Hierarchical Spatio-Temporal Modeling of Malaria Incidence in Kenya.

Abstract

Malaria remains a significant public health challenge in Kenya, with substantial spatial and temporal variability in incidence and mortality rates across the country. Despite the growing availability of disease surveillance data, existing analyses often overlook the joint spatial and temporal dependencies that characterize malaria epidemiology. This study addresses this by applying spatial and spatio-temporal Bayesian models to analyze malaria incidence rates in Kenya over the period 2010–2021. Using integrated nested Laplace approximations (INLA) within a hierarchical modeling framework, the study proposed a skewed spatio-temporal (SKEW ST) model and compared its performance to the intrinsic conditional autoregressive (ICAR) and Besag-York-Mollié (BYM) models in modelling malaria incidence rates in Kenya. Model diagnostics were based on WAIC, lppd, and pWAIC which indicated superior performance of the BYM model. Spatial trend analysis revealed significant regional disparities, with high incidence clusters observed in Lake Victoria and coastal regions. The temporal trend demonstrated fluctuating malaria patterns with notable surges in 2015 and 2021. Exceedance probability and relative risk maps further highlighted hotspots with persistent transmission intensity in Northern Kenya, thereby validating spatially structured vulnerability. These findings emphasize the necessity of integrating spatial and temporal patterns in malaria surveillance to guide data-driven policy.