In 2017, the World Health Organization recognized obesity as an epidemic and an important public health issue; thus, our goal was to look at obesity prevalence trends in the United States over time to understand how it may progress in the future. In this article, obesity prevalence was computed using data published by the Centers for Disease Control and Prevention (CDC) between the years 1990 and 2019. We studied the obesity prevalence time series for all fifty states within the United States and the District of Columbia. The states were then clustered using the nonparametric dynamic time warping (DTW) approach to identify significant regional differences in the development of obesity prevalence. To differentiate our study from the CDC, we focused on clustering obesity prevalence both spatially and temporally. This helped us understand obesity patterns throughout the states over time, rather than focusing on one year at a time. DTW helped us define six heterogeneous regions of obesity prevalence rates in the United States. We observed that the states grouped in the same region exhibit similarities in temporal dynamics measured by relationships in their obesity prevalence time series. While all six regional time series exhibit an upward trend, there are distinct differences in how their obesity prevalence patterns behave over time. For policy making, leaders must pay attention to regional commonalities, as we observed that states in each region share some similar characteristics regardless of their geographic proximity. Once we determined regions, we implemented Autoregressive Integrated Moving Average models to estimate obesity prevalence and develop future forecasts through the year 2022 for each region. Through the discoveries and forecasts made in our research, we hope to better understand obesity prevalence disparities across regions in the United States in past, current, and future times.
Authors: Sierra Hessinger, Rebekah Poth
Advisor: Tatjana Miljkovic, Statistics
Graduate Advisor: Katie Vorpe, Mathematics
You must be logged in to post a comment.