Bayesian Modeling of Geostatistical Survival Data With Misaligned Covariates: A Simulation‐Based Study of Malaria Risk in Gabon

Malaria remains a leading cause of morbidity and mortality in sub‐Saharan Africa, with spatial heterogeneity in transmission posing significant challenges to effective disease control. Accurate estimation of survival risk in relation to local malaria prevalence is often hampered by spatial misalignment, a scenario in which infection covariates and health outcomes are measured at different geographic locations. This study addresses this critical issue by developing a Bayesian hierarchical framework that jointly models malaria infection risk and survival outcomes under spatial misalignment. We employ a two‐stage approach: First, malaria prevalence is modeled geostatistically using a spatial binomial model with Gaussian random effects; second, the predicted infection risk and its associated uncertainty are incorporated into a Bayesian accelerated failure time (AFT) survival model. Through a simulation study calibrated using empirical malaria survey data from Gabon, we evaluated the performance of this joint model under varying degrees of spatial misalignment, censoring, and spatial resolution. Results demonstrate that ignoring spatial misalignment leads to biased effect estimates and poor predictive accuracy, particularly in high‐risk zones. The proposed Bayesian model consistently outperforms competing methods in terms of parameter recovery, credible interval coverage, and spatial prediction of mortality risk. It also proves robust to high levels of right‐censoring and covariate prediction error. By integrating spatial processes and uncertainty propagation, our framework enables more reliable identification of high‐risk regions for targeted intervention, offering a robust tool for malaria risk mapping in settings with sparse or misaligned data.