Advancing Complexity Science in Healthcare Research: The Logic of Logic Models

Background Logic models are commonly used in evaluations in healthcare research to represent how interventions produce outcomes in simple diagrammatical form. Uses of logic models include: serving as a framework in evaluations, forging consensus among stakeholders about a proposed change and knowledge transfer. However, whether logic models can be used for describing complex interventions which adapt to context remains unclear. This paper advances the field by proposing a more dynamic approach to logic modelling that focuses less on causal pathways from inputs to outcomes and more on contextual factors which shape causal pathways. Methods Various logic model types were tested as part of an evaluation of a Patient Experience (PE) Toolkit intervention, designed to guide healthcare professionals through a facilitated process of reflecting and acting on patient experience data. Implemented through action research across 6 diverse hospital wards, the PE Toolkit intervention was found to bear the hallmarks of a highly complex intervention. In particular, the facilitation provided by the action researchers was crucial to its functioning, allowing the intervention to take on different form in the different settings. An initial logic model, developed iteratively on the basis of a large qualitative dataset containing multiple data sources, was rejected at the halfway point of the study as it failed to capture intervention dynamics. Other logic models used in healthcare research were assessed as to whether they could account for the complexity of the intervention and a typology of logic model types, including strengths and weaknesses, was created. Implementation models were then explored and the PHARIS model was identified as being able to successfully model interaction between interventions, facilitation and context. The Venn-diagram used in the PHARIS model was incorporated into the logic model for the PE Toolkit intervention, after which a set of guiding principles for advancing the field was formulated. Results Three dominant types of logic model exist in healthcare research, each of which fail to capture the dynamics of complex interventions. These models may have potentially negative consequences, such as promoting the wrong courses of action in particular settings. However, a fourth type is possible that specifically models how contextual factors shape causal processes. These can be created by drawing upon theories and frameworks of implementation and context, using diverse shapes and arrows to highlight dynamic relationships and contingencies, specifying at what level “moderators” exert an effect and incorporating diverse stakeholder views. Implications Logic models can be used to represent how complex interventions work but more dynamic models are required than is currently the case in healthcare research. A potential implication of this is for logic models to take on a different role in improvement projects in future. That is, less of a role for logic models forging consensus among stakeholders or as precise guides for practitioners and more of a role stimulating discussion about how interventions work across different settings