Assessing Measurement Invariance of a Land Use Environment Construct Across Levels of Urbanicity

Variation in the land use environment (LUE) impacts the continuum of walkability to car dependency, which has been shown to have effects on health outcomes. Existing objective measures of the LUE do not consider whether the measurement of the construct varies across different types of communities along the rural/urban spectrum. To help meet the goals of the Diabetes Location, Environmental Attributes, and Disparities (LEAD) Network, we developed a national, census tract-level LUE measure which evaluates the road network and land development. We tested for measurement invariance by LEAD community type (higher density urban, lower density urban, suburban/small town, and rural) using multiple group confirmatory factor analysis. We determined that metric invariance does not exist; thus, measurement of the LUE does vary across community type with average block length, average block size, and percent developed land driving most shared variability in rural tracts and with intersection density, street connectivity, household density, and commercial establishment density driving most shared variability in higher density urban tracts. As a result, epidemiologic studies need to consider community type when assessing the LUE to minimize place-based confounding. Plain Language Summary A community's land use environment (LUE) describes its citizens ability to walk from place to place versus their reliance on vehicles for transportation. Some factors that might influence LUE include the street network, size of road blocks, density of walkable establishments (food stores, restaurants, schools, etc.), and the mixture of residential and commercial establishments. Existing objective measures of the LUE do not consider urbanicity, which can lead to differences due only to places being more or less rural. The Diabetes Location, Environmental Attributes, and Disparities (LEAD) Network created a national, census tract-level LUE measure and assessed whether the LUE construct differed across LEAD community type groups (higher density urban, lower density urban, suburban/small town, and rural). We found that the LUE construct does vary across LEAD community type. Future epidemiologic studies examining the LUE need to consider urbanicity to account for these differences.