Modeling the spatial spread of COVID-19 in Kenya

This study examines the spatial diffusion of COVID-19 across Kenyan counties using gravity based and spatial autoregressive (SAR) models. We model transmission as a one way process originating from Nairobi, which reported Kenya's first confirmed case and serves as the country's Main center of mobility, commerce, and governance. Using county level data on confirmed cases, population, gross domestic product, poverty rates, household count, and access to media, we estimate multiple Linear and SAR regressions to identify structural and spatial determinants of disease burden. By July 2021, the extended gravity model demonstrated strong explanatory power ([Formula: see text]), with distance from Nairobi, number of households, poverty rate, and television access emerging as significant predictors. SAR models indicated minimal spatial autocorrelation after accounting for covariates, suggesting that transmission was primarily centralized around Nairobi. Cluster analysis revealed consistent regional patterns in both socioeconomic vulnerability and COVID-19 prevalence. As a sensitivity analysis, we re-estimated the model using Mombasa as the origin, which produced similar clustering outcomes but yielded different model coefficients, highlighting the importance of hub selection in gravity modeling.