Bayesian Inference of Nonlinear Malaria Dynamics in Ghana via an Ensemble Markov Chain Monte Carlo Sampler Journal: Artificial Intelligence In Medicine
T. Ansah-Narh, Y. Asare Afrane, J. Bremang Tandoh · 2026
Read Bayesian Inference of Nonlinear Malaria Dynamics in Ghana via an Ensemble Markov Chain Monte Carlo Sampler Journal: Artificial Intelligence In Medicine by T. Ansah-Narh, Y. Asare Afrane, J. Bremang Tandoh (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-00783.txt.
Introduction
Ansah-Narh and co-authors model malaria in Ghana with care for doubt in this 2026 paper. Health records there are short, noisy, and uneven across districts. Hospital data from 2014 to 2023 show wavy, age-split patterns in admissions. Past methods miss the random swing and give weak bounds on their own error.
The team builds a Bayesian frame with a cubic baseline and a damped wavy kernel. An ensemble Markov chain Monte Carlo sampler fits it. Fits are strong: R-squared values of 0.9958 for children under five and 0.9956 for older groups. Errors stay below 2 percent, and the chains mix well. District gaps are stark, from calm urban Kumasi to wild swings in Mpohor and Bia East. Forecasts for 2024 to 2026 show a slow rise in cases. Under-five cases grow from 137,000 to 149,000, with doubt widening over time.
Catalog record
| CREATOR | T. Ansah-Narh, Y. Asare Afrane, J. Bremang Tandoh |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 72 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.00783
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Bayesian Inference of Nonlinear Malaria Dynamics in Ghana via an Ensemble Markov Chain Monte Carlo Sampler Journal: Artificial Intelligence In Medicine T. Ansah-Narh, Y. Asare Afrane, J. Bremang Tandoh Abstract Reliable quantification of malaria dynamics in sub-Saharan Africa remains hindered by short, noisy, and spatially heterogeneous surveillance records that challenge the assumptions of conventional deterministic models. In Ghana, health-facility data between 2014 and 2023 reveal highly non-linear and age-specific fluctuations in hospital admissions, yet existing approaches struggle to capture such stochastic variability or provide credible uncertainty bounds. This study develops a Bayesian nonlinear inference framework that integrates a cubic deterministic baseline with a damped oscillatory kernel, estimated via an affine-invariant ensemble Markov Chain Monte Carlo sampler. The framework accommodates limited data, explicitly models parameter uncertainty, and generates probabilistic forecasts of malaria admissions for children under five years and individuals aged five years or more. Results demonstrate that the proposed hybrid cubic–damped oscillatory kernel model achieves strong empirical adequacy ( R 2 = 0.9958 R^{2}=0.9958 for < 5 <5 years; R 2 = 0.9956 R^{2}=0.9956 for ≥ 5 \geq 5 years) with residual errors below 2 % 2\% and unimodal, well-mixed posterior distributions confirming robust convergence. District-level analysis reveals pronounced spatial heterogeneity, with coefficients of variation ranging from < 0.07 <0.07 in stable urban centres such as Kumasi to > 3.3 >3.3 in peripheral districts including Mpohor and Bia East. Forecasts for 2024–2026 indicate a gradual resurgence in admissions, increasing from approximately 137,000 to 149,000 cases among children under five and from 348,000 to 375,000 among older individuals, with uncertainty widening modestly over time. By producing interpretable probabilistic forecasts, this Bayesian framework provides a principled tool for anticipating short-term malaria fluctuations, guiding resource allocation, and strengthening data-driven decision-making within Ghana’s national malaria control strategy. † † highlights: Bayesian model for Ghana’s malaria admissions from limited data. Hybrid kernel captures smooth trends and transient fluctuations. Used an affine-invariant MCMC sampler for efficient analysis. Malaria admissions fell 35\% in under-5s from 2014–2023. Spatial variation was extreme: under 0.07 in cities to over 3.3 in rural areas. 1 Introduction Malaria remains a dominant cause of morbidity across sub-Saharan Africa and continues to impose a significant public-health and economic burden, despite decades of targeted interventions. The World Health Organization estimates that in 2022 there were over 249 million malaria cases globally, resulting in more than 600,000 deaths, with the majority occurring in African countries ( 43 ) . In Ghana, malaria is hyperendemic and transmission persists throughout the year, though it peaks during the rainy seasons. The disease accounts for a substantial proportion of hospital outpatient visits and inpatient admissions, particularly among children under five years of age and pregnant women ( 23 ) . While the implementation of vector control measures, improved diagnostics, and treatment campaigns has contributed to a gradual reduction in prevalence, national surveillance data indicate marked inter-annual and regional variability in malaria admissions. These persistent fluctuations underscore the complex, nonlinear, and spatially heterogeneous nature of malaria dynamics in Ghana. The reliable quantification and forecasting of malaria admissions are essential for informed planning of healthcare resources and preventive interventions. However, the available datasets used for such assessments are typically limited in temporal resolution and spatial detail. Most health-facility records and national surveillance data provide aggregated annual or monthly summaries, often spanning only a decade, as is the case for Ghana’s 2014--2023 malaria admission data. Such short or moderately sized time series constrain the applicability of classical statistical and machine-learning models. Linear and autoregressive frameworks (such as ARIMA 1 1 1 Autoregressive Integrated Moving Average. ) require assumptions of stationarity and consistent seasonal cycles, which are rarely satisfied in malaria datasets characterised by irregular peaks, long-tailed fluctuations, and structural breaks caused by climatic variability or intervention rollouts ( 32 ) . On the other hand, data-intensive machine-learning models, though flexible, demand larger training samples to achieve stability and generalisability ( 34 ) . In low-data regimes, they tend to overfit, yielding implausible extrapolations and inflated uncertainty. Hospital-based malaria admission data also present intrinsic representativeness challenges. Admission counts reflect both epidemiological transmission and contextual factors such as access to healthcare, reporting completeness, and facility catchment dynamics. Consequently, raw admission trends can obscure true transmission intensity or exaggerate local fluctuations due to variations in service utilisation ( 5 ) . Spatial delineation of health-facility catchments and adjustment for reporting biases have been shown to alter the perceived geography of malaria burden ( 18 ) . Ignoring such spatial heterogeneity risks misinterpreting temporal patterns as epidemiological trends when they may instead reflect reporting artefacts. A joint temporal–spatial approach that separates the temporal signal from the spatial variability, therefore becomes crucial to reveal underlying dynamics and to identify consistent high-burden or highly variable regions. Malaria transmission dynamics are also shaped by multiple nonlinear processes. These include rainfall-driven mosquito breeding cycles, temperature-dependent parasite development, and socio-behavioural drivers such as preventive bed-net use or treatment uptake ( 16 ) . Such interactions produce epidemic curves with abrupt inflexion points, plateaux, and intermittent surges that cannot be adequately captured by simple linear regressions or fixed-seasonal autoregressive models. Even where covariates are available, parametric forms may fail to express complex phase relationships among drivers. A modelling framework that allows flexible curvature, accommodates stochastic variability, and directly quantifies parameter uncertainty is therefore needed to provide credible inference under these conditions. Bayesian inference provides a framework that combines prior information with observed data, allowing Bayesian models to effectively handle the uncertainty associated with small samples and reflect it within predictive intervals. This approach is particularly valuable for health surveillance systems with limited temporal depth or incomplete spatial coverage ( 44 ; 19 ; 15 ) . A Bayesian nonlinear structure, such as a hybrid cubic component augmented by a bounded nonlinear term, can capture smooth long-term changes while flexibly representing transient anomalies. Posterior sampling techniques, such as ensemble Markov chain Monte Carlo (MCMC), allow efficient exploration of parameter space and generation of full posterior distributions for key epidemiological quantities, rather than point estimates alone. This probabilistic perspective aligns with public-health needs for decision-making under uncertainty, where credible intervals provide more informative guidance than single deterministic forecasts. To avoid ambiguity regarding scope, the modelling framework adopted in this study is deliberately parsimonious, reflecting the limited temporal depth of routine malaria surveillance data in Ghana. The emphasis is placed on Bayesian inference, uncertainty quantification, and epidemiological interpretability, rather than on maximising functional complexity. This design choice ensures stable estimation and credible probabilistic forecasting in a low-sample regime typical of national health-facility datasets. In this study, we develop and apply a Bayesian nonlinear modelling framework to characterise malaria admission dynamics in Ghana over the period 2014–2023. The approach is designed to handle limited data while retaining interpretability and statistical robustness. We model separate temporal trajectories for children under five years and individuals aged five years or older to capture demographic contrasts in burden and variability. Posterior inference is used to evaluate parameter uncertainty and to forecast near-future admissions for 2024–2026. The resulting posterior predictive distributions provide a measure of forecast reliability and enable probabilistic evaluation of potential increases or decreases in hospital admissions. Beyond temporal inference, the study also examines spatial heterogeneity in malaria admission variability.