Geospatial Modelling of COVID-19 Outcomes: Insights from Composite Indicators for Pandemic Preparedness

Abstract

Effective pandemic preparedness requires a comprehensive understanding of the spatial distribution of the outbreak and its local driving factors. We applied geographically weighted random forest (GWRF) to model the spatial dynamics of COVID-19 incidence and mortality at the county level across the United States. Unlike most studies that used standalone factors, this study integrated multiple standalone factors into a single score. We used robust geographically weighted principal component analysis (GWPCA) to construct nine composite indicators across several domains, including epidemiological, socio-economical (adversity and prosperity), environmental, and vaccination indicators. Our findings indicate that the epidemiological indicator was the most important contributor to the dynamics of COVID-19 incidence and mortality, with significant regional disparities observed. Counties in the West and South exhibit higher risks, while those in the Midwest and Northeast show comparatively lower risks. Interestingly, lower vaccination vulnerability does not necessarily correlate with lower COVID-19 mortality, highlighting the complexity of pandemic risk dynamics. By incorporating spatially adaptive composite indicators, this study provides enhanced insights for public health planning and intervention strategies. The integration of spatial analytics with epidemiological modelling offers a scalable framework for regionalised pandemic response, ensuring more equitable resource allocation and targeted mitigation efforts in future health crises.

Keywords: Spatial modelling, composite indicators, pandemic preparedness, United States