Host university: Newcastle University
PhD research project: Issues in the design and analysis of (non-randomised) stepped-wedge trials with small numbers of clusters
Project summary: Stepped wedge trials (SWTs) are cluster randomised trials where all clusters, such as hospitals, schools, or community centres, begin in a control phase before receiving the intervention in a randomised time period.
This creates a staircase-looking design for which the SWT gets its name. All of the clusters stay in the intervention once they receive it, meaning the effect of time must be considered as more clusters will have received the intervention by the end of the study than at the start – for example, the prevalence of what we are studying may decrease independently from the treatment over time.
My work concentrates on SWTs with a binary outcome as most of the methodology available so far assumes a normal distribution for the outcome, yet many trials utilising stepped wedge designs have a binary outcome. I will be exploring what the most efficient time to move clusters from control to intervention is, with the aim to maximise power and make these designs more easily accessible to researchers by minimising the number of participants needed in a study.
Further work will also investigate, and provide guidance for, the role of randomisation in a stepped wedge trial. The majority of the work of SWTs labours under the assumption of having clusters in randomised sequences of when they cross from control to intervention.
For clusters such as hospitals, these randomisations may not always be logistically feasible and where we may have to consider a nonrandomised variation of a SWT, the literature to guide the design and analysis of such designs is scant.
Having the methodology to design robust SWTs is important to assess the effect of public health intervention which must be rolled out to whole communities to provide evidence of the cost effectiveness of these interventions.
Expected PhD completion date: Sept 2024
Contact: [email protected]