Multi-level Analysis of the Impact of Community Factors on Geographic Disparities in Diabetes and Obesity Nationwide and among U.S. Veterans, 2008-2017

American Public Health Association (APHA) Annual Meeting, November 10-14, 2018, San Diego, California

Priscilla Lopez, MPH, David Lee, MD, Pasquale Rummo, PhD, MA, Stephanie Rosoff, MS, Andrew Vinson, MS, GISP, Mark Schwartz, MD, Sanja Avramovic, PhD, Farrokh Alemi, PhD, Andrea Troxel, Phd, Ingrid Gould Ellen, PhD, MA, Brian Elbel, PhD, MPH, Lorna E. Thorpe, PhD, MPH and Rania Kanchi, MPH 

BACKGROUND: While most chronic disease prevention efforts have focused on addressing individual-level risk factors, a growing body of research suggests that modifying social and environmental characteristics may reduce the risk of developing chronic disease and mitigate health disparities, particularly with respect to diabetes and obesity. A number of studies have shown that community-level factors are associated with obesity, including neighborhood sociodemographic composition, walkability, the food environment, community safety measures and more. Evidence to date is sparser regarding the impact of the food environment and other social determinants on diabetes risk. The current literature fails to address the fact that geographic distributions of high diabetes and high obesity prevalence do not always overlap, which suggests discordance in risk profiles of the two diseases.

OBJECTIVES: We aim to use county-level prevalence data from the Behavioral Risk Factor Surveillance System (BRFSS), American Community Survey (ACS), and individual-level electronic health record (EHR) data from a national Veteran’s Administration (VA) cohort (~3.8 million patients diabetes-free at baseline as of January 2008) to examine the relationship between community-level food and housing environment determinants and three outcomes: diabetes, obesity and discordance between diabetes and obesity prevalence.

METHODS: In initial analyses, we constructed meaningful measures of the food and housing environment and calculated change in diabetes and obesity prevalence between 2004 and 2013 using BRFSS data. For the food environment, we will be examining food swamps (relative measure of the percentage of less healthy food outlets to total food outlets) using county-level retail data; and our primary housing affordability measure is residual income (household’s gross income –gross housing costs) adjusted for regional cost-of-living using ACS data. In future analyses, we will use these metrics to predict change in community-level diabetes and obesity prevalence nationwide; and examine the impact these metrics on diabetes incidence among U.S. Veterans using EHR data.

RESULTS: In longitudinal nationwide analysis of BRFSS data 2004-2013, we found that on average, county-level diabetes prevalence has increased 1.27 fold, and county-level obesity prevalence has increased 1.23 fold. However, the correlation between these relative changes in diabetes and obesity prevalence was only 0.37 as the rate of increases in diabetes and obesity did not always match in magnitude. The relative increases in diabetes were outpacing relative increases in obesity within two clusters; one over Alabama, Georgia and Tennessee, and the other over Indiana, Illinois and Kentucky. Analyses to quantify the impacts of the food and housing environment on change in the burden of disease over time are ongoing.

DISCUSSION: Within the Diabetes LEAD Network, the NYU team offers innovative insight into novel food and housing environment measures and the application of multi-level approaches to a national cohort of VA patients with high incidence of diabetes