**Background:** Almost 14% of 2–5-year-old children in the US have obesity, with higher rates among Hispanic and African American children. Evidence-based prevention strategies exist but are often not adopted without broader policy and environmental changes. Cross-sector coalitions are promising but lack empirical evidence on how they effect change. The Stakeholder-driven Community Diffusion (SDCD) theory posits that multi-sector groups can diffuse knowledge and engagement through social networks to change policies, systems, and environments (PSE). This study prospectively tested the SDCD theory through the Shape Up Under 5 (SUU5) intervention.
**Methods:** SUU5 was a two-year study (2015–2017) in Somerville, MA, where childhood obesity prevalence among children under five was 14.4%. A 16-person coalition (SUU5 Committee) was formed, representing early childhood education, healthcare, parks and recreation, public health, and public schools. The Committee met 16 times over two years, facilitated by Tufts University researchers and Community-based System Dynamics (CBSD) experts. The intervention had three phases: reviewing evidence-based strategies, group model building (GMB) to develop shared understanding and prioritize actions, and stakeholder-led action. Knowledge and engagement were measured using the validated COMPACT SDCD survey at baseline and every six months (five timepoints total). Knowledge comprised 18 items across five domains; engagement comprised 25 items across five domains. Committee members nominated up to 20 professional contacts (first-degree alters) at each timepoint; these alters also completed the survey (n=193 total). PSE changes were assessed via an online survey at baseline, 12 months, and 24 months. A community communications campaign (Small Steps: Eat, Play, Sleep) was created by the Committee, with materials in English, Spanish, Portuguese, and Haitian Creole. Reach was estimated via logs and a survey of first-degree alters. Mixed effects regression models with random intercept and trend were used to analyze changes over time.
**Key Results:** The Committee was 88% female, 94% white, 6% Hispanic, with a mean age of 50 years and mean 20 years of experience. First-degree alters were 85% female, 89% white, 10% Hispanic, mean age 43. Overall knowledge among Committee members increased significantly over two years (β=0.027, SE=0.006, p=0.0002). Knowledge of resources increased significantly at every follow-up compared to baseline. Overall engagement increased significantly from baseline to 24 months (β=0.011, SE=0.005, p=0.03), with influence and power showing a significant increase (β=0.029, SE=0.011, p=0.01). Among first-degree alters, overall knowledge increased significantly (β=0.008, SE=0.002, p=0.001), particularly knowledge of the problem (p=0.0001) and resources (p=0.0003). Overall engagement did not increase among alters; influence and power significantly decreased (β=-0.013, SE=0.005, p=0.004), while trust increased (β=0.009, SE=0.003, p=0.004). PSE scores improved significantly from baseline to 24 months (mean difference 0.084, 95% CI: 0.028–0.140, p=0.003). Changes in Committee members' engagement were significantly associated with PSE score changes (correlation coefficient=0.56, p=0.037). The baseline network included 90 stakeholders and 131 ties; the end-of-study network included 217 stakeholders and 356 ties. Of 55 first-degree alters surveyed about campaign reach, 73% reported seeing Small Steps materials; posters (83%) and brochures (75%) were most frequently viewed.
**Clinical Implications:** This study provides prospective evidence that a structured, theory-driven community coalition intervention can increase knowledge and engagement about early childhood obesity prevention among coalition members and partially diffuse that knowledge through professional networks. The significant improvements in PSE and the association with engagement suggest that upstream interventions targeting knowledge and engagement may translate into midstream environmental changes. The community communications campaign reached a substantial proportion of network members across multiple settings. The findings support the use of cross-sector coalitions and systems thinking methods (e.g., group model building) as a strategy for childhood obesity prevention. Limitations include the lack of a comparison group, a predominantly white and female sample, and reliance on self-reported PSE data. Future studies should include comparison groups and more objective PSE measures.