Title: Utilizing an Ensemble Markov Chain Monte Carlo Sampler for Bayesian Inference of Nonlinear Malaria Dynamics in Ghana

Abstract

Accurate quantification of malaria trends across sub-Saharan Africa is often impeded by surveillance data that are characterized by short time spans, high noise levels, and significant spatial heterogeneity. In Ghana, health-facility records spanning from 2014 to 2023 demonstrate that hospital admissions exhibit complex, nonlinear, and age-specific variations. However, conventional methods frequently fail to adequately capture this stochastic variability or establish reliable uncertainty intervals. To address these limitations, this research introduces a Bayesian nonlinear inference framework. This approach combines a cubic baseline function with a damped oscillatory kernel, utilizing an affine-invariant ensemble Markov Chain Monte Carlo (MCMC) sampler for estimation.

The proposed framework is designed to handle sparse datasets, explicitly model parameter uncertainty, and deliver probabilistic projections for two distinct demographic groups: children under five and individuals aged five and older. The model demonstrates robust empirical fit, achieving an $R^2$ value of 0.9958. Notably, peripheral districts such as Mpohor and Bia East reported 3.3 cases. Looking ahead, projections for the period 2024–2026 suggest a gradual resurgence in malaria incidence. Specifically, cases among children under five are estimated to rise from 137,000 to 149,000, while cases among older individuals are projected to increase from 348,000 to 375,000. As the forecast horizon extends, the associated uncertainty margins also widen. By generating these probabilistic forecasts, this Bayesian methodology offers a rigorous instrument for predicting malaria fluctuations, thereby supporting evidence-based decision-making within Ghana’s national malaria control initiatives.

Source: arXiv Generated at: 2026-06-02 00:00:00 UTC