exam stats are not complicated — they’re just unfamiliar under time pressure. The goal is a tiny set of automatic moves: when to use likelihood ratios, how to convert ARR to NNT, and how to interpret odds ratios without getting tricked by intuition.
The minimum set to automate
LR+ and LR- (diagnostics), ARR → NNT (treatment effect), and odds ratio interpretation (association). If you can do these fast, you unlock a lot of marks.
1
Move 1 — Use likelihood ratios to update probability
Know what they’re for: translating a test result into post-test probability. LR is often more stable across populations than predictive values, which shift with prevalence.
2
Move 2 — NNT is just 1 / ARR
If ARR is 0.10 (10%), NNT is 10. If ARR is 0.02 (2%), NNT is 50. Practise 3–5 quick examples until it’s automatic.
3
Move 3 — Odds ratios aren’t risk ratios
ORs can overstate effect size when outcomes are common. Don’t over-interpret magnitude; focus on direction, confidence interval, and context.
4
Move 4 — Use a 60-second stats drill
Once daily: one LR question, one NNT question, one OR interpretation. Small reps beat long painful cramming.
Common failure mode
Memorising definitions but never doing timed examples. Stats is a skill: you need retrieval under pressure.
SourceOxford CEBM — Likelihood ratios (tool + explanation)
Open Link SourceJAMAevidence — Numbers Needed to Treat (NNT = 1/ARR)
Open Link SourceJAMA Guide — Correct use and interpretation of odds ratios
Open Link Save your place and choose what comes next
Work through this guide (4 steps)
Use the actions from this article to track what you have worked through.
Sign in using the reading controls above to save your steps.
Know what they’re for: translating a test result into post-test probability. LR is often more stable across populations than predictive values, which shift with prevalence.
Return to this sectionIf ARR is 0.10 (10%), NNT is 10. If ARR is 0.02 (2%), NNT is 50. Practise 3–5 quick examples until it’s automatic.
Return to this sectionORs can overstate effect size when outcomes are common. Don’t over-interpret magnitude; focus on direction, confidence interval, and context.
Return to this sectionOnce daily: one LR question, one NNT question, one OR interpretation. Small reps beat long painful cramming.
Return to this section
Study methods · Article 20 of 48