Forecasting Obesity and Type 2 Diabetes Incidence and Burden: The ViLA-Obesity Simulation Model

Abstract

Background: Obesity is a major public health problem affecting millions of Americans and is considered one of the most potent risk factors for type 2 diabetes. Assessing future disease burden is important for informing policy-decision making for population health and healthcare.

Objective: The aim of this study was to develop a computer model of a cohort of children born in Los Angeles County to study the life course incidence and trends of obesity and its effect on type 2 diabetes mellitus.

Methods: We built the Virtual Los Angeles cohort-ViLA, an agent-based model calibrated to the population of Los Angeles County. In particular, we developed the ViLA-Obesity model, a simulation suite within our ViLA platform that integrated trends in the causes and consequences of obesity, focusing on diabetes as a key obesity consequence during the life course. Each agent within the model exhibited obesity- and diabetes-related healthy and unhealthy behaviors such as sugar-sweetened beverage consumption, physical activity, fast-food consumption, fresh fruits, and vegetable consumption. In addition, agents could gain or lose weight and develop type 2 diabetes mellitus with a certain probability dependent on the agent's socio-demographics, past behaviors and past weight or type 2 diabetes status. We simulated 98,230 inhabitants from birth to age 65 years, living in 235 neighborhoods.

Results: The age-specific incidence of obesity generally increased from 10 to 30% across the life span with two notable peaks at age 6-12 and 30-39 years, while that of type 2 diabetes mellitus generally increased from <2% at age 18-24 to reach a peak of 25% at age 40-49. The 16-year risks of obesity were 32.1% (95% CI: 31.8%, 32.4%) for children aged 2-17 and 81% (95% CI: 80.8%, 81.3%) for adults aged 18-65. The 48-year risk of type 2 diabetes mellitus was 53.4% (95% CI: 53.1%, 53.7%) for adults aged 18-65.

Conclusion: This ViLA-Obesity model provides an insight into the future burden of obesity and type 2 diabetes mellitus in Los Angeles County, one of the most diverse places in the United States. It serves as a platform for conducting experiments for informing evidence-based policy-making.

Keywords: agent-based model; obesity; prediction; simulation; type 2 diabetes.

Figure 1

Conceptual directed acyclic diagram underlying the data-generating process. SSB, sugar-sweetened beverage consumption; BMI, body mass index; FFV, Fresh fruit and vegetable consumption; T2DM, type 2 diabetes; Ado, Adolescence. T is an index of time. The smaller dotted square represents the neighborhood variables and the larger dotted square represents the individual level variables.

Figure 2

Model initialization diagram of the ViLA-Obesity model. The neighborhood attributes are first initialized at time t=0 and subsequently followed by the initialization of the agent. The neighborhood food and physical activity attributes are predicted as a function of neighborhood socio-demographics. The individual time-invariant variables are set to their baseline values or predicted from socio-demographic variables.

Figure 3

Model execution diagram of the ViLA-Obesity model. BMI, body mass index; T2DM, type 2 diabetes; SES, Socio-economic status. Individual behaviors are predicted based on the agent's socio-demographics, previous behavior and depending on the behavior neighborhood characteristics. The body mass index and type 2 diabetes are predicted based on the agent's socio-demographics, previous behaviors, and previous body mass index in the case of body mass index.

Figure 4

Calibration of the ViLA-Obesity model. This figure depicts the results of the model calibration. It compares the observed (plain lines) to the simulated data (dotted lines).

Figure 5

Obesity and type 2 diabetes prevalence (A), cumulative incidence (B), age-specific incidence proportion (C), and annual incidence rates (D) in the ViLA-Obesity model. The incidence measures were calculated for first-time diagnosis of obesity or type 2 diabetes among at-risk individuals (i.e., without the diagnosis).

Figure 6

Proportion of obesity- and type 2 diabetes-related health behaviors over time in the ViLA-Obesity model. This figure highlights the health behaviors prevalence across the population.