Delineating Health Enumeration Areas in Florida

Abstract

Consideration of the spatial dimensions of health is often necessary for small area health investigations. This study introduces a novel approach to neighborhood delineation by designing geographic units defined by similarity in both behavioral and social determinants of health alongside criteria to ensure the robustness of health metrics (≥ 1000 individuals, ≥ 50 deaths). To create the Health Enumeration Areas (HEA), we first aggregate individual death records to census blocks and calculate drive time from each block's centroid to the nearest hospital. The number of deaths and decedents' smoking status and drive time to nearest hospital serve as constraints in the enriched constraints Max-P algorithm (EMP) regionalization technique. Across Florida counties, HEA generally outperformed census tracts, as evidenced by higher path silhouette scores associated with decedent tobacco use patterns and reduced within-unit heterogeneity. HEA also reduced unstable life expectancy estimates from ∼14% to ∼4% and accounted for a larger share of the population in those estimates, with improvements ranging from 20% to over 70% in many counties. The algorithm produced geographic units, similar in size to census tracts, that prioritized homogeneity and more accurately represents the conditions impacting health by addressing the uncertainty inherent in traditional areal units.