A China‐USA Comparative Study of Peer Influence on Childhood Obesity and Eating Behavior: Social Network Analysis and Agent Based Modeling (ABM)

Hong Xue,Jun Zhang,Youfa Wang

DOI: https://doi.org/10.1096/fasebj.30.1_supplement.900.4

2016-01-01

Abstract:Background Peer influence on health behaviors in children is widely recognized, but the understanding of peer influence on childhood obesity and eating behavior and the between‐country differences is limited. Aims Using innovative social network analysis and agent‐based models (ABM), we examined: 1) peer influence on obesity in children; 2) association/resemblance of eating behavior among classmates; 3) different peer influence/resemblance between China and USA. Methods We built ABMs in a utility maximization framework and parameterized the models based on empirical longitudinal data collected in both countries: 1) Nanjing, China, the Childhood Obesity Prevention through Physical Activity Promotion data (N=5122, 4 th graders, followed during 2013–2014); 2) USA national data, the Early Childhood Longitudinal Study – Kindergarten Cohort (ECLS‐K) (N=760, 5 th graders, followed during 2006–2008). Peer influence was modeled as “follow‐the‐average” social tendency, i.e. children tend to have their BMI close to the average BMI in a given social network they were in (i.e. classroom). Results The combined prevalence of overweight and obesity in the Chinese 4 th graders at baseline was 36%(43% in boys and 28% in girls) based on the IOTF BMI reference. For US 5 th graders, the prevalence of overweight or obesity at baseline was 37% (44% in boys and 29% in girls) based on The 2000 CDC Growth Charts). Our ABM simulated children's BMI growth and fruit and vegetable (FV) consumption in the follow‐up period. We used the Mean Squared Error (MSE) between the simulated data and the empirical data to calibrate and validate our ABM. The agreement between our ABM predicted weight status and weight status based on measured weight and height was high, Kappa was 0.8 and 0.7 for the Chinese children and US children respectively, indicating the validity and high predicting power of the ABMs in both China and the US. The peer influence on children's BMI and weight status was evident in both countries, causing about 0.025 (kg/m 2 /year) BMI increase in US children, in contrast to 0.05 (kg/m 2 /year) BMI increase in Chinese children, at the population level excluding the influence of other contributing factors. However, our simulation suggested high heterogeneity in children's FV consumption behavior in China and the US. The Kappa measuring the agreement of FV consumption patterns between the ABM simulated results and the empirical data was only 0.15 for Chinese children and 0.26 for US children. Our results also indicate network structure and behavior including out degrees, homophily parameters, and transitive triplets are important and would modify peer influence effect on children's BMI changes. Conclusions High obesity prevalence would lead to a continuous increase in children's BMI due to increased socially acceptable mean BMI (as an indicator of social norms). New social interventions such as to foster desirable social norms need be tested and may be potentially effective in obesity prevention. Support or Funding Information Funding: NIH U54 center grant, HD070725.