Explaining high and low performers in complex intervention trials: A new model based on diffusion of innovations theory

© McMullen et al. Background: Complex intervention trials may require health care organisations to implement new service models. In a recent cluster randomised controlled trial, some participating organisations achieved high recruitment, whereas others found it difficult to assimilate the intervention and were low recruiters. We sought to explain this variation and develop a model to inform organisational participation in future complex intervention trials. Methods: The trial included 40 general practices in a London borough with high HIV prevalence. The intervention was offering a rapid HIV test as part of the New Patient Health Check. The primary outcome was mean CD4 cell count at diagnosis. The process evaluation consisted of several hundred hours of ethnographic observation, 21 semi-structured interviews and analysis of routine documents (e.g., patient leaflets, clinical protocols) and trial documents (e.g., inclusion criteria, recruitment statistics). Qualitative data were analysed thematically using-and, where necessary, extending-Greenhalgh et al.'s model of diffusion of innovations. Narrative synthesis was used to prepare case studies of four practices representing maximum variety in clinicians' interest in HIV (assessed by level of serological testing prior to the trial) and performance in the trial (high vs. low recruiters). Results: High-recruiting practices were, in general though not invariably, also innovative practices. They were characterised by strong leadership, good managerial relations, readiness for change, a culture of staff training and available staff time ('slack resources'). Their front-line staff believed that patients might benefit from the rapid HIV test ('relative advantage'), were emotionally comfortable administering it ('compatibility'), skilled in performing it ('task issues') and made creative adaptations to embed the test in local working practices ('reinvention'). Early experience of a positive HIV test ('observability') appeared to reinforce staff commitment to recruiting more participants. Low-performing practices typically had less good managerial relations, significant resource constraints, staff discomfort with the test and no positive results early in the trial. Conclusions: An adaptation of the diffusion of innovations model was an effective analytical tool for retrospectively explaining high and low-performing practices in a complex intervention research trial. Whether the model will work prospectively to predict performance (and hence shape the design of future trials) is unknown.