Proceedings of a Workshop
Using Systems Applications to Inform Obesity Solutions
Proceedings of a Workshop—in Brief
The Roundtable on Obesity Solutions of the Health and Medicine Division of the National Academies of Sciences, Engineering, and Medicine held a virtual public workshop, Using Systems Applications to Inform Obesity Solutions, on September 16, 2020. The workshop, which built on the roundtable’s spring 2020 workshop1 that introduced complex systems thinking and systems science approaches to obesity solutions, explored various systems science applications (i.e., methodologies and tools) that could guide future obesity research and action and featured examples of how such approaches have informed decision making within policy and program areas. Workshop presentations also discussed the support structures (e.g., data sources, modeling expertise, training, and partnerships and collaborations) that encourage and engage researchers and decision makers to use systems science approaches to more effectively and efficiently understand the causes of and solutions to the obesity epidemic. Throughout this proceedings the use of systems science “approaches,” “applications,” “methods,” and “models” are used interterchangeably to describe analytical methodologies and tools.
This Proceedings of a Workshop—in Brief highlights the presentations and discussions that occurred at the workshop and is not intended to provide a comprehensive summary of information shared during the workshop.2 The information summarized here reflects the knowledge and opinions of individual workshop participants and should not be seen as a consensus of the workshop participants, the Roundtable on Obesity Solutions, or the National Academies.
THE PROMISES AND PITFALLS OF SYSTEMS SCIENCE APPROACHES: A PRACTITIONER’S PERSPECTIVE
The workshop’s first session featured speaker Jack Homer, director of Homer Consulting, who introduced systems science models that can help address complex public health issues, such as obesity, and discussed opportunities and challenges associated with various stakeholders’ use of these applications.
Homer began by affirming that public health stakeholders want to make decisions that stand the test of time and avoid shortsightedness, but said this is a challenge for systems science models because uncertainties in social systems hamper the ability to make precise forecasts. He referenced statistician George Box’s famous quote, “all models are wrong, but some are useful,” to highlight a key characteristic of useful models, which is that they can correctly an-
1 Integrating Systems and Sectors Toward Obesity Solutions: Part 1, see https://www.nationalacademies.org/event/04-06-2020/ integrating-systems-and-sectors-toward-obesity-solutions-part-1 (accessed October 28, 2020). Integrating Systems and Sectors Toward Obesity Solutions: Part 2, see https://www.nationalacademies.org/event/06-30-2020/integrating-systems-and-sectorstoward-obesity-solutions-part-2 (accessed October 28, 2020).
2 The workshop agenda, presentations, and other materials are available at https://www.nationalacademies.org/event/09-16-2020/ using-systems-applications-to-inform-obesity-solutions-a-workshop (accessed October 28, 2020).
ticipate intervention impacts and serve as useful decision-making tools even if their baseline predictions are imprecise. He added that sensitivity tests to assess the level of confidence in results often reveal that a model’s policy conclusions are robust despite uncertainty in point predictions.
Homer mentioned three simulation approaches to systems science modeling—system dynamics simulation, discrete event simulation, and microsimulation (individual actors without interaction)/agent-based models (individual actors with interaction)3,4—and underscored that despite differing levels of analysis, time constraints, approaches to uncertainty, and heterogeneity of individuals, all three approaches agree that models should be testable, focused, and scientifically developed.5
He moved on to contrast systems science models, which he described as quantitative simulation approaches, compared with systems maps, which he described as qualitative pictures that can be useful despite limitations such as not being problem focused. Referencing an ongoing academic debate about the appropriate roles of qualitative maps and group model building6 in systems science efforts, Homer summarized that both approaches have value in enriching the modeling process, but have limitations such as the inability to predict behavior.
Homer shifted to focus on quantified systems science simulation models and practical considerations for their reliability. He described a conceptual diagram illustrating the “possibility frontier” of reliable systems science modeling, which he explained as a depiction of modeling choices and their feasibility and value (see Figure 1).
The number of interacting concepts in a model (i.e., the breadth) moves from very narrow to very broad as one moves up the y-axis, he explained, and the x-axis represents the level of detail (i.e., the number of subcategories per concept), from little detail to very disaggregated details. The dotted line in the graph area represents the possibility frontier, beyond which lies a “black box” area of broad, detailed models that are too difficult to validate, modify, and understand. Homer described several types of reliable systems science models within the possibility frontier, highlighting the “policy sweet spot,” which he said is marked by models with adequate detail and breadth to be useful for decision makers.
Homer next described a conceptual diagram for considering a systems science model’s level of evidence (i.e., reliability), graded as A, B, or C. The broader the model’s scope, the longer it takes to achieve a high (“A”) level of reli-
3 Borshchev, A., and A. Filippov. 2004. From system dynamics and discrete event to practical agent based modeling: Reasons, techniques, tools. Proceedings of the International Conference of System Dynamics Society, 22nd, Oxford.
4 Luke, D. A., and K. A. Stamatakis. 2012. Systems science methods in public health: Dynamics, networks, and agents. Annual Review of Public Health 33:357–376.
5 Homer, J. B. 1996. Why we iterate: Scientific modeling in theory and practice. System Dynamics Review 12(1):1–19. Levy, D. T., P. L. Mabry, Y. C. Wang, S. Gortmaker, T. T-K. Huang, T. Marsh, M. Moodie, and B. Swinburn. 2011. Simulation models of obesity: A review of the literature and implications for research and policy. Obesity Reviews 12(5):378–394. doi: 10.1111/j.1467-789X.2010.00804.x.
6 Siokou, C., R. Morgan, and A. Shiell. 2014. Group model building: A participatory approach to understanding and acting on systems. Public Health Research & Practice 25(1):e2511404.
ability. As an example, he said a narrow model might take days or weeks to achieve a high level, whereas a broader model could take months or years to achieve the same level.
Homer reviewed several system dynamics models that either focus on or include obesity. He highlighted a 2005 system dynamics model that depicted a cascade in which caloric imbalances translate to changes in weight and eventually changes in body mass index (BMI) categories across the U.S. population, stratified by sex and single age cohorts. The model forecasted patterns of weight change over time by age category, he explained, and one of its key findings was that ideal policy responses would combine preventive interventions with effective weight-loss options for people with obesity.
Lastly, Homer suggested a blueprint for an ideal systems science modeling project with adequate funding and more than 1 year of time allocated to it. He urged following best practices for building, testing, and documenting systems models and their outputs; anchoring models to well-established datasets and the best studies; enlisting analysts to extract data as well as statisticians who can help interpret it; including stakeholders, decision makers, and subject-matter experts as advisors on modeling teams; keeping other thought leaders in the loop; and involving junior modelers as apprentices.
THE USEFULNESS OF SYSTEMS SCIENCE APPROACHES FOR STAKEHOLDERS IN DIFFERENT SECTORS
The workshop’s second session included three panels of speakers who discussed the usefulness of systems science approaches for stakeholders in different sectors—businesses and the private sector, communities, and policy makers—and that illustrate the coordination and partnerships that can develop.
The Business and Private Sectors: Project Play
In the first panel, three speakers discussed a business case study of Project Play, an effort designed to address the U.S. obesity epidemic by building healthy communities through sport.
Tom Farrey, executive director of The Aspen Institute Sports & Society Program, explained that the first phase of Project Play focuses on children ages 12 and younger and aims to ensure that every child has the opportunity to engage in either structured or unstructured sports regardless of zip code or ability. He described three steps the initiative has taken toward that aim.
The first step was to compile and organize existing research on the lifelong individual- and community-level benefits of physical activity and sport activity. This research informed the development of a playbook called Sport for All, Play for Life,7 Farrey continued, the nation’s first cross-sector framework for action in youth sports. He noted the playbook was an outgrowth of 2 years of information gathering and roundtable discussions with stakeholders and thought leaders during which the best ideas were surfaced and distilled into an attractive, easy-to-read document. The playbook highlights eight sectors that touch children’s lives and presents eight strategies to apply in those sectors to encourage more children to be active through sport.
Farrey reviewed the playbook’s eight strategies: (1) putting children at the center of designing sport experiences; (2) reintroducing free play; (3) encouraging sport sampling; (4) revitalizing in-town, local sport leagues; (5) thinking small, which Farrey explained with an example of brokering shared use agreements to use community spaces creatively; (6) designing for development (i.e., anchoring a sport system in the principles of developmentally appropriate play); (7) training all coaches in how to make sport a positive experience for children, including recognizing the differences in their physical, mental, and emotional capacities at different ages; and (8) emphasizing prevention, which Farrey said emerged to address parental concerns about children sustaining physical and emotional injuries from participating in sports.
Farrey moved on to describe Project Play’s second step in pursuing its mission, which was mobilizing organizations. This began by convening the initiative’s first annual summit, Farrey recalled, highlighted by a call to action around the value of youth sports to promote physical activity and public health. Since then, Project Play has conducted “state of play” audits that help communities assess where they are and how they could design investments and policies and mobilize stakeholders to activate and improve the quality of play in their communities. Project Play convenes stakeholders to share new tools, projects, and resources at its annual summit, Farrey continued, which also showcases the release of an annual State of Play report featuring trend data on children’s sport participation.
Farrey next described Project Play 2020, a convening that gathered some of the largest and most influential organizations in the Project Play network to discuss how to collectively invest in increasing activity among children through age 12. This led to the development of howtocoachkids.org, the nation’s first freely available resource that
aggregates coaching materials for any organization to use with its coaches, and the Healthy Sport Index, which helps parents and stakeholders assess the relative risks and benefits of participating in the 10 most popular sports for high school students.
Farrey described a final step: mobilizing parents. He noted research that suggests that most children play sports for an average of 2.9 years and typically stop by age 11. He highlighted a media campaign introduced in August 2019 called “Don’t Retire, Kid,” which directed parents to advice and resources related to maintaining children’s participation in sports.
Bruce Y. Lee, professor of health policy and management at the City University of New York Graduate School of Public Health & Health Policy and executive director of Public Health Informatics, Computational, and Operational Research (PHICOR), began by reiterating that decision makers benefit from information quantifying the impact of a problem as well as the potential value of different types of interventions. But producing this type of information is a challenge when the problem is complex, he explained, such as the physical inactivity epidemic in the United States.
Lee described a suite of systems science models called Virtual Population Obesity Prevention (VPOP), developed by PHICOR, to illustrate its utility to help decision makers better grapple with the complexity of diet- or physical activity–related problems and their potential solutions for stakeholders engaged in Project Play. In these agent-based models, each person is represented by a computational agent with autonomous decision making, which means that their actions may be influenced by others but are ultimately determined by individual freedom of choice. He added that agents also exhibit basic complex adaptive behavior, which means they can learn from the past or do things differently depending on previous events.
The VPOP model uses a synthetic population that was built using U.S. Census data, Lee continued, and each agent has assigned physical and demographic characteristics. The initial model was built for Baltimore, Maryland, and the same structure then informed building models for additional cities as well as the entire United States. Modelers assign the agents a daily schedule of activities, behaviors, and decisions that mimic a real person’s patterns, Lee explained, and each agent is embedded with a personalized metabolic model that attempts to represent what happens to an individual’s weight status when varying amounts of calories are ingested and expended.
Lee explained that his team attached the metabolic model and agent characteristics to an embedded clinical and economic outcomes model representing key health conditions potentially associated with changes in BMI. In this model of clinical and economic outcomes for different BMIs over the lifetime, each agent had probabilities of developing diabetes, cardiovascular disease, or cancer over time, he elaborated, and accrued varying levels of medical costs and productivity losses based on the condition that developed and the health care resources consumed. This work led to a general model representing all youth in the United States, Lee continued, which simulated the economic and health impact of increasing physical activity in that segment of the U.S. population.
Lee reported that right now less than one-third of youth in the United States maintain the Sports & Fitness Industry Association’s recommended “active to a healthy level” (i.e., 20 minutes of physical activity three times per week). Results from Lee’s model suggested that increasing this number to at least 50 percent of the current youth population would avert billions of dollars in medical care costs and lost productivity over time.8
Dev Pathik, chief executive officer and founder of the Sports Facilities Companies, described his company’s mission to improve the health and economic vitality of communities by serving as a trusted resource for communities who want to plan, fund, develop, or manage youth and amateur community-based sport and recreation facilities. Pathik elaborated that the Sports Facilities Companies advise municipalities and large institutions on infrastructure planning and resource allocation to activate their objectives around sport and recreation.
Pathik, a partner in Project Play with access to the systems science modeling previously described by Lee, reviewed five access-related factors for healthy communities, emphasizing that infrastructure plays a role in these factors, which are intended to improve health and social outcomes for children. The first factor, he began, is safe and free spaces to play. A second factor is activity-focused urban design, which Pathik explained by listing example measures such as the walkability and the interconnectivity of recreation and sport assets. A third factor is the utilization and programming of school assets and sport and recreation facilities. A fourth factor is investing in multigenerational programs and events and a fifth factor is multimodal transportation. According to Pathik, these five access-related factors help communities think more broadly about how they participate in, contribute to, and fund or use partnerships to promote wellness in their communities.
Pathik said the last recession brought significant defunding of parks and recreation, which prompted the private sector to capitalize on opportunities in that marketplace. But without the programming previously offered by
8 Lee, B. Y., A. Adam, E. Zenkov, D. Hertenstein, M. C. Ferguson, P. I. Wang, M. S. Wong, P. Wedlock, S. Nyathi, J. Gittelsohn, S. Falah-Fini, S. M. Bartsch, L. J. Cheskin, and S. T Brown. 2017. Modeling the economic and health impact of increasing children’s physical activity in the United States. Health Affairs (Millwood) 36(5):902–908. doi: 10.1377/hlthaff.2016.1315.
parks and recreation departments, he explained, facilities that may still be well maintained are now more expensive for people to use. The loss of public commitment to funding sports and recreation at the programmatic level results in a system designed toward capitalism, he elaborated, which provides a place to play but lacks the programmatic structure to help kids develop competence and confidence to engage in activity for a lifetime.
To build the political will for investment in and activation of sport and recreation infrastructure that can reach those most in need, Pathik advocated for a multistakeholder approach. Reflecting on past experiences, he suggested that potential key participants include a major medical partner, major financial institution, elected officials, representatives from the city manager’s office and economic development organizations, parks and recreation directors, and other community health and wellness partners.
To equip and empower parks and recreation directors and advocates to express the value of recreation and sports, Pathik said that the Sports Facilities Companies and the Florida Recreation & Park Association produced calculators that generate reports about how individual parks contribute to property values, health care savings, environmental and public spending, and job creation in the community. He reported that this kind of data resulted in billions of dollars in public investment into sport and recreation assets.
Farrey concluded the presentation with a summary of Project Play’s impact since the 2015 launch of its Sport for All, Play for Life playbook. In summary, he reported that more than 100 organizations and companies have taken actions as a result of an activated broad network of organizations that have participated in Project Play activities, and have introduced an array of aligned programs.
Communities: Baltimore, Maryland
In the second panel, three speakers discussed the use of systems science models in Baltimore and offered another example of coordination and partnership of efforts. They provided an overview of the Baltimore food system, discussed a specific community project, and shared three examples of how systems science models informed potential policies and interventions in the city and engaged community members and other stakeholders. The speakers, who took turns presenting as they covered different portions of the presentation, included Lee, who also spoke in a previous panel; Joel Gittelsohn, professor in the Center for Human Nutrition, Johns Hopkins Bloomberg School of Public Health; and Sarah Buzogany, food resilience planner, Baltimore Food Policy Initiative, Baltimore City Department of Planning.
Lee briefly described the Global Obesity Prevention Center, previously housed at Johns Hopkins University, and its goal to use different types of systems science approaches, including mapping and modeling, to understand the complex system of factors involved in the obesity epidemic. One of its main initiatives, he said, was a multiscale, multicomponent intervention project (led by Gittelsohn) to address obesity in Baltimore City by applying different systems science models to understand all of the complexities involved in Baltimore City’s food system.
Buzogany then discussed Baltimore’s food system and food policy landscape. For the past 10 years, the Baltimore Food Policy Initiative has driven the city’s food policy agenda in partnership with and alongside residents, organizations, and partners such as Johns Hopkins University. The relationships among these institutions and the city have been instrumental, she declared, because academic partners have used systems science models to help community stakeholders identify needs, understand the potential implementation of strategies and plans, and suggest viable policy solutions.
Buzogany stated that one of Baltimore’s frames is to use food as a catalyst to address health, economic, and environmental disparities in healthy food priority areas. The term healthy food priority areas9 describes a geographic area of the city where residents may face structural barriers to accessing healthy foods, she explained, and evolved from the term “food desert.” The Baltimore Food Policy Initiative mapped the city’s healthy food priority areas, she recalled, and partnered with Gittelsohn to use the maps to explore policy questions such as the effect of a staple foods ordinance on food quality. By collaborating on systems science modeling efforts, she added, the initiative allowed planners to understand how they could best implement strategies to improve the food environment.
Gittelsohn discussed B’More Healthy Communities for Kids, a multilevel program implemented by the Global Obesity Prevention Center. He focused on the program’s policy working group, which was comprised of Baltimore City’s food policy director, food resilience planner, and representatives from various sectors: city council, city health department, city schools, family league, recreation and parks, wholesale companies, and academia. Gittelsohn reported that the group engaged in a series of planning exercises and activities to develop and consider solutions for improving the city’s food environment. Gittelsohn mentioned that he and his team were specifically asked by policy working group members to develop a simulation model to provide evidence for a proposed urban farm tax credit. Gittelsohn
9 For more information, see https://planning.baltimorecity.gov/baltimore-food-policy-initiative/food-environment (accessed November 24, 2020).
explained that the credit would provide a 90 percent tax reduction to owners of vacant lots if they converted them to urban farms. Buzogany noted that policies like the urban agriculture tax credit were created to help put the city’s vacant land to productive use and to support ownership by urban farmers, particularly black farmers and other farmers of color, without burdening them with large property tax bills.
Gittelsohn elaborated on the Global Obesity Prevention Center’s use of a systems science agent-based model called BLIFE (Baltimore Low Income Food Environment), an acronym that he said was modified to provide evidence in support of the potential impact of the proposed urban agriculture tax credit. The modeling simulation was provided to the city council in a testimony as evidence, he recounted, and legislation was passed to establish the tax credit in Baltimore City. Lee emphasized that although the BLIFE model—the first example of a systems science model they highlighted, with its 300 computational agents—was useful for testing different policies and interventions, the expanded model, VPOP, which Lee discussed in the session’s first panel, included millions of computational agents.
Lee described this second example model and how the VPOP model forecasted the potential effect (on purchase and consumption behaviors) of placing sugar-sweetened beverage warning labels in different combinations of grocery stores, corner stores, schools, and other settings. Results indicated that a warning label would result in people choosing an alternative to sugary beverages 8 percent of the time, a finding that Lee said was used as a baseline to forecast the impact of no labels compared with labels at 4 percent, 8 percent, and 12 percent efficacy in three major cities. He reported that the model indicated that as the efficacy of warning labels increased, a greater decrease in the prevalence of obesity occurred—effects that were consistent in direction but not in size across the cities.10 Lee explained that the heterogeneity of effects across cities is a result of differences in geographic layouts, types of food sources, and combinations of population BMI distributions and health conditions, which he said shows the importance of using systems science models that explicitly represent the population and food environments in an intervention’s target city.
Next, Buzogany provided an introduction to the third example of using systems science modeling to inform Baltimore’s policy making. A community advisory group tasked to co-create policy with the city planning department and government developed a recommendation to require corner stores to carry a minimum stock of health-promoting foods, she explained, leading the group to partner with Gittelsohn to develop a system dynamics model to simulate the impact of a staple foods ordinance.
According to Gittelsohn, modelers manipulated variables relating to supply, consumer behavior, and ordinance characteristics to simulate four scenarios in which Baltimore corner stores could implement a staple foods ordinance. The scenarios varied with regard to levels of enforcement and required a minimum stock of different foods and beverages. Among the model’s many outputs, reported Gittelsohn, was a projection of weekly profits that corner stores could expect to reach if they implemented a staple foods ordinance under each of the four simulated scenarios. He reported that three of the scenarios were projected to turn a small yet significant profit, but one scenario had minimum stock requirements that were burdensome to the point of being unsustainable in the simulation. That scenario would likely lead implementing stores to quickly go out of business, Gittelsohn clarified, essentially tabling it from policy consideration.
Lee ended the panel by emphasizing the value of collaboration among modeling experts and community decision makers and stakeholders, which he said enables an iterative feedback loop that informs and improves study design, data collection, and systems science modeling parameters.
Policy Makers: SALURBAL
The third panel comprised three speakers who shared experiences with policy maker engagement in an international collaborative initiative called SALURBAL (Salud Urbana en América Latina [Urban Health in Latin America]), which incorporates systems thinking and systems science approaches to promote urban health and environmental sustainability in Latin American cities.
Ana Diez Roux, dean and distinguished university professor of epidemiology in the Dornsife School of Public Health at Drexel University, opened the panel with an overview of SALURBAL’s goals and aims. She highlighted its goal of engaging policy makers and the public in a new dialogue about urban health and sustainability and implications for societal action; and the aim to employ systems thinking and simulation models to evaluate urban–health–environment links and plausible policy impacts. The study has employed multiple strategies to engage policy makers in its systems science aim, Diez Roux said, and she asked her fellow panelists to elaborate on examples of policy maker engagement.
Brent Langellier, assistant professor of health management and policy at Drexel University, stressed that urban systems are complex and that policy makers generally recognize and accept this attribute of their cities. He expounded
10 Lee, B. Y., M. C. Ferguson, D. L. Hertenstein, A. Adam, E. Zenkov, P. I. Wang, M. S. Wong, J. Gittelsohn, Y. Mui, and S. T. Brown. 2018. Simulating the impact of sugar-sweetened beverage warning labels in three cities. American Journal of Preventive Medicine 54(2):197–204. doi: 10.1016/j.amepre.2017.11.003.
on SALURBAL’s participatory group model–building exercise and its two policy-oriented, agent-based models. The former, he explained, involved three group model–building workshops conducted in three countries with 62 policy stakeholders from academia, public health and planning agencies, and nongovernmental organizations that influence local policies. The outcomes of interest were healthy eating and mobility/transport in Latin American cities, he continued, which the policy stakeholders explored through scripted activities to introduce systems thinking and work toward mapping systems.
Langellier explained that the first scripted activity, graphs over time, was intended to engage participants in framing a problem, initiating mapping, generating variables, and ranking the priority of variables as preparation for creating causal loop diagrams. In the second activity, participants were instructed to create a causal loop diagram in small groups and then present it to the large group. A synthesis activity followed, where the small groups worked with facilitators to unify their diagrams into a single synthesis causal loop diagram that explained a hypothesis of the food system/transportation system factors that influence a healthy urban environment. He pointed out that this is where the diverse and often overlapping perspectives of the stakeholders emerged, as the synthesis model took shape and participants quickly recognized the complexity and interconnectedness of the systems involved.
After the causal loop diagrams were created, Langellier said that participants completed a final scripted activity called action ideas. After they reflected on the system’s structure and feedback loops to help generate potential policy interventions to improve health outcomes, they placed the interventions on a 2 × 2 grid that ranked them by ease of implementation (easy or hard) as well as by level of impact (high or low). This helped them consolidate the easy-to-implement, high-impact ideas, said Langellier, which in the workshops included advocating for food and transport, promoting cycling tours among healthy food sources in the city, and microcredit and credit assistance for urban farming.
Langellier shifted to review the ongoing development of SALURBAL’s two agent-based simulation models of urban policy, beginning with one about purchasing policy for ultra-processed food. He explained that the model will explore how food labeling, advertising, and taxes can be most effectively combined to reduce purchases of ultra-processed food in Latin American cities, as well as the differences in policy effects across different population segments (i.e., people with high versus low levels of income and educational attainment). Though the model uses population demographics and ultra-processed food purchasing data from Mexico City, Langellier explained that modelers can quickly change the parameters to apply the model to future policy proposals in different locations.
Langellier moved on to discuss the second model, which was developed to simulate commuter decision making and behavior in a Bogotá-inspired city where there are five commute modes: car, motorbike, bus, bus rapid transit, and using a bicycle or walking. This model explores how public transportation policies, taxation, and interventions to improve personal safety from crime affect mode share, physical activity, and air pollution. The model is spatially explicit, he pointed out, which accounts for residential segregation in a city, income inequity, and commutes between homes and workplaces. He said agents make a daily commute-to-work decision based on the modes of transportation available to them, their determination of the highest-utility mode (a function of travel time and cost), and their evaluation of the perceived safety of each mode (i.e., avoiding modes that they consider too unsafe based on their threshold for safety).
Felipe Montes Jimenez, associate professor in the Department of Industrial Engineering at the Universidad de los Andes, shared a case study of an effort called TransMiCable that implemented cable car transport in a low-income, densely populated area of Bogotá, Columbia, in 2018. The objective was to assess the effect of TransMiCable’s implementation on environmental and social determinants of health, physical activity (for both leisure and transport), and health outcomes (quality of life, respiratory diseases, and homicides). A key component of the evaluation was its use of citizen science to identify, prioritize, and communicate the intervention’s negative and positive features on health and quality of life in the community, which Montes Jimenez said contributed to empowering the community and increasing its uptake of the intervention.
Montes Jimenez explained that the first step in the evaluation’s conceptual framework was to convene multisector stakeholders, including community members and academics. The stakeholders participated in a group model–building workshop, he recounted, where they engaged in scripted activities for the purpose of developing a causal loop diagram that complemented the conceptual model and identified and explored policy alternatives within the TransMiCable evaluation. After the workshop, researchers interviewed each stakeholder individually to review the diagram, collect input on the feedback loops, and clarify terminology for accuracy with respect to the jargon of each discipline and sector. Following these interviews, the researchers revised the diagram so that it more clearly broke out four key aspects of the system: health conditions, social and economic development, operationalization of TransMiCable, and citizen’s culture. Montes Jimenez explained that the aspect of citizen’s culture involves community participation and ownership of the intervention, as well as reinforcement of inclusive behaviors that nurture well-being and reduce vandalism and gender-based violence in the area where the cable cars operate.
Reflecting on the TransMiCable evaluation, Montes Jimenez observed that the involvement of multisector stakeholders leveled the playing field by giving everyone the same opportunities to contribute to a shared mental model of the system; and that policy makers’ involvement in the conceptual framework for research and project evaluation resulted in their support for the evaluation’s conduct and results dissemination.
REFLECTIONS ON USING SYSTEMS APPLICATIONS TO INFORM OBESITY SOLUTIONS
In the workshop’s final session, each member of the workshop planning committee provided brief remarks on a variety of issues to encourage and help support the use of systems thinking and systems science approaches to address the complexity of obesity drivers and solutions. Following this portion of the session, a closing speaker delivered final reflections on the workshop’s discussions.
Workshop Organizers’ Reflections
Douglas Luke, the Irving Louis Horowitz Professor in Social Policy and director of the Center for Public Health Systems Science at Washington University in St. Louis, offered three strategies for translating systems science modeling results and disseminating resources and tools for community partners. The first is to make a plan for dissemination at the outset of a project, which he said can be facilitated by ongoing dialogue with community partners about the type of resources they need to support their decision making. The second, Luke continued, is to create highly visual, tailored resources, because policy makers are more likely to be engaged when they see a model’s projections for their specific area of concern. The third is to emphasize tools that are interactive, dynamic, and explorable, such as interactive dashboards that visually present a variety of information, maps or other interfaces that show a physical area of interest, before-and-after presentations that display how an environment might look pre- and post-intervention, and resources that allow exploration of what-if scenarios.
Sara Czaja, professor of gerontology and director of the Center on Aging and Behavioral Research at Weill Cornell Medicine and emeritus professor of psychiatry and behavioral sciences at the University of Miami, spoke to the importance of leveraging partnerships and multidisciplinary teams to address complex challenges. She observed that systems science approaches involve the interplay of multisector stakeholders, and urged early planning for how these layers of participants will communicate effectively despite coming from different arenas. In terms of training and education, Czaja continued, it is important for participants in systems science approaches to understand each other’s purposes, contributions, and goals for the targeted effort.
Czaja also encouraged understanding and managing participants’ expectations. For example, it is important for researchers and systems science modelers to determine how community stakeholder input will be used and to understand how those intentions align with community stakeholders’ expectations regarding implementation of their inputs. It is not always feasible or even possible to integrate every input in the model, she noted, but it helps if community stakeholders understand that stipulation from the start.
David Mendez, associate professor in the Department of Health Management and Policy at the University of Michigan School of Public Health, stated that as systems science modeling helps solve a specific problem or answer a particular question, modelers can put various levels of complexity into a model to represent the processes that govern the topic of interest. If the problem is discrete, Mendez suggested that the corresponding mental map and systems science model might not be as complex as a map or a model that shows all of the intertwined aspects of the greater systems that influence a broader problem.
Mendez shared that a challenge of his work in evaluating tobacco policy is that the impact of policies of interest take a long time to materialize. The right model can help indicate a range of time in which policies of interest will be successful, he said; therefore, his team developed a model to gauge the overall mortality caused by smoking over time. He explained that by better understanding the magnitude of the problem the team was able to derive a set of robust policies to address that problem. Mendez also suggested the use of comparative models; that is, different models for the same topic, to assess how they differ in predictions and results. He explained that this tactic can help inform a level of confidence in the potential outcomes.
Jamy Ard, professor in the Department of Epidemiology and Prevention and the Department of Medicine at the Wake Forest University Baptist Medical Center and member of the 2020–2025 Dietary Guidelines Advisory Committee, suggested that the process of developing dietary guidelines should naturally be all about systems because food, nutrition, and the consequences of dietary intake on health are interrelated. Yet, the process of developing dietary guidance has historically taken a reductionist approach, he pointed out, whereby recommendations are more nutrient based versus food or dietary pattern based until recent cycles of the guidelines. Much more can be done, Ard indicated, to integrate a systems science approach into the process of developing dietary recommendations to reduce chronic disease risk in both populations and individuals.
With regard to his role as a medical provider, Ard proposed that health care providers would benefit from training in systems thinking to help shape their perspectives about the context in which their prescribed interventions occur. He appealed for helping providers understand how they can play an influencer role within systems, such as by helping link patients to community supports.
Stella Yi, assistant professor in the Department of Population Health at the New York University Grossman School of Medicine, discussed the formative work involved with engaging stakeholders and community members in participatory model building (refer to SALURBAL presentation). She said that her research group’s long-term relationships with Asian American communities and community-based organizations have afforded much dialogue about issues that are important to residents. Yi noted that these conversations have been a precursor to the group’s current efforts to engage those communities in group model–building exercises. She encouraged participants, saying that although it takes time and effort to foster the connections that lead to long-term, sustainable change in communities, relationship building can start simply with an initial conversation and progress organically as parties discuss mutual interests in a particular population or topic area. Different stakeholders have unique contributions to offer, Yi added, sharing an example in which her university provided data collection tools to community-based organizations that wanted to collect data but lacked the capacity to do so.
Daniel Rivera, professor of chemical engineering and program director of the Control Systems Engineering Laboratory at Arizona State University, offered suggestions for initiating community and stakeholder engagement in the modeling process. Regardless of an intervention’s outcomes of interest, Rivera maintained that some degree of tension typically exists with regard to different language (jargon) and expertise across sectors and disciplines. He mentioned the usefulness of guiding groups to draw path diagrams, which he described as an iterative process that ultimately results in the development of dynamical models that are used for intervention and optimization. He also mentioned the usefulness of anchoring models in strong theoretical foundations and noted an example of an effort that developed a dynamical systems science model for social cognitive theory where one did not exist before.
Lee proposed a change in the paradigm of how modeling is used, maintaining that stakeholders are harnessing only a small percentage of modeling capabilities. He referenced public perceptions that the initial models that predicted COVID-19 mortality counts were “wrong” because they did not exactly match what occurred, but argued that predictions represent only a small sliver of a model’s capability. He described the process of building a budget as an analogy for developing a model and determining its purpose. One would not use a budget spreadsheet (i.e., the financial model) to predict how much money would be earned in a year, he affirmed, because too many variables are involved. But building the spreadsheet would be useful, he said in contrast, to help conceptualize one’s financial context and better understand the relative contribution of each expense category to total spending.
To address a common stakeholder misperception that all models are the same, Lee pointed out that wide diversity exists within modeling approaches and that modelers sometimes use the same technique in different ways. Lee suggested that several modeling teams assess a problem or question so that results can be compared and contrasted.
Patricia (Patty) L. Mabry, interdisciplinary scientist at HealthPartners Institute, delivered the final reflections on the workshop’s discussions and interspersed bits of additional content that was not covered by the speakers. She first commended Homer’s chart comparing characteristics of three simulation approaches for systems science modeling, and added that an entire field of network science and its subfields exist within the column for microsimulation or agent-based models. She also referenced Homer’s blueprint for an ideal, well-funded project, and suggested that another important component is providing tailored education to different stakeholders to explain the appropriate use and expectations of a model.
Mabry next recalled the presentation about the Baltimore Food Policy Initiative and the synthetic population it built using U.S. Census data. She provided a list of secondary sources of data that can be used to populate systems science models and voiced her support for reusing models “when it makes sense.” She pointed to a quantitative mathematical model of the dynamics of childhood growth and obesity as an example of a model that may be applied to other modeling efforts. Mabry also stressed the importance of using spatial models of food and physical environments to help convey inequalities.
Next, Mabry elaborated on the complexity of behavioral and social science problems and the associated challenges of applying systems science approaches to these topics. Challenges include getting appropriate sources of data, particularly longitudinal data; maintaining stakeholder engagement; dealing with impatience to produce actionable results, particularly for policy makers who have limited time in office; building interdisciplinary teams and communicating across disciplinary boundaries; doing what one knows instead of learning what one should do (i.e., lack of self-efficacy); relying too much on randomized controlled trials as the gold standard; following through on a course of
action that models indicate, particularly when it goes against ideology; and dealing with the “prevention conundrum,” which Mabry described as the difficulty in proving that an intervention indicated by a model avoided an impending fate. Humans are not always successful in solving complex problems, she suggested, because it takes hard work and complex solutions. Furthermore, she indicated that people tend to pursue low-hanging fruit; that is, easy opportunities waiting to be seized and requiring little effort. Mabry urged participants to “go for the hard part.”
Mabry moved on to the topic of recruiting systems science modeling teams as she listed various types of stakeholders and experts to consider including. She appealed for including interdisciplinary scientists to help facilitate communication across disciplinary boundaries.
Mabry wrapped up her presentation with a list of additional resources to help people who want to dive into systems science modeling. These included records of prior National Institutes of Health convenings on the topic;11 a theme issue of the American Journal of Public Health;12 Donella Meadows’s book, Thinking in Systems: A Primer;13 the SIPHER (systems science in public health and health economics research) Consortium;14 and the CSART (computer simulation and advanced research technologies) initiative,15 among others.
Finally, Mabry outlined future directions for systems science modeling efforts, with an emphasis on building a more structured community for systems scientists. She also reiterated the value of self-reflection and self-correction in order to enable progress. ◆◆◆
11 See https://www.preventionresearch.org/conferences/training/2007-symposia-series-on-systems-science-and-health (accessed October 23, 2020)
12 Sterman, J. D. 2006. Learning from evidence in a complex world. American Journal of Public Health 96(3):505–514.
13 Meadows, D. H. 2008. Thinking in systems: A primer. White River Junction, VT: Chelsea Green Publishing Company.
DISCLAIMER: This Proceedings of a Workshop—in Brief was prepared by Emily A. Callahan as a factual summary of what occurred at the workshop. The statements made are those of the rapporteur or individual workshop participants and do not necessarily represent the views of all workshop participants, the planning committee, or the National Academies of Sciences, Engineering, and Medicine.
*The National Academies of Sciences, Engineering, and Medicine’s planning committees are solely responsible for organizing the workshop, identifying topics, and choosing speakers. The responsibility for this published Proceedings of a Workshop—in Brief rests with the rapporteur and the institution.
REVIEWERS: To ensure that it meets institutional standards for quality and objectivity, this Proceedings of a Workshop—in Brief was reviewed by Bruce Y. Lee, City University of New York, and Tiffany M. Powell-Wiley, National Heart, Lung, and Blood Institute. Lauren Shern, National Academies of Sciences, Engineering, and Medicine served as the review coordinator.
SPONSORS: This workshop was partially supported by Academy of Nutrition and Dietetics; Alliance for a Healthier Generation; American Academy of Pediatrics; American College of Sports Medicine; American Council on Exercise; American Society for Nutrition; Banner Health; Bipartisan Policy Center; BlueCross BlueShield of North Carolina Foundation; General Mills, Inc.; Intermountain Healthcare; The JPB Foundation; The Kresge Foundation; Mars, Inc.; National Recreation and Park Association; Nemours; Novo Nordisk; Obesity Action Coalition; The Obesity Society; Partnership for a Healthier America; Reinvestment Fund; Robert Wood Johnson Foundation; SHAPE America; Society of Behavioral Medicine; Wake Forest Baptist Medical Center; Walmart; WW International; and YMCA.
For additional information regarding the workshop, visit nationalacademies.org/obesitysolutions.
Suggested citation: National Academies of Sciences, Engineering, and Medicine. 2021. Using systems applications to inform obesity solutions: Proceedings of a workshop—in brief. Washington, DC: The National Academies Press. https://doi.org/10.17226/26043.
Health and Medicine Division
Copyright 2021 by the National Academy of Sciences. All rights reserved.