Research methods guide
Retrospective cohort protocol template
A retrospective cohort protocol starts with the source population, index date, exposure, comparator, and follow-up. Then specify outcomes, confounding, censoring, and analysis.
Compare two outpatient strategies in a synthetic cohort. Align eligibility, assignment and follow-up to avoid crediting pre-exposure event-free time to the exposed group.
Retrospective cohort protocol skeleton
Replace examples with auditable definitions in your own data.
| Item | Worked example | Verification |
|---|---|---|
| Entry and time zero | Include eligible new users with no use in the preceding 180 days. Index at first qualifying initiation of either strategy. | Verify comparable eligibility, start dates and data coverage. |
| Outcome and follow-up | Unplanned admission within 90 days; retain death and exit separately. | Prespecify the algorithm, administrative cutoff, competing events and censoring. |
| Bias and analysis | Consider pre-initiation severity, age and prior utilization; report risk difference with interval. | Prespecify overlap, missingness and alternative definitions; avoid treating mediators as baseline confounders. |
Apply the method to your materials
Run a feasibility table first
Check populations, observable windows, key missingness, duplicate outcome codes and site coverage dates before estimating effects.
Freeze protocol and log deviations
Version definitions, primary and sensitivity analyses, and document data-driven changes to comparators and their implications.
Questions that arise in practice
Can the first completed follow-up be treatment start?
If eligibility begins earlier, account for exposure classification and observation in the intervening period before selecting an analysis.
Verified sources
The official sources below were checked on 2026-11-12. Access conditions and update dates are rechecked before publication.
RigorMed workspace
Start with your research materials
Bring your research question, available materials, and the steps you want to check.
