Use of cluster randomised trials in implementation research
Background. Implementation research is the scientific study of methods to promote the uptake of research findings in clinical practice. Cluster randomised trials are commonly adopted in implementation research, where interventions are generally targeted at health professionals or policymakers, in order to protect against contamination that would occur if individual patients were randomised. The application of cluster randomisation has important implications for design and analysis of trials evaluating implementation strategies. Case study: The Urological referral guidelines evaluation (URGE) has been used throughout this thesis, to explore the design and analysis issues of adopting a cluster randomised trial design in implementation research. URGE aimed to evaluate the effectiveness and efficiency of a guideline-based open access urological investigation service. This cluster randomised study adopted a 2X2 balanced incomplete block (BIB) design and collected data both prior to and following introduction of the intervention. The unit of randomisation was general practice and patients were recruited upon referral to secondary care. Aim: To investigate the implications of cluster randomisation for the design and analysis of trials evaluating implementation strategies. Objectives: This thesis has four distinct components. 1. A review of published cluster randomised trials in the field of implementation research. The methodological quality of these studies is assessed (Chapter 2). 2. An exploration of clustering within the URGE trial. Estimates of clustering and the imprecision in these estimates are considered for a number of endpoints, including process and outcome of care indicators and costs (Chapters 4 and 7). 3. The application of statistical methods in the analysis of cluster randomised trials. A number of approaches to the analysis of cluster randomised trials are described, applied and compared empirically. Incorporation of the BIB design and pre-intervention performance are also considered (Chapters 5 and 6). 4. Analysis of cost data collected from the economic evaluation conducted within the URGE trial. The analysis of skewed cost data collected within a cluster randomised trial design is considered (Chapter 7).