Missing binary outcome sensitivity bounds
StatMedAnalysis | Verified 2026-11-04

Instructions: Save a copy, replace blanks, and retain the completed version with your analysis plan. Use synthetic examples or approved aggregate counts only; do not add patient identifiers. Record unresolved decisions rather than inventing answers. This worksheet documents decisions and does not provide clinical advice.

Project/version: __________ Date: __________ Prepared by: __________

1. DEFINE THE ENDPOINT AND POPULATION
Binary event and fixed assessment window: ____________________
Group definitions and intended denominator: ____________________
Which unresolved outcomes can legitimately be either event or non-event: ____
Death, competing events or unequal follow-up needing separate handling: ____

2. RECONCILE COUNTS
Group | observed events e | observed outcomes o | missing m | total o+m
1     | _________________ | ___________________ | _________ | _________
0     | _________________ | ___________________ | _________ | _________
Check 0<=e<=o; total must be positive. Include observed non-events in o.

3. CALCULATE BOUNDS
Group 1 minimum e1/N1: ____ maximum (e1+m1)/N1: ____
Group 0 minimum e0/N0: ____ maximum (e0+m0)/N0: ____
Lowest risk difference=min1-max0: ____
Highest risk difference=max1-min0: ____
Observed-outcome-only difference (if both observed denominators positive): ____
These are sensitivity bounds, not confidence intervals and not causal estimates.

4. INTERPRET AND FOLLOW UP
Does the range cross zero or a prespecified meaningful threshold? _______
Plausible narrower assumptions and their evidence: ____________________
Retrieval actions, owner and deadline: ____________________
Sampling uncertainty and other biases still unaddressed: ______________

Final review
Arithmetic checked by/process: __________ Date: __________
Unresolved limitations and next action: ____________________
Guide: https://www.statmedanalysis.com/blog/missing-binary-outcome-bounds