Assessing community-level exposure to social vulnerability and isolation: spatial patterning and urban-rural differences

BACKGROUND: Environmental health disparity research involves the use of metrics to assess exposure to community-level vulnerabilities or inequities. While numerous vulnerability indices have been developed, there is no agreement on standardization or appropriate use, they have largely been applied in urban areas, and their interpretation and utility likely vary across different geographies. OBJECTIVE: We evaluated the spatial distribution, variability, and relationships among different metrics of social vulnerability and isolation across urban and rural settings to inform interpretation and selection of metrics for environmental disparity research. METHODS: For all census tracts in North Carolina, we conducted a principal components analysis using 23 socioeconomic/demographic variables from the 2010 United States Census and American Community Survey. We calculated or obtained the neighborhood deprivation index (NDI), residential racial isolation index (RI), educational isolation index (EI), Gini coefficient, and social vulnerability index (SVI). Statistical analyses included Moran’s I for spatial clustering, t-tests for urban-rural differences, Pearson correlation coefficients, and changes in ranking of tracts across metrics. RESULTS: Social vulnerability metrics exhibited clear spatial patterning (Moran’s I ≥0.30, p<0.01). Greater educational isolation and more intense neighborhood deprivation was observed in rural areas and greater racial isolation in urban areas. Single-domain metrics were not highly correlated with each other (rho≤0.36), while composite metrics (i.e., NDI, SVI, principal components analysis) were highly correlated (rho>0.80). Composite metrics were more highly correlated with the racial isolation metric in urban (rho: 0.54–0.64) versus rural tracts (rho: 0.36–0.48). Census tract rankings changed considerably based on which metric was being applied. SIGNIFICANCE: High correlations between composite metrics within urban and rural tracts suggests they could be used interchangeably; single domain metrics cannot. Composite metrics capture different facets of vulnerabilities in urban and rural settings, and these complexities should be examined by researchers applying metrics to areas of diverse urban and rural forms.