Spatially Informed Public Health Responses to COVID-19 in Asia: Evidence from Advanced Spatial Regression Techniques

Abstract

Understanding the spatial dynamics of COVID-19 is essential for designing effective and equitable public health responses.  However, most existing analyses have relied on traditional regression or basic spatial models, often failing to capture the complex, multi-dimensional nature of infectious disease spread. This study highlights the importance of adopting flexible, data-driven approaches that account for both statistical and spatial heterogeneity in pandemic data. Using data from Asian countries, we examined geographic variation in COVID-19 cases and deaths through a combination of classical, robust, and spatial regression models. Preliminary analysis revealed non- normal distributions, extreme outliers, and significant spatial clustering—underscoring the need for robust, spatially aware modeling. Robust Weighted Least Squares (RWLS) was first applied to handle heteroscedasticity and outliers. The detection of spatial dependence prompted the use of spatial models including the Spatial Lag Model (SLM), Spatial Error Model (SEM), and Geographically Weighted Regression (GWR). While GWR addressed spatial non-stationarity, it remained sensitive to outliers. To overcome this, we employed a Robust Geographically Weighted Regression (RGWR) model that integrates Huber loss and optimal bandwidth selection. RGWR outperformed all other models, explaining 91% and 92% of the variance in case and death rates, respectively, and revealed substantial regional heterogeneity in predictor influence.  Key drivers of COVID-19 outcomes included the proportion of elderly population, air pollution levels, and healthcare infrastructure. By integrating distributional, spatial, and robustness dimensions of the data, this study demonstrates that nuanced, context-sensitive analysis can yield more reliable and actionable insights, supporting better-targeted public health interventions and resource allocation.