Abstract
Objective:
Whole-of-community interventions are a promising systems-based approach to childhood obesity prevention. A theorized driver of success is “Stakeholder-Driven Community Diffusion” (SDCD): the spread of knowledge about and engagement with obesity prevention efforts from a committee of stakeholder representatives. We focus on the potential of SDCD to affect the broader community.
Methods:
We use an agent-based model of SDCD to dynamically represent the interpersonal interactions that drive community diffusion of knowledge and engagement. We test its explanatory power using longitudinal data from a sample of community members and then use simulations to extrapolate from this limited sample to the unobserved community at large. We also consider counterfactual scenarios that show how changes in implementation strategy might have led to different patterns of community change.
Results:
Our model can reproduce real-world patterns of diffusion. Simulations show a substantial increase in knowledge (an approximate doubling) and a slight increase in engagement throughout the broader community. A relatively small amount of this change in knowledge (∼10%), and all the change in engagement is attributable to direct intervention effects on committee members.
Conclusions:
SDCD is premised on creating preconditions for sustainable change. Previous work has estimated impact on small samples closely linked to the stakeholder committee, but the degree to which this translates into the much broader diffusion envisioned by SDCD theory is unknown. This analysis demonstrates the potential of interventions to do just that. Additionally, the counterfactual scenarios suggest that simulation can help tailor implementation of SDCD interventions to increase impact.
Keywords: agent-based modeling, childhood obesity prevention, social network, Stakeholder-Driven Community Diffusion, whole-of-community
Introduction
The obesity epidemic remains a serious challenge for American society. Childhood obesity is of particular concern not only because of its implications for immediate health outcomes but also because of its association with long-term health effects, including an increased likelihood of obesity in adulthood. Recently, it has been estimated that 14.4 million children and adolescents have obesity, including 20% of children aged 6 to 11 years and 13% of children aged 2 to 5 years; underlying health disparities are reflected in prevalence being highest among Black and Hispanic children.1
An especially promising approach to meaningful and sustainable reduction of childhood obesity is whole-of-community intervention that takes a broad systems-based perspective, strategically guiding changes to policies and practices across community settings.2–5 In recent years, evidence has emerged showing that a number of whole-of-community interventions have tangibly reduced child overweight and obesity.6–14 Common features of successful whole-of-community childhood obesity prevention interventions include “buy-in” from and involvement of community stakeholders; leveraging and aggregating existing resources and networks; a focus on inclusive community engagement; synergistic design and implementation of intervention actions across multiple community sectors; and recognition of and adaptation to local needs and conditions.3,5,15,16 A recent review found that these features—and through them, positive impact—can be facilitated by enlisting and regularly convening a “steering committee” of stakeholders from key community sectors.16
In this and previous work, we refer to the hypothesized mechanism through which steering committees drive intervention success as “Stakeholder-Driven Community Diffusion” (SDCD).17,18 It comprises a multistage process through which a relatively small group of community stakeholders can catalyze changes in the broader community to facilitate shifts in policy, practice, and environment and ultimately have positive effects on child health. Members of the steering committee work to purposively cultivate their own understanding about challenges their community faces in reducing child obesity and overweight; insights into how these can be addressed; and capacity and enthusiasm for doing so.
We broadly refer to these efforts as building two multidimensional constructs referred to as “knowledge” and “engagement.” As this happens, steering committee members spread (i.e., “diffuse”) this knowledge and engagement through their direct professional and social networks and then potentially through the much larger set of existing networks in the rest of the community. This is an important goal because previous work has argued that sufficient levels of knowledge about and engagement with child obesity prevention in the larger community are an important prerequisite for sustainable and context-appropriate “midstream” changes in policies and practices in settings such as child care, health care, and local government.3,19–21
A nascent body of evidence has emerged supporting the importance of these domains as measured here,19 and potential for translation into downstream impacts on behaviors and health outcomes.20,21 However, direct large-scale measurement of change in these domains across entire communities is challenging and has not yet been conducted, so the full impact of the SDCD approach (beyond the local steering committee) is not yet known.
To address this gap, we build on previous research using a simulation model of SDCD during a successful whole-of-community childhood obesity prevention intervention.18 Previous work used retrospective evidence from a completed intervention supporting the presence and impact of the SDCD mechanism. Here, we contemporaneously studied an active intervention, using the model to make testable predictions, and then to extrapolate insights beyond where direct measurement was possible. The intervention studied here was designed with SDCD at its core, and data collection was informed by relevant prior research and collaboratively planned a priori by a team of intervention experts and complex systems scientists.17,18
The new analysis presented here provides additional insight into the SDCD process in two ways. First, testing the model's explanatory power using longitudinal data from a sample of community members allows us to refine our understanding of the hypothesized core causal mechanism. Second, as longitudinal measurement of entire communities is not currently practical, the use of modeling and simulation adds important insight into the broader implications of changes observed in the small sample directly measured. To assist in further practical refinement of the SDCD approach, we also consider counterfactual scenarios that consider how changes in implementation strategy might have led to different patterns of community change.
Methods
Shape Up Under 5: Setting and Data
Shape Up Under 5 was a 2-year early childhood obesity prevention pilot study conducted from 2015 to 2017 in Somerville, Massachusetts.17 A 16-member stakeholder committee was recruited and convened to address early childhood obesity prevention representing a variety of community settings and sectors. The Committee met 16 times over the course of the study and used group model building, a participatory research method, to (1) increase their knowledge and engagement around early childhood obesity prevention; (2) shift their mental models toward a systems perspective on challenges and solutions; (3) design a community-wide, culturally appropriate messaging campaign (Small Steps: Eat Play Sleep) about the importance of healthy growth in children 0–5 years; and (4) intentionally grow members' social networks, leveraging these to implement policy, practice, and environmental changes throughout the community.22,23
Data were collected using a web-based survey conducted semiannually from Committee members and their first-degree alters (i.e., those who they identify as having direct contact related to issues of childhood obesity in their community) using an SDCD survey to measure social networks, knowledge, and engagement.23,24 Knowledge was characterized as a vector of five “domains” representing salient areas of expertise identified during previous research: obesity as a problem; intervention factors and modifiable determinants; the roles of stakeholders and others; available resources; and sustainability of efforts.19
Engagement describes an individual's enthusiasm and agency for preventing childhood obesity in their community.19,24 In addition, three semi-structured qualitative interviews were conducted annually with each Committee member. An annual survey to measure policies, practices, and environments related to early childhood obesity prevention efforts in the community was distributed to Committee members. All study procedures and activities were approved by the institutional review board at Tufts University.
Agent-Based Model Design
We built upon an existing agent-based model (ABM) of SDCD that we had previously developed and tested using retrospective data.18 ABM is a computational simulation approach well suited to our research goals as it can capture individual-level interactions occurring in a social network framework in situations such as this one, where aggregate outcomes may be highly dependent on who interacts with whom and when. Because individuals' characteristics and interactions with one another can all be explicitly represented over time, we can effectively capture the SDCD process beyond the stakeholder committee and into the whole community (Fig. 1).
Figure 1.
Conceptual framework. The top panel depicts our theoretical model of how inputs (in boxes) flow through processes (in triangles) and also shown in greater detail in lower panels (A–C) to affect dynamic variables (in ellipses). Elements to the left of the dashed line (with the exception of exogenous influences) are represented in our computational model. Individual committee members along with other stakeholders in the community are shown as circles in the lower panels, with border colors representing community group sector affiliation, fill shading representing knowledge, and size representing engagement; social network connections are lines connecting circles. Each individual is labeled consistently across the lower panels, with red labels indicating SC membership. Thus, lower panel (A) shows members of the community (X, Y, and Z) joining the SC; panel (B) shows those individuals gaining knowledge, engagement, and connections to one another while attending a meeting; and panel (C) shows diffusion of knowledge and engagement out from the SC, with resulting changes experienced by other community members explicitly represented. SC, Stakeholder Committee. Color image is available online.
We ground our simulations in analogous, prospectively collected real-world data, compare simulated trends with those observed in data, and then extrapolate beyond available data. Empirically observed changes in the community were limited by practical and ethical considerations to the Committee members and first-degree alters23—these included time and understanding required to complete the survey, incentive costs for participants, and willingness of participants to provide contact information for further sampling. However, changes in the much larger overall community are considered essential to effective obesity prevention. Here, we present an analysis that bridges this gap by calibrating the ABM with observed data from the intervention and then using it to extrapolate to the (unobserved) community at large. We summarize our model and its use of data below, and fully describe these in Supplementary Appendix.
Model Initialization
Simulated “agents” in our model each represent all community stakeholders (i.e., Committee members and others). They have initial knowledge and engagement values, positions within the community social network, and affiliations with community sectors that have the potential to directly or indirectly affect childhood obesity prevention (e.g., child care providers or parents). The model parameter values used for initialization are derived from data sources listed in Table 1; the processes used to derive specific values as well as the values themselves are detailed in Supplementary Appendix.
Table 1.
Description of Model Parameterization and Collected Output
| Category | Model input parameter or expected model behavior | Sources of data |
|---|---|---|
| Model inputs | ||
| Knowledge and engagement (initial) | Group knowledge (mean) | Survey data (first round responses completed by identified alters); for 2 of 11 groups supplemented with expert estimates due to small sample size relative to group size |
| Group knowledge (standard deviation) | ||
| Group engagement (mean) | ||
| Group engagement (standard deviation) | ||
| Committee member knowledge | Survey data (first round, Stakeholder Committee) | |
| Committee member engagement | Survey data (first round, Stakeholder Committee) | |
| Social network structure | Group membership | Survey data (first round responses from identified alters) and intervention records |
| Group connectivity (average number of links per agent) and associativity table (pairwise probability of connections between the groups) | Combination of survey data (first round) and expert estimates | |
| Committee member connectivity | Survey data (first round) | |
| Shape Up Under 5 Committee Meetings | Committee composition | Survey data (first round) and intervention records |
| Effects on knowledge, engagement, and links between Committee members | Survey data (longitudinal, five rounds) | |
| Model outputs | ||
| Knowledge and engagement (end of intervention) | Community knowledge (mean) | Average of group data taken from model runs |
| Community engagement (mean) | Average of group data taken from model runs | |
A full description of parameter values and the parameterization process appears in Supplementary Appendix.
Model Dynamics
During simulation runs, agents interact with their social network contacts, affecting each other's knowledge and engagement over time; in aggregate, this captures the “diffusion” aspect of SDCD throughout the community. We also explicitly simulate the core intervention element: a subset of agents who are members of the Committee. These individuals attend regularly scheduled meetings that increase their knowledge and engagement; these agents also form additional social connections to one another as a result of meeting participation. During the period in which meetings occur, changes in knowledge or engagement for Committee members are “hard-coded” (i.e., changes in knowledge and engagement correspond exactly to analogous survey data). Parameters that determine model dynamics are derived from data sources listed in Table 1. Model dynamics are fully detailed in Supplementary Appendix.
Model Analysis
Each model run consists of 2000 simulated days, with Committee meetings taking place during the first portion (∼2 years). This follows the meeting schedule that took place during the real-world intervention but allows us to explore trends over a longer time frame in our simulation than were observed in the real world. We use model data for three distinct but related analyses.
Direct comparison to analogous real-world data
This analysis consisted of testing whether and to what extent our individual-level simulations of daily interactions are capable of generating patterns similar to those observed in the real world in our measurements of interest (i.e., knowledge and engagement). Our challenge was finding a strategy that utilized rich real-world data (i.e., measurements for the same individuals at multiple time points) for which we could produce simulated analogues that provide useful information about our model's explanatory power. By design, our model explicitly reproduces observed changes in Committee members' values by using these to characterize the effect of meetings on attendees. Therefore, we restricted our analysis to first-degree social network alters who were identified by Committee members and asked to complete up to five rounds of knowledge and engagement surveys throughout the project and compared these observations with analogous simulated individuals (i.e., those who are socially connected to simulated Committee members).
Extrapolation to a wider community
One advantage of our ABM is its ability to capture a larger population of community members than were observed using survey instruments. Once we have ascertained using “apples-to-apples” comparison of simulated and real-world trends in knowledge and engagement of Committee members' immediate social network peers, our confidence in our specific characterization of the interpersonal interactions that comprise SDCD allows us to extend our analyses. Specifically, in this analysis, we use simulation data to estimate the magnitude and speed of change in knowledge and engagement in the whole community.
Exploration of counterfactual scenarios
Similar to the previous analysis, we leveraged the ability of our ABM to extend beyond observational data. Here, we use the same SDCD and community characterizations as in the previous analyses, but we explicitly vary the “strength” of the intervention (i.e., changes in Committee members' knowledge and engagement relative to what was observed in survey data) and observe the impact that this has on the broader community. This allows us to provide insights that might be useful for the design and implementation of future whole-of-community childhood obesity prevention efforts guided by steering committees.
Results
Direct Comparison to Analogous Real-World Data
Our model produces output that is consistent with the observed real-world data within the parameter space explored; a more detailed quantitative comparison of the results appears in Supplementary Appendix. In Figure 2, we present results from 10 simulation runs using a parameter combination that produced the best fit between simulated and real-world data. Simple comparisons of means indicate statistical equivalence for each time point considered (within ranges appropriate given the small sample size and potential for missing data). These results build confidence in the explanatory power of the model to appropriately project patterns of change observed in available analogous data.
Figure 2.
Comparison of real-world distributions of observed knowledge and engagement for alters identified by Committee members to analogous model output, across five rounds of longitudinal survey deployment. Model data are taken from simulated Committee members' immediate (i.e., “first degree”) social network peers at simulated time points that correspond to when each round of real-world survey data was collected). Dots represent distribution means and lines one standard deviation. Color image is available online.
Extrapolation to a Wider Community
In a second analysis, we used the simulation model to project impact of the intervention beyond the limits of currently available data. This is particularly important given the premise of SDCD theory (that diffusion across the broader community facilitates sustainable change). Figure 3 represents distributions of knowledge and engagement for the same parameterization of the SDCD process used in Figure 2, but for a longer time period and for the much larger number of individuals in the wider community who might affect the implementation of childhood obesity prevention efforts. These results indicate a substantial increase in community knowledge—effectively doubling mean knowledge (the black line in the figure) during the simulated time period—along with a slight increase followed by stabilization at high levels of community engagement.
Figure 3.
Trends in knowledge (averaged across domains) and engagement values for the entire community across 10 simulation runs. The black lines depict community mean values, with ranges of values observed across simulated community members displayed in shades of blue. Color image is available online.
Exploration of Counterfactual Scenarios
In a third set of analyses, we make use of the ability of simulation models to explore “counterfactuals” (scenarios that differ from those observed in the specific case measured empirically). Here, we apply the model to understand how changes in implementation choices in the intervention may have led to different outcome patterns. Because the following results are based on directly comparable counterfactual scenarios, they can be interpreted as causal.25 Figures 4 and 5 show the impact relative to the real-world “baseline” (i.e., the scenario depicted in Fig. 3) on community changes in mean engagement and knowledge, respectively, of different meeting impact scenarios (including no impact, effectively capturing a counterfactual condition where meetings were not convened).
Figure 4.
Distributions (means and a one standard deviation range) of intervention impact on engagement within the entire community across different counterfactual scenarios (repeated 10 times) across different “Intervention Strength” scenarios. Intervention Strength represents the change in meeting attendees' knowledge and engagement relative to what occurred during “baseline” runs that use observed changes in these values (shown here as the Intervention Strength 100% scenario). Impact is given as mean change in engagement throughout the entire community during the period during which meetings took place relative to the baseline scenario (on the left Y-axis) as well as relative to initial values (on the right Y-axis).
Figure 5.
Distributions (means and a one standard deviation range) of intervention impact on knowledge within the entire community across different counterfactual scenarios (repeated 10 times) across different “Intervention Strength” scenarios. Intervention Strength represents the change in meeting attendees' knowledge and engagement relative to what occurred during “baseline” runs that use observed changes in these values (shown here as the Intervention Strength 100% scenario). Impact is given as mean change in knowledge throughout the entire community during the period during which meetings took place relative to the baseline scenario (on the left Y-axis) as well as relative to initial values (on the right Y-axis).
For mean community knowledge, the marginal impact of varying effects on Committee meeting attendees is small (e.g., a scenario without an active intervention saw a mean increase in community knowledge that is ∼90% of a “baseline” representing the observed intervention). Conversely, although Figure 3 shows that community engagement levels were initially high and did not change much, intervention impact on community engagement increases nearly linearly with meeting impact.
Discussion
The SDCD approach is premised on creating the preconditions for sustainable change in communities by diffusing knowledge and engagement widely. Previous work has estimated the impact of diffusion interventions on a small sample closely linked to the stakeholder steering committee,18 but the degree to which this translates into the much broader diffusion envisioned by SDCD theory is unknown. To the best of our knowledge, this is the first attempt to apply rigorous theory-informed computational modeling in conjunction with data collected beyond individuals involved in a steering committee specifically for this purpose to examine a contemporaneous whole-of-community childhood obesity intervention.
Our ABM of the SDCD process generates key outcome patterns similar to those observed in real-world data. This builds confidence in our previously tested operationalization of the SDCD process into a mathematical model and also strengthens support for the underlying SDCD hypothesis about how work with a central committee of stakeholder representatives can positively impact knowledge about and engagement with childhood obesity prevention outside the committee itself. We next went one step further: our simulations allow us to isolate the SDCD process from other confounding factors that likely exist in the real world.
In this way, we can extrapolate whether and to what extent SDCD alone is likely to increase knowledge and engagement throughout the entire community over a longer time frame than what was observed. Much like analogous physical phenomena (e.g., sound waves), we anticipate that SDCD effects diminish given greater distance; across a large social environment such as a nation, intervention impact from a single community would be imperceptible. In this case, our model was able to detect positive changes in knowledge and engagement throughout the relevant Somerville community; these changes were substantial (an approximate doubling) for knowledge and slight for engagement. Taken as a whole, this has substantially increased our optimism about the potential for the SDCD process—and interventions that are designed to intentionally harness it—to act as important drivers of positive sustained change that supports children's health in other communities.
Finally, we explored counterfactual scenarios where steering committee meetings had smaller or larger effects on attendees than those in the real world. This effectively separates out diffusion processes—which we hypothesize occur regardless of an active intervention—from the intervention itself, which can “boost” these processes. Results indicate that at least some of the community-wide changes are attributable to the intervention itself and might have differed had implementation been altered. Specifically, ∼10% of the estimated change in knowledge and all the change in engagement are attributable to direct intervention effects on committee members.
These results can serve as both a note of caution and of promise for the implementation of similar interventions in the future: context probably matters a lot, and you should thus look before you leap. Our exploration of “counterfactual” scenarios suggests an initial community context (i.e., social network structure and community members' knowledge and engagement levels before the intervention) in which knowledge was primed to increase substantially with or without active input, while engagement was initially high but could be increased with intentional effort.
There are quite likely community contexts where active interventions have much greater and lesser potential impacts. One of the benefits of the simulation model that we outline here is in its possible application a priori to a given context to help answer whether such an intervention is advisable at all and, if so, how it should be designed. If equipped with such insights, intervention experts might adjust their approach to yield greater positive impact by “targeting” high-potential contexts and “tailoring” to them.26 The use of the data collection and ABM tools that we describe here can thus serve a critical role in the design and implementation of future interventions, and our research team is actively participating in these endeavors.27
We note two main potential limitations of this research. First, one can almost always have more and better data; we acknowledge that we rely on a sample of the community in conjunction with expert estimates to parameterize our model, either of which could be a source of potential noise. Second, as we note above, we are concerned here with “upstream” outcomes (i.e., knowledge and engagement levels). Although the linkage between these and “mid-stream” effects on policy systems, environment and practice and, through these, downstream changes in health are theorized, they have not yet been quantitatively specified. This presents a hindrance for a thorough (immediate) assessment of model explanatory power and interpretation of output.
We envision three desirable extensions of the work described here. The first is application to other contexts (e.g., different communities with varying levels of knowledge and engagement, and diverse intervention designs). One of the core tenets of systems science is iterative improvement of models; additional testing across settings and subsequent generalization can help strengthen explanatory and predictive power. Second, the model can be used to generate insights based on output with finer granularity. As data collection and scientific understanding in this topic space continue to advance, we anticipate that a computational model such as this one will be able to produce more specificity in its guidance and help midstream interventions that use evidence-based strategies take hold more effectively.
Finally, we anticipate an expansion of the model boundary in future work. By building upon what we have learned from our ABM of Shape Up Under 5, we believe that it will be possible to develop a computational tool that can directly incorporate how changes in knowledge and engagement, moderated by other community contexts (e.g., demographics and built environment), translate into effective implementation of policies and practices.
Supplementary Material
Funding Information
This work was supported by a U.S. National Institutes of Health grant, “Systems Science to Guide Whole of Community Childhood Obesity Interventions” (1R01HL115485).
Author Disclosure Statement
No competing financial interests exist.
Supplementary Material
References
- 1. CDC. Childhood Obesity Facts | Overweight & Obesity | CDC. 2021. https://www.cdc.gov/obesity/data/childhood.html (Last accessed April 22, 2021).
- 2. Bleich SN, Vercammen KA, Zatz LY, et al. Interventions to prevent global childhood overweight and obesity: A systematic review. Lancet Diabetes Endocrinol 2018;6:332–346. [DOI] [PubMed] [Google Scholar]
- 3. Economos CD, Hammond RA. Designing effective and sustainable multifaceted interventions for obesity prevention and healthy communities. Obesity (Silver Spring) 2017;25:1155–1156. [DOI] [PubMed] [Google Scholar]
- 4. National Academies of Sciences, Engineering and Medicine. Using Systems Applications to Inform Obesity Solutions: Proceedings of a Workshop. The National Academies Press: Washington, DC, 2021. [PubMed] [Google Scholar]
- 5. Wolfenden L, Wyse R, Nichols M, et al. A systematic review and meta-analysis of whole of community interventions to prevent excessive population weight gain. Prev Med 2014;62:193–200. [DOI] [PubMed] [Google Scholar]
- 6. Bell AC, Simmons A, Sanigorski AM, et al. Preventing childhood obesity: The sentinel site for obesity prevention in Victoria, Australia. Health Promot Int 2008;23:328–336. [DOI] [PubMed] [Google Scholar]
- 7. Borys J-M, Richard P, du Plessis HR, et al. Tackling health inequities and reducing obesity prevalence: The EPODE community-based approach. Ann Nutr Metab 2016;68(Suppl 2):35–38. [DOI] [PubMed] [Google Scholar]
- 8. Chomitz VR, McGowan RJ, Wendel JM, et al. Healthy living Cambridge kids: A community-based participatory effort to promote healthy weight and fitness. Obesity (Silver Spring) 2010;18:S45–S53. [DOI] [PubMed] [Google Scholar]
- 9. de Silva-Sanigorski AM, Bell AC, Kremer P, et al. Reducing obesity in early childhood: Results from Romp & Chomp, an Australian Community-Wide Intervention Program. Am J Clin Nutr 2010;91:831–840. [DOI] [PubMed] [Google Scholar]
- 10. Economos CD, Hyatt RR, Must A, et al. Shape Up Somerville Two-Year Results: A community-based environmental change intervention sustains weight reduction in children. Prev Med 2013;57:322–327. [DOI] [PubMed] [Google Scholar]
- 11. Economos CD, Curtatone JA. Shaping up Somerville: A community initiative in Massachusetts. Prev Med 2010;50:S97–S98. [DOI] [PubMed] [Google Scholar]
- 12. Hoelscher DM, Springer AE, Ranjit N, et al. Reductions in child obesity among disadvantaged school children with community involvement: The Travis County CATCH trial. Obesity (Silver Spring) 2010;18:S36–S44. [DOI] [PubMed] [Google Scholar]
- 13. Hoelscher DM, Butte NF, Barlow S, et al. Incorporating primary and secondary prevention approaches to address childhood obesity prevention and treatment in a low-income, ethnically diverse population: Study design and demographic data from the Texas Childhood Obesity Research Demonstration (TX CORD) Study. Child Obes 2015;11:71–91. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Spears-Lanoix EC, McKyer ELJ, Evans A, et al. Using family-focused garden, nutrition, and physical activity programs to reduce childhood obesity: The Texas! Go! Eat! Grow! Pilot Study. Child Obes 2015;11:707–714. [DOI] [PubMed] [Google Scholar]
- 15. Hunt H, Anderson R, Coelho H, et al. The effectiveness of whole system approaches to prevent obesity. Lond Natl Inst Health Care Excell 2011. [Google Scholar]
- 16. Korn AR, Hennessy E, Tovar A, et al. Engaging coalitions in community-based childhood obesity prevention interventions: A mixed methods assessment. Child Obes Print 2018;14:537–552. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Appel JM, Fullerton K, Hennessy E, et al. Design and Methods of Shape Up Under 5: Integration of systems science and community-engaged research to prevent early childhood obesity. PLoS One 2019;14:e0220169. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Kasman M, Hammond RA, Heuberger B, et al. Activating a community: An Agent-based model of Romp & Chomp, a whole-of-community childhood obesity intervention. Obesity (Silver Spring) 2019;27:1494–1502. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Korn AR, Appel J, Hammond RA, et al. Validation and refinement of the stakeholder-driven community diffusion survey for childhood obesity prevention. Implement Sci 2021;16:91. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Hennessy E. Designing an agent-based model for childhood obesity interventions: A case study of ChildObesity180. Prev Chronic Dis 2016;13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Hennessy E, Korn AR, Economos CD. Chapter 23—A community-level perspective for childhood obesity prevention. In: Bagchi D (ed), Global Perspectives on Childhood Obesity, 2nd ed. Cambridge, MA: Academic Press, 2019, pp. 287–298. [Google Scholar]
- 22. Calancie L, Fullerton K, Appel JM, et al. Implementing group model building with the Shape Up Under 5 Community Committee Working to Prevent Early Childhood Obesity in Somerville, Massachusetts. J Public Health Manag Pract 2022;28:E43–E55. [DOI] [PubMed] [Google Scholar]
- 23. Korn AR, Hammond RA, Hennessy E, et al. Evolution of a coalition network during a whole-of-community intervention to prevent early childhood obesity. Child Obes 2021;17:379–390. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Korn AR, Hennessy E, Hammond RA, et al. Development and testing of a novel survey to assess stakeholder-driven community diffusion of childhood obesity prevention efforts. BMC Public Health 2018;18:681. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Imbens GW, Rubin DB. Causal Inference in Statistics, Social, and Biomedical Sciences. Cambridge, UK: Cambridge University Press, 2015. [Google Scholar]
- 26. Gillman MW, Hammond RA. Precision treatment and precision prevention: Integrating “below and above the Skin.” JAMA Pediatr 2016;170:9–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Catalyzing Communities. Catalyzing Communities | Helping Children Thrive in Healthy Environments. 2022. https://catalyzingcommunities.org/ (Last accessed February 2, 2022).
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.