Pre-viva seminar: Robust Analysis of Stepped Wedge Trials
Date: Tuesday 28th November
Venue: Room LG9, Keppel Street
Speaker: Jennifer Thompson
Supervisors: Katherine Fielding and James Hargreaves
Stepped-wedge cluster-randomised trials are a trial design where clusters are randomised to sequences. Each sequence switches from the control to the intervention condition at a different time during the trial. This means that at the start of the trial, all or most of the clusters are receiving a control condition, but by the end of the trial all or most of the cluster are receiving the intervention. The analysis of these trials are complex; thought must be given to changes in the outcome over time as well as to the correlation of observations over time. The most commonly used analysis methods make strong and often unrealistic correlation assumption. In this talk I will describe the impact of using different analysis methods in this setting. When an analysis assumes an incorrect correlation structure, the standard errors are biased leading to confidence intervals that are too narrow. To overcome this problem, I will describe a novel analysis method that makes fewer assumptions, and I will demonstrate use of the method with an example.