AtlasSession 3L9: Variations in study designp.21–24, 42–43
Cluster and stepped-wedge
- 3 earlier topics
- Types of distribution
- Normal vs non-normal
- Choosing a statistical test
- Parallel and multi-arm trials
- Cluster and stepped-wedge
Fig. 1
Burns units: hydrogel dressing protocol vs standard dressings, wound infection at 21 days (patients per unit)
Fig. 1. 18 burns units of about 20 patients; whole burns units move as blocks: 9 to standard dressings and 9 to hydrogel protocol.
- Burns units
- 18 (9 + 9)
- Patients
- 360
- What is randomised
- whole burns units
Trial Design Board · First look
Level 1 · First look · Lecture 9 · p.21–24, 42–43
See why some trials randomise whole groups, and what a staged rollout is.
Explain why randomising groups inflates the sample size by a design effect and needs a clustering-aware analysis, and when a stepped-wedge rollout fits.
Schools, not pupils (the step you are on)
A new lunch menu goes to whole schools, and pupils in one school tend to be alike.
A new lunch menu goes to whole schools, not single pupils, and pupils in one school tend to be alike. A dressing protocol goes to whole burns units the same way. Burns units: 18 (9 + 9); patients: 360.