Analysis of Geospatial Variations in Healthcare Across Rural Communities in the US Using Machine Learning

Rural public health is significantly impacted by social drivers of health (SDOH), a set of community-level factors, with rural areas facing challenges such as a higher rate of aging population, fewer jobs, lower income, higher mortality, and poor healthcare access. While much research exists on rurality and SDOH, methodological issues remain, including a narrow definition of SDOH that often overlooks the critical location aspect of healthcare.

This study utilized county-level data from the 2020 Agency of Healthcare Research and Quality SDOH database to investigate geospatial variations in healthcare across the spectrum of rurality. This study employed a set of novel spatial-statistical methods: gradient boosting machines (GBM), Shapley additive explanations (SHAP), and multiscale geographically weighted regression (MGWR).

The analysis of 262 variables across 1976 counties identified 20 key variables related to rural healthcare. These variables were grouped into three categories: health insurance status, access to care, and the volume of standardized Medicare payments. The MGWR model further revealed both global and local effects of specific healthcare characteristics on rurality, demonstrating that geographically varying relationships were strongly associated with socio-geographical factors.

To improve the SDOH in vulnerable rural communities, particularly in Southern states without Medicaid expansion, policymakers must develop and implement equitable and innovative care models to address social determinants of health and access-to-care issues, especially given the potential cuts to public health programs.