The purpose of the project is to study the dynamics of obesity in the United States over the past three decades to build a generic system dynamics model that can be used for obesity policy analysis at multiple levels. In this regard the model will have the following features:
The developed model is multi-level in the sense that it builds on individual level Energy Models for both childhood and adulthood (Butte, Christiansen et al. 2007; Hall 2010) to capture the energy balance and weight change throughout the life of individuals, and aggregates individual level models to population level trends. The model also captures the dependence of RMR on age among adults, a factor that becomes relevant for modeling age-heterogeneous populations or dynamics over long time horizons. Overall, our approach extends the current models to capture variations across individuals, that is critical for population health policy analysis.
In order to estimate relevant model parameters, ideally we require panel data on demographics, body weight and composition, EI, and PA for a large sample of individuals and over a long time. However such a database was not identified, and the most comprehensive study which has data on all the relevant variables for a reliably large sample sizes and also provides a lot of information in terms of the distribution of weights (and body composition) for different subgroups in the population, is the National Health and Nutrition Examination Survey (NHANES). A good population level model should be able to match those distributions closely, and the quality of that match can inform parameter estimation and hypothesis testing. An innovative feature of this study is its methodological contribution towards leveraging estimation methods that can use this cross-sectional data base to estimate dynamic models that span over multiple years.
This approach enables community, state, or national policy analysis building on a calibrated model. The model distinguishes individuals based on sex, age, ethnicity, and socio-economic characteristics and takes physical activity and energy intake as inputs and provides the dynamics of body weight and body composition as outputs. An overview of the model is provided in figure below.
Hazhir Rahmandad, PhD
Grado Department of Industrial and Systems Engineering
Alice Ammerman, DrPH
Department of Nutrition
University of North Carolina at Chapel Hill