the knowledge platform

ebm stats in 30 minutes: lr, nnt, and odds ratios

a clinician-friendly stats cheatsheet for exams: likelihood ratios, nnt/arr, and how to avoid odds-ratio traps.

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