GeoAI-Based Analysis of Spatial Associations Between Access to Specialized Healthcare and Premature Mortality From Respiratory Tract Cancer

Premature mortality from respiratory tract cancer (RTC) remains unevenly distributed across territories, reflecting inequalities in access to specialized healthcare services. This study applied a geographic artificial intelligence (GeoAI) framework to investigate spatially varying associations between healthcare accessibility indicators and premature RTC mortality in Southern Brazil. A population-based ecological study was conducted across 1,191 municipalities using mortality data from 2018–2022. Spatial autocorrelation was assessed using Moran’s I and Local Indicators of Spatial Association (LISA). Model performance was compared between Ordinary Least Squares (OLS), Geographically Weighted Regression (GWR), and Geographically Weighted Gradient Boosting Regression (GWRBoost). Spatial block cross-validation was implemented to reduce spatial leakage and overfitting. GWRBoost improved model fit (R² = 0.926) compared with OLS (R² = 0.053) and GWR (R² = 0.473), while reducing residual spatial dependence (Moran’s I = –0.048). These findings highlight the potential of geographically weighted spatial machine learning for geographically targeted cancer control strategies.