A Spatial Analysis of Obesity and Its Associations with the Built and Natural Environment, Physical Inactivity, and Socioeconomic and Demographic Conditions in the United States of America

Abstract

Obesity has strong genetic determinants but the genetic composition of the population does not change rapidly. Thus in this study, the major changes in non-genetic factors such as the development of obesogenic environments and shifting socioeconomic status and lifestyle of individuals are hypothesized to increase the risk of obesity. As the prevalence of obesity continue to increase worldwide with substantial attention in the US, a clearer understanding of how spatial associations between obesity and confounding factors are interrelated is crucial to better tackle the issue of obesity. This study employs the ‘global’ and ‘local’ Exploratory Spatial Data Analysis (ESDA) methods including the Ordinary Least Squares (OLS) regression and Geographically Weighted Regression (GWR) to investigate obesity and its spatial associations with environmental, behavioral, socioeconomic, sociodemographic, and population based dynamics at the county level. The results from this study have generated empirically-based and useful insights for the 3,105 counties and county-equivalents across the 48 contiguous states, also known as the continental US. A major contribution of this study is exploring obesity and its confounding associations with various factors not only spatially but also temporally for the first time, revealing the temporal changes from 2004 to 2007 and to 2010. By utilizing the ESDA methods, a consistent answer obtained significantly indicates that positive spatial associations exist between obesity and physical inactivity (PIA), poverty, and population-weighted distance (PWD) to parks. Conversely, negative spatial associations exist between obesity and ratio of jobs to employed residents (JER) and population density. Another major contribution of this study is examining and revealing geographic variability in the association between obesity rates and various explanatory variables both nationwide and regionally at the county level for the entire US. By utilizing the GWR, a significant spatial nonstationarity is identified. This finding suggests that the strength of associations between obesity and each of the explanatory variables vary depending on the spatial location. It is also revealed that the confounding variables PIA, high educational attainment, African-American population, and poverty are identified as the top four variables by having relatively stronger effects in explaining obesity rates at the county level both nationwide and regionally.