Dynamic Empirically Based Model for Understanding Future Trends in US Obesity Prevalence in the Context of Social Influences

Abstract

Objective: This study aimed to (1) identify mechanistic model structures that produced quality fit to historic obesity prevalence trends and (2) evaluate the sensitivity of future obesity prevalence to social transmission and nonsocial parameters.

Methods: An age- and gender-structured compartmental model was used to describe transitions between weight status groups. Four model structures with different combinations of social transmission and nonsocial mechanisms were calibrated to match historic time series and assessed for quality of fit. Projections of overall obesity prevalence to 2052 were simulated, and sensitivity analyses were conducted.

Results: The model structure that included only nonsocial mechanisms indicated that the overall obesity prevalence in the United States has already stabilized and will increase little more; however, it underestimated observed obesity prevalence since 2013. If social transmission mechanisms influence obesity, the model estimated continued increases in obesity prevalence, reaching 48.0% to 55.1% by 2050. Obesity prevalence was most sensitive to changes in the adult social transmission parameters, especially among women.

Conclusions: The model projected that US obesity prevalence in the overall population will likely continue to increase for decades. The findings that obesity prevalence was most sensitive to adult parameters can be used to inform conversations about priorities for public health and health care programs and policies.

Figure 1.. Diagram describing age- and gender-structured compartments and flow of the model

Note: Death rates were also applied to each compartment

Figure 2.. Dynamic prediction of calibrated models and comparison to historic data - child and adolescent obesity prevalence

Notes: The figure shows the outcome of calibrations and projected future trends with four different model structures as follows: (1) non-social transition rates only, (2) social transmission rates only, (3) both non-social transition and social transmission rates, and (4) the third model structure with an additional population-level social transmission factor in which population-wide prevalence of obesity also influenced normal to overweight and overweight to obese transitions across all age groups (Pop-ST). The different styles and colors of lines represent the model calibration using different structures, as indicated in the figure. A linear trend line is also provided for reference. The yellow shaded portion represents confidence bounds of the model prediction using the parameter sets identified with GLUE in the third calibration (non-social and social transmission rates). A yellow line is drawn for each of 100 parameter sets that provided acceptable fit. The black diamonds represent point estimates from NHANES historical data along with their 95% confidence intervals.

Figure 3.. Dynamic prediction of calibrated models and comparison to historic data - adult and middle-age adult obesity prevalence

Notes: The figure shows the outcome of calibrations and projected future trends with four different model structures as follows: (1) non-social transition rates only, (2) social transmission rates only, (3) both non-social transition and social transmission rates, and (4) the third model structure with an additional population-level social transmission factor in which population-wide prevalence of obesity also influenced normal to overweight and overweight to obese transitions across all age groups (Pop-ST). The different styles and colors of lines represent the model calibration using different structures, as indicated in the figure. A linear trend line is also provided for reference. The yellow shaded portion represents confidence bounds of the model prediction using the parameter sets identified with GLUE in the third calibration (non-social and social transmission rates). A yellow line is drawn for each of 100 parameter sets that provided acceptable fit. The black diamonds represent point estimates from NHANES historical data along with their 95% confidence intervals.

Figure 4.. Box plots of the calibrated parameter sets from the best-fit model structure (combination of social and non-social parameters)

Note: The box plots represent the range of values identified from the GLUE procedure for all calibrated parameter sets that produced acceptable behavior (when compared to historic data).

Figure 5.. Dynamic model prediction of prevalence of obesity among US adults and children

Notes: The figure shows the outcome of calibrations and projected future trends with four different model structures as follows: (1) non-social transition rates only, (2) social transmission rates only, (3) both non-social transition and social transmission rates, and (4) the third model structure with an additional population-level social transmission factor in which population-wide prevalence of obesity also influenced normal to overweight and overweight to obese transitions across all age groups (Pop-ST). The different styles and colors of lines represent the model calibration using different structures, as indicated in the figure. A linear trend line is also provided for reference. The yellow shaded portion represents confidence bounds of the model prediction using the parameter sets identified with GLUE in the third calibration (non-social and social transmission rates). A yellow line is drawn for each of 100 parameter sets that provided acceptable fit. The black diamonds represent point estimates from NHANES historical data along with their 95% confidence intervals.

Figure 6.. Sensitivity of overall obesity prevalence in 2052 to calibrated social transmission and non-social transition parameters

Note: The figure visualizes the results from the one-way sensitivity analyses, which were conducted over the base-case calibrated values of the social transmission and non-social transition parameters. We allowed each value to range across a full range of potential values (0 to 1.0). The 12 parameters that reduced overall obesity the most by 2052 are shown.