Original articleReducing racial disparities in obesity: simulating the effects of improved education and social network influence on diet behavior
Introduction
There is no doubt that most behavioral, social, environmental, and biological factors that affect health and health disparities are the result of complex, multifactorial, and multilevel processes, taking place over time. However, for the most part, our analytic approaches struggle with this complexity [1]. The search for independent causes in epidemiology and other health and social sciences is attractive because it can suggest targeted narrowly defined interventions. However, when appreciating the complex determination of population health outcomes, approaches that seek to isolate independent causes as a guide to intervention may mislead by oversimplification. On the other hand, approaches that embrace a dynamic, complex, multifactorial, and multilevel perspective can be daunting from both a conceptual and analytical perspective.
Obesity is an important, and growing, public health problem [2]. However, it is an extraordinarily complex problem, perhaps best exemplified by the Foresight group's systems map of the drivers of obesity that include many dozens of pathways over many levels with considerable interaction and feedback [3]. The number of obesogenic factors and their interconnections does not easily lend itself to traditional analytical approaches such as multivariate regression.
Understanding the causes of obesity and how to prevent it becomes even more challenging when we consider racial and ethnic disparities in levels of obesity and differences in trends in obesity [4]. Because levels and trends in obesity vary considerably between groups, and race and ethnicity stand for the complex intersection of many levels of determinants [5], it is not clear what types of interventions would be most effective in reducing disparities in obesity.
Agent-based modeling is an alternative approach with considerable potential to capture the key characteristics of the obesity system: multiple-levels of analysis, interdependence between levels, heterogeneity in agents or actors, and interdependence within levels (e.g., between individual people in a social network). Furthermore, agent-based models can capture key causal mechanisms and transparently highlight policy levers. This approach implements the key features of a phenomenon as a computer simulation, in which individuals are represented as agents, and a set of explicit rules are defined that dictate how individuals interact with the environment and with each other [1], [6], [7].
With this in mind, we introduce an agent-based model that allows us to capture a large set of dynamic and interactive factors that contribute to obesity and to generate simulations that examine the short- and long-term effects of various approaches to intervening on the obesity epidemic. Although not encompassing anywhere near all the factors identified by the Foresight group [3], the model does encompass behavioral, social, and environmental determinants of dietary intake and physical activity, the chief proximal determinants of obesity.
In this initial presentation of our model, we focus on diet alone. This affords the most straightforward view of how the model works and the potentials of such a model for policy decisions. It should be noted, however, that individual levels of physical activity and neighborhood opportunities for physical activity were included in the model.
Our approach in this article was to model the potential effects of a social policy, an improvement in school quality for neighborhoods in which it is poor, on the reduction in black–white disparities in diet. School quality may affect health through multiple pathways, both directly, affecting individual choices and economic viability, and indirectly affecting neighborhood-level contextual factors [8]. Furthermore, in our model, we included the effects of endogenous social networks and norms on diet behaviors because of the increasing and strong evidence that diet behavior is influenced by social norms [9], [10], [11].
Our hypotheses are as follows. Hypothesis I: policy effects and social network effects will have independent effects on the diet of both blacks and whites, but will also interact; when social norms are aligned with a healthy diet, the effects of policy will be boosted; when not aligned with a healthy diet, the effects of policy will be reduced.
Hypothesis II: targeting neighborhood school quality will reduce the black–white disparities in diet because neighborhoods in our model with the poorest quality of schools also tended to have the greatest concentration of black residents. However, we do not expect to completely eliminate the black–white disparity because both subpopulations will be affected by changes in neighborhood school quality.
Hypothesis III: the effects of targeting neighborhood school quality will be self-sustaining, even after the policy is no longer in place, because of the propagation of diet behaviors via social network effects.
Section snippets
Description of the model
The population of agents in our model represented the economic and racial distributions of black and non-Hispanic whites in the 100 largest metropolitan statistical areas in the United States. This was accomplished by constructing 64 neighborhoods and matching the race or ethnic and economic distributions of these neighborhoods to empirical data sources [12]. Each neighborhood had 25 households, resulting in 1600 housing units. Please see the online supplemental material for details on these
Experiment 1
Experiment 1a manipulated student-to-teacher ratio when no social network effects were in place. For this experiment, the policy was continuously applied from time step 10 to 110. Figure 2A illustrates the full time series of healthy diet for experiment 1a. Table 2 shows the beginning and ending values of healthy diet for experiment 1a along with the main effects for both policy and period. For blacks, a two-way mixed ANOVA yielded a main effect for policy, F(1,38) = 49.72, P < .001, no main
Discussion
These results reinforce and elaborate on several themes in the literature on determinants of population health and racial disparities in health. First, there exists a potential to leverage upstream macrosocial policies to affect the change in population-level racial disparities in health behaviors [19]. Although our model focused on only one aspect of school quality (student-to-teacher ratio), the results fall in-line with contemporary thought regarding the potential effects of education on
Acknowledgment
This work was supported in part by an Epidemiology Merit Fellowship to M.G.O. from the Mailman School of Public Health, Columbia Univ., grant No. 60466 from the Robert Wood Johnson Foundation to S.G. and G.A.K., a TIIH Scholar Award to G.A.K. from the Institute for Integrative Health, and by a small grant from the Network on Inequalities, Complex Systems, and Health (HHSN276200800013C). The authors would like to thank Nathan Osgood, Ronald Mintz, and Dylan Knowles regarding technical
References (29)
- et al.
Social epidemiology and complex system dynamic modelling as applied to health behaviour and drug use research
Int J Drug Policy
(2009) - et al.
Influence of peers and friends on children's and adolescents' eating and activity behaviors
Physiol Behav
(2012) - et al.
Evaluation of the Healthy Eating Index-2005
J Am Diet Assoc
(2008) - et al.
Availability of healthy foods and dietary patterns: the Multi-Ethnic Study of Atherosclerosis
Am J Clin Nutr
(2009) - et al.
Causal thinking and complex system approaches in epidemiology
Int J Epidemiol
(2010) - Glickman D, Parker L, Sim LJ, Cook H, Miller EA. Accelerating progress in obesity prevention: Solving the weight of the...
- Vanderbroeck P, Goossens J, Clemens M. Tackling obesities: Future choices–building the obesity system map In: (UK)...
- et al.
Prevalence among US adults, 1999-2008
J Am Med Assoc
(2010) What's wrong with social epidemiology, and how can we make it better?
Epidemiological Rev
(2004)- et al.
Are network-based interventions a useful antiobesity strategy? An illustration of simulation models of causal inference in epidemiology
Am J Epidemiol
(2013)
Educational attainment and obesity: a systematic review
Obes Rev
Peer clustering of exercise and eating behaviours among young adults in Sweden: a cross-sectional study of egocentric network data
BMC Public Health
Social norms and dietary behaviors among young adults
Am J Health Behav
Migration, segregation, and the geographic concentration of poverty
Am Sociological Rev
Cited by (33)
Social network-structural position and obesity: Evidence from a national study
2023, Social Science and MedicineDo socio-structural factors moderate the effects of health cognitions on COVID-19 protection behaviours?
2021, Social Science and MedicineComplex Systems Approaches to Diet: A Systematic Review
2019, American Journal of Preventive MedicineCitation Excerpt :This increased exposure, in turn, increased each person's propensity to consume the food. Several of the ABM and SDM studies generated insights about the implementation of one or multiple policy interventions.4–6,9,10,12,14,27–34,36 For example, Orr and colleagues32 found that an intervention to improve the quality of the lowest-performing schools in a community had the potential to reduce disparities in healthy diet between black and white students, and that improvements in diet were greatest when the policy was paired with healthy social norms.
A multivalued agent-based model for the study of noncommunicable diseases
2019, Journal of Biomedical InformaticsAn agent-based simulation of persistent inequalities in health behavior: Understanding the interdependent roles of segregation, clustering, and social influence
2016, SSM - Population HealthCitation Excerpt :In brief, complex systems methods are simulation-based approaches that allow researchers to examine the potential impacts of feedback loops, non-linear effects, adaptation, and other dynamic processes that change over the course of time (Mabry, Marcus, Clark, Leischow, & Méndez, 2010). Health researchers, for example, have used agent-based models (ABMs) to examine adaptive behaviors including the spread of health behaviors through social networks via social influence (Hammond & Ornstein, 2014; Orr, Galea, Riddle, & Kaplan, 2014), adaptation of food stores in response to purchasing patterns in a given neighborhood (Auchincloss, Riolo, Brown, Cook, & Diez Roux, 2011), and the impact of past experiences on future behavior (Yang, Roux, Auchincloss, Rodriguez, & Brown, 2011). These mechanistic models seek to identify and understand features of complex systems (e.g., feedback loops, social influence) that contribute to population-level patterns in health behaviors and outcomes.