Background Malaria is a significant public health problem in Cameroon where spatial and temporal heterogeneities in transmission make early-warning and control complicated. The traditional forecasting methods have the limitation that they are not always able to model both the spatial dependency and the uncertainty; therefore, restricting their functionality. Methods We used a spatio-temporal bayesian model that incorporates a Stochastic Partial Differential Equation (SPDE) that models the spatial correlations and an autoregressive time prior AR(1), solved by Integrated Nested Laplace Approximation (INLA). The analysis of severe malaria cases in 2015–2024 was done separately for children <5 years, adults and pregnant women, and rainfall as a climatic covariate and population adjusted offsets for at-risk individuals. Exceedance probabilities were calculated considering historical risk to identify hotspots. Model performance was evaluated through DIC, WAIC and calibration diagnostics. Results Severe malaria risk was positively related to rainfall and this relationship was observed for all demographic groups. The Far North had long-term high risk foci; secondary risks were noted in the North, Centre and Littoral. The most intense hotspots were observed in pregnant women and children. Conclusions The SPDE-AR(1) model offers a strong and uncertainty-based early warning system for severe malaria in Cameroon, which allows interventions to be targeted in areas and populations at high risk.
Nji et al. (Mon,) studied this question.