Evaluating the interface of health data and policy: Applications of geospatial analysis to county-level national data

ABSTRACT

Introduction: The objective of this research was to spatially analyze linked health data for geographic trends in factors impacting children’s health. Traditional linear regression analyses of county-level data tend to inflate R². Spatial regression represents a robust approach for improved analysis of geographic data. Methods: We used GeoDa 1.6.0 to regress 3,221 U.S. county-level child health outcomes (e.g., infant mortality, child mortality) on independent variables (e.g., low birth weight, percent race/ethnicity, uninsured, emotional support). Statistical analyses included spatial R², Moran’s I, and multicollinearity measures. The data source was the 2014 County Health Rankings. Results: Three spatial regression models (health, socioeconomic, and combined) were compared for infant and child mortality. The combined model for infant mortality rate yielded the largest adjusted R² = 0.428 (F = 110.9, p < 0.001), similarly for child mortality rate R² = 0.411 (F = 94.3, p < 0.001). The strongest predictors in both models were obesity, smoking, teen birth rate, severe housing problems, no social supports, and urbanicity. Discussion: The results demonstrate correlations between county-level conditions and child health outcomes, supporting previous research linking poor health/education and low socioeconomic conditions. Geospatial information can assist policymakers to apply health education interventions.