Context: In the evaluation of the Healthy Kids, Healthy Communities initiative, investigators implemented Group Model Building (GMB) to promote systems thinking at the community level. As part of the GMB sessions held in each community partnership, participants created behavior-over-time graphs (BOTGs) to characterize their perceptions of changes over time related to policies, environments, collaborations, and social determinants in their community related to healthy eating, active living, and childhood obesity. Objective: To describe the process of coding BOTGs and their trends. Design: Descriptive study of trends among BOTGs from 11 domains (eg, active living environments, social determinants of health, funding) and relevant categories and subcategories based on the graphed variables. In addition, BOTGs were distinguished by whether the variables were positively (eg, access to healthy foods) or negatively (eg, screen time) associated with health. Setting: The GMB sessions were held in 49 community partnerships across the United States. Participants: Participants in the GMB sessions (n = 590; n = 5-21 per session) included key individuals engaged in or impacted by the policy, system, or environmental changes occurring in the community. Main Outcome Measures: Thirty codes were developed to describe the direction (increasing, decreasing, stable) and shape (linear, reinforcing, balancing, or oscillating) of trends from 1660 graphs. Results: The patterns of trends varied by domain. For example, among variables positively associated with health, the prevalence of reinforcing increasing trends was highest for active living and healthy eating environments (37.4% and 29.3%, respectively), partnership and community capacity (38.8%), and policies (30.2%). Examination of trends of specific variables suggested both convergence (eg, for cost of healthy foods) and divergence (eg, for farmers' markets) of trends across partnerships. Conclusions: Behavior-over-time graphs provide a unique data source for understanding community-level trends and, when combined with causal maps and computer modeling, can yield insights about prevention strategies to address childhood obesity.
