Independent binary effect-measure audit
Method review date: 2026-11-03

Instructions: Save a copy, complete blanks and record unresolved questions. Use fictional or aggregate values only; never patient identifiers. This planning aid does not certify an analysis.

DEFINE GROUPS AND ORIENTATION
Group1: ____; group2(reference): ____
Event, time window and population: ____
Independent groups? ____; matched/clustered/case-control selection? ____
Chosen estimand and reason: ____

EXTRACT COMPLETE TABLE
Group1 events a: ____; non-events b: ____; total a+b: ____
Group2 events c: ____; non-events d: ____; total c+d: ____
Missing outcomes/exclusions by group: ____
Source locations and coding checks: ____

KEEP MEASURES DISTINCT
p1=a/(a+b): ____; p2=c/(c+d): ____
Risk difference p1−p2: ____ (multiply 100 for percentage points)
Odds ratio ad/bc: ____ (not a risk ratio)
OR log-SE=sqrt(1/a+1/b+1/c+1/d): ____
OR 95% limits=exp(log(OR)±1.959964*SE): ____
This OR approximation requires every cell>0; do not silently add 0.5.

DOCUMENT INTERVAL CHOICE
For RD, linked tool uses independent Wilson intervals L1,U1,L2,U2.
Lower=RD−sqrt((p1−L1)²+(U2−p2)²): ____
Upper=RD+sqrt((U1−p1)²+(p2−L2)²): ____
Method: Newcombe hybrid-score method 10, no continuity correction.
Independent check, sparse-data limitations, bias/confounding review: ____
For case-control sampling, do not interpret sampled proportions as population risks.

Practical guide: https://www.statmedanalysis.com/blog/risk-difference-newcombe
Project/version: ____  Completed by role/date: ____