Smart density: A more accurate method of measuring rural residential density for health-related research.

Background: Studies involving the built environment have typically relied on US Census data to measure residential density. However, census geographic units are often unsuited to health-related research, especially in rural areas where development is clustered and discontinuous.

Objective: We evaluated the accuracy of both standard census methods and alternative GIS-based methods to measure rural density.

Methods: We compared residential density (units/acre) in 335 Vermont school neighborhoods using conventional census geographic units (tract, block group and block) with two GIS buffer measures: a 1-kilometer (km) circle around the school and a 1-km circle intersected with a 100-meter (m) road-network buffer. The accuracy of each method was validated against the actual residential density for each neighborhood based on the Vermont e911 database, which provides an exact geo-location for all residential structures in the state.

Results: Standard census measures underestimate residential density in rural areas. In addition, the degree of error is inconsistent so even the relative rank of neighborhood densities varies across census measures. Census measures explain only 61% to 66% of the variation in actual residential density. In contrast, GIS buffer measures explain approximately 90% of the variation. Combining a 1-km circle with a road-network buffer provides the closest approximation of actual residential density.

Conclusion: Residential density based on census units can mask clusters of development in rural areas and distort associations between residential density and health-related behaviors and outcomes. GIS-defined buffers, including a 1-km circle and a road-network buffer, can be used in conjunction with census data to obtain a more accurate measure of residential density.