Application of a Bayesian Hierarchical Spatio-temporal Model on Malaria–Typhoid Co-infection in Kenya

A Bayesian hierarchical spatio-temporal model for malaria–typhoid co-infection has beendeveloped and validated using simulation studies in (Kazungu et al., 2026); however, its performance and epidemiological relevance when applied to real-world data have not been assessed. Consequently, its capacity to reveal spatial and temporal heterogeneity and quantify the contribution of climatic factors to malaria–typhoid co-infection dynamics across Kenya remains uncertain. This paper aimed to apply a Bayesian hierarchical spatio-temporalmodel to investigate malaria–typhoid co-infection dynamics and climatic influences across47 counties of Kenya over the period 2020-2025. Parameter estimation was performed withthe integrated nested Laplace approximation (INLA). The adoption of INLA is due to itscomputational efficiency and suitability for Bayesian latent Gaussian models, particularlycomplex generalized linear mixed models (GLMMs), where traditional Markov chain MonteCarlo (MCMC) approaches may require substantially higher computational costs. Temperaturewas identified as the only statistically credible climatic predictor (β1 = 0.015, 95% CrI[0.003, 0.026]), confirming its dominant influence on co-infection risk consistent with thestrong temperature effect observed in the preceding simulation study. The county-level spatialvariation was substantially greater than temporal and interaction effects, highlightingspatial heterogeneity as the main determinant of disease distribution and suggesting thatthe space–time interaction may unnecessarily increase model complexity. The low varianceinflation factor (VIF) values confirmed the absence of multicollinearity among climatic predictors, while Moran’s I statistic demonstrated significant positive spatial autocorrelation.The model achieved good predictive performance with an area under curve(AUC) of 0.877and accuracy of 0.777, indicating reliable discrimination between co-infected and non-coinfected county–year observations. The spatial model risk predictions revealed clear clustering of malaria–typhoid co-infection, with high-risk areas concentrated mainly in western, northwestern, and coastal Kenya, including Turkana, Busia, Migori, Baringo, Uasin Gishu, Trans Nzoia, Kilifi, Kwale, and Mombasa counties. These findings emphasize the importance of climate sensitive, spatially targeted surveillance strategies for integrated malaria and typhoid control in Kenya.