Disability and Food Insecurity across the Contiguous United States: A Geographically Weighted Regression Analysis

Background and Purpose:
Food insecurity remains a significant public health concern in the United States, disproportionately affecting individuals with disabilities and increasing risk for numerous adverse health outcomes. While prior research has established a positive association between disability and food insecurity, there is limited evidence on how this relationship varies spatially across the United States. Thus, this study aims to examine the geographic variability of the association between disability prevalence and food insecurity across counties in the contiguous United States.

Methods:
Disability prevalance by county was retrieved from the Behavioral Risk Factor Surveillance Survey (BRFSS). Food insecurity rate by county was retrieved from Feeding America’s Map the Meal Gap. Covariates included housing cost burden, smoking prevalence, percentage of adults age 65 and older, and limited access to health food. Covariates were retrieved from BRFFS, the American Community Survey (ACS) and the USDA Food Environment Atlas. The final analytic sample included 3,064 counties across the contiguous United States, representing 99.9% of counties. Economic covariates were intentional excluded as they were previously incorporated into Map the Meal Gap’s food insecurity estimates. Spatial autocorrelation for food insecurity rate and disability prevalence was assessed using Moran’s I. Next, an Ordinary Least Squares (OLS) regression was conducted and significant spatial dependence and non-stationarity as indicated by the Koenker BP statistic (p < 0.001) justified the use of Geographically Weighted Regression (GWR) to estimate local relationships between disability and food insecurity.

Results:
Food insecurity rates ranged from 5.8% to 28.8% (M = 14.4%), while disability prevalence ranged from 17.9% to 54.1% (M = 32.5%). Significant spatial autocorrelation was observed for both food insecurity (Moran’s I = 0.71, p < 0.01) and disability prevalence (Moran’s I = 0.67, p < 0.01), as well as for residuals from the OLS regression (Moran’s I = 0.64, p < 0.01). GWR results indicated substantial spatial variability in the association between disability and food insecurity, with local coefficients ranging from –0.35 to 1.45 (M = 0.38). Strong positive associations were concentrated along the Gulf Coast, the Texas–Mexico border, the North Carolina coast, and parts of the Northern Plains, while non-significant relationships were observed across much of the Midwest and Northeast. The GWR model demonstrated strong overall fit (adjusted R² = 0.85; AIC = 10,943.74).

Conclusions and Implications:
The findings reveal geographic heterogeneity in the relationship between disability prevalence and food insecurity across the contiguous United States. Notably, within states with relatively low overall food insecurity such as North and South Dakota, certain counties exhibited strong associations, suggesting that global and aggregated statistics may mask vulnerabilities among individuals with disabilities. These results underscore the importance of geographically targeted interventions and policies to address food insecurity among individuals with disabilities. Future research should explore more granular geographic scales and include longitudinal analysis to better understand causal pathways and inform equitable resource allocation.