A descriptive spatial analysis of multiple health-related resources in Baltimore, New York City, and Winston-Salem

Background: Resource access may contribute to widening health disparities. Research has established that some resources, like supermarkets, are less prevalent in minority areas. There is little empirical evidence using spatial analytic techniques on the patterning of multiple health-related resources. Methods: We geocoded locations of supermarkets, recreational facilities, parks, and retail areas in New York City (819 block groups), Baltimore (737), and Forsyth County, NC (169). We constructed 1/2 mile kernel densities of resources as indicators of locational access to healthy foods, recreational resources, and land-use mix. Densities were analyzed individually and collectively as a summary score. Correlation coefficients and Moran's I statistics quantified statistical and spatial resource clustering. Spatial regression models were fit to estimate the relationship between block group race and resource access, while accounting for spatial correlation. Results: After adjusting for population density, access to different resources was positively correlated (0.09 to 0.62, p<.05 for all) except for access to parks which was not correlated with other resources. Spatial autocorrelation varied across resource and site (Moran's I from 0.08 to 0.95) and was highest in New York. Spatial regression models showed significant negative relationships between percent minority and resource access. In Baltimore, the relationship was more significant for the summary score (-10.2, p<.0001) than any one individual resource. Discussion: There is evidence of clustering in locational access to resources, suggesting that some areas are deficient in multiple resources. The pattern of clustering is dependent on area race/ethnicity and the metropolitan area studied.