What is complexity and what do you do with it? Reflections on use of controlled trials to assess complex public health interventions

Abstract

Listen

Translate article

BackgroundDebate about the appropriate means to assess the effects of complex public health interventions is very much ongoing, particularly the feasibility and value of randomised controlled trials. There has been little systematic investigation of complexity in studies of public health interventions, but an approach to evaluation that explicitly investigates the signs and sources of complexity in an intervention has the potential to increase understanding of why an intervention works or fails. We review the debate about how randomised controlled trials can accommodate complex interventions, how complexity can be investigated within the controlled trial framework, and what the added benefits of such an approach might be. We use the Well London cluster randomised trial as a case study. Case studyLocal adaptation is a commonly cited feature of complexity, as are complicated intervention delivery logistics and interactions between multiple intervention components. A key unresolved objection to the use of controlled trials for complex intervention evaluation is the reductionist conceptualisation of the intervention. The effectiveness component of the Well London trial necessarily treats this multiple-component, locally adapted, community engagement for health programme as simple, to allow valid causal inference. Although the trial includes quantitative community-level outcomes and examines interaction between the intervention and context, this analysis is in a subgroup model. Only the qualitative study nested within the trial directly addresses signs of complexity (synergies between intervention components, phase changes, unexpected outcomes). We propose that the notion of an intervention as simple or complex is dictated by the research question. From this view, although the effectiveness component of a trial treats the intervention as simple, and as such could be seen as overly reductionist, it is not theoretically inconsistent for an assessment to include some additional components investigating aspects of complexity. Additional methods could be easily accommodated, both quantitative and qualitative, including collection of outcome and exposure data continuously throughout the intervention period from a subcohort of study participants; diaries and other ongoing reflective documentation by community members and practitioners delivering the intervention; and social network analysis. Logic models play a central part in identification of potential unexpected outcomes and sources of complexity for investigation. However, methods for logic model development must accommodate non-linear aspects of community engagement interventions to be useful. InterpretationThe benefit of a focus on complexity in the controlled trial framework is to provide rich data about how the intervention interacts with context to assess transferability and explain the findings of the effectiveness component, whether they are positive, null, or negative. The knowledge produced will expand the comparatively sparse basic science evidence base underpinning upstream, preventive public health interventions, which do not receive developmental funding akin to pharmaceutical interventions, to understand in detail how the intervention takes effect before an effectiveness trial. Complexity is important to consider in the process of development of research questions with relevant stakeholders. Consideration of complexity at that stage provides a less reductionist way of thinking about public health interventions and the social systems within which they exist, motivating selection of different research questions and, importantly, different outcomes for the study. FundingWellcome Trust (grant 083679/Z/07/Z).