Purpose: Addressing complex problems such as obesity requires engagement of actors and organizations across multiple levels of action. Whole-of-community interventions – those that are multilevel, multicomponent and implemented through multiple sectors and settings of an entire community – have been recommended and show promise for prevention. However, less is known about mechanisms underlying their success or failure. A growing call to integrate systems science into implementation science hopes to address these gaps—successful implementation is more likely to be effective when efforts are informed by both knowledge of the intervention levers likely to have the most impact and appreciation of the system in which the intervention is implemented. The utilization of tools to help understand complex systems has been gaining traction within implementation science as a means to identify potential predictors of implementation success, such as stakeholders’ social networks. However, most applications to date employ only a single complex systems method.  Approach:  We illustrate the retrospective qualitative development of a systems map representing community change dynamic within the Shape Up Somerville intervention.  We then describe how this systems map, and complementary work of other successful interventions (Romp & Chomp intervention) informed the COMPACT (childhood obesity modeling for prevention and community transformation) study. COMPACT’s design aligns complex systems science principles and community-engaged research to better understand stakeholders’ leadership roles in whole-of-community interventions. We provide an overview of the complex systems tools used in COMPACT: agent-based modeling (ABM), group model building (GMB) and social network analysis (SNA), and describe how whole-of-community intervention stakeholders (“agents”) use their social networks to diffuse knowledge about and engagement with childhood obesity prevention efforts, laying the groundwork for community readiness for sustainable change.  Conclusion: Complex systems approaches appear feasible and useful to study whole-of-community obesity prevention interventions and provide novel insights that expand upon those gained from traditional approaches. The specific multi-method research process described—qualitative domain-specific systems mapping, theory-building by an interdisciplinary team informed by multiple fields of science, and quantitative tools like SNA and ABM—may provide a roadmap for future applications in implementation science.

Purpose: Successful whole‐of‐community childhood obesity prevention interventions tend to involve community stakeholders in spreading knowledge about and engagement with obesity prevention efforts through the community. We refer to this process as stakeholder‐driven community diffusion (SDCD). We will describe how we have used agent-based modeling (ABM) in conjunction with data collected from interventions deployed in multiple sites to understand how SDCD operates and under what conditions it has the potential to effect substantial, sustainable changes in communities. Methods: In collaboration with intervention experts, we designed an ABM that simulates the SDCD process in communities of socially connected individuals. We have used this model to retrospectively simulate multiple interventions in communities in the United States and Australia. We utilize stakeholder survey data, intervention records, and expert estimates to accurately represent each intervention that we simulate. We then compared model output to real-world observations to test hypotheses about SDCD. Results: Across settings, we find that increases in knowledge about and engagement with childhood obesity prevention interventions produced by our simulation model could match those observed in the real world. This was neither a “knife edge” result produced only under a small set of parameter combinations nor a “hard-coded” outcome that occurred regardless of model parameters. Thus, our findings allow us to make statements about the role of SDCD in driving intervention outcomes as well as how SDCD operates. Conclusions: We found strong suggestive evidence in support of a hypothesis that SDCD was a key driver of the success of multiple interventions. Model exploration also provided additional insights about salient aspects of how SDCD operates across settings. We will discuss the important implications that this has for the design and implementation of future interventions, as well as what additional data collection in conjunction with those interventions might prove most beneficial to gaining a better understanding of how communities can effectively address childhood obesity.

Purpose: In this presentation we will present examples of the application of systems modeling techniques in intervention design, describe how these fit in the trial design of interventions across large regional areas, and pose questions and challenges for evaluation and interpretation of results using systems informed process and outcome data. Method: The Global Obesity Centre (GLOBE) along with community partners have multiple active trials of childhood obesity prevention in 24 Victorian communities reaching more than 45,000 children. These trials have a stepped wedge cluster randomized design, supporting a traditional intervention measurement approach to evaluation. The intervention design, implementation and process evaluation adapt and apply techniques arising from the COMPACT collaboration to test their efficacy and capacity building impacts.  System dynamics (causal loop diagrams and simulation modelling), social network analysis and agent-based modelling have been applied in practical implementation of community-based obesity prevention efforts. Partners to these interventions include departments of health and education, health services, civic leaders in government and representatives from retail, education, sport and recreation sectors. Results: Promising initial changes have been observed in obesity rates alongside the trial and in light of higher than expected prevalence figures. Longer term behaviors changes have been observed. Communities have been able to develop causal loop diagrams, social network analysis has been used to inform trials and understand relative contribution to implementation, and knowledge and engagement information have been collected to further inform computational modelling of diffusion. Conclusions: We found applying the principles underpinning complex systems modeling as well as formal system modelling methods congruous with community-based interventions. Participating communities demonstrated variance in their uptake of these methods and in the ability to build capacity to apply them. Practical application and implementation of these key methods for addressing the challenges posed by the complexity of obesity remains a critical area for future research.