Practical issues that impact statistical design, analysis and synthesis of cluster randomised controlled trials

Student thesis: Phd

Abstract

When conducting randomised controlled trials assessing the effect of introducing a new health service or policy, researchers face challenges relating to organisation, implementation and contamination. Cluster randomised trials, where participants are randomised in groups, offer a solution to these challenges. In this PhD thesis I present seven published research articles as evidence of my contribution as an applied statistician to the design, analysis and synthesis of cluster randomised trials. I present two papers from the Organising Support for Carers of Stroke Survivors (OSCARSS) trial to demonstrate the methods I developed to minimise bias in in the design and analysis of a cluster randomised trial of a complex intervention. I present a Cochrane systematic review to display innovative methods that I developed to incorporate trials with non-standard cluster designs, such as cluster crossover trials, into systematic reviews. SOCIAL was a large systematic review of social norms interventions to change healthcare professional behaviour. I present 4 papers relating to this review to illustrate how I overcame a number of challenges which included synthesis of multiple complex interventions, a mixture of outcome measurements, and a variety of cluster designs. I have shown that cluster randomised trials reveal a number of challenges in their design, analysis and evidence synthesis, over and above those of individually randomised trials. I have offered practical methods to deal with these challenges and critically appraised their merits and weaknesses.
Date of Award31 Dec 2022
Original languageEnglish
Awarding Institution
  • The University of Manchester
SupervisorSarah Cotterill (Supervisor)

Keywords

  • cluster trials
  • systematic review
  • biostatistics

Cite this

'