Abstract
Although the malaria burden has been significantly decreased over the last two decades, it is still a major public health problem in sub-Saharan Africa, such as Malawi. Transmission dynamics have evolved to be more heterogeneous with micro-epidemics and peaks. The existing surveillance infrastructure is predominantly reactive and has little capability for early detection and response. Early warning systems (EWS)present an opportunity to strengthen preparedness and support timely, data-driven action, yet they remain underutilised in many endemic settings. This thesis, comprising five interrelated published papers, aims to improve malaria prediction and early warning capabilities.The first paper is an analysis of the malaria prevalence trends in Malawi from 2000 to 2022. Data from household surveys and model-based geostatistical methods were utilised to generate fine-scale subnational risk maps. The outputs, which included fine-scale maps of malaria burden across different geographic areas, offered an empirical basis for stratification to guide targeted interventions and prioritisation of malaria control efforts.
The second paper, using geostatistical analysis, explores subnational variations in malaria prevalence across Kenya. In this paper, we show the transmission hotspots and spatial heterogeneity at a finer scale.
The third paper details the development of the Malawi national malaria burden stratification framework. The framework helps pinpoint where interventions, such as EWS interventions, are most likely to help. Using estimates based on the prevalence of malaria in children aged 2 to 10 years, historical incidence, mortality, and expert opinion on malaria, districts are classified by transmission intensity and programmatic requirements.
The fourth paper synthesises global evidence on malaria EWS, reviewing their implementation settings, methodological designs, performance assessment approaches, and operational challenges. The review brings out the key limitations of existing EWS, including limited integration into national surveillance platforms, lack of clearly defined outbreak thresholds, and minimal involvement of end-users in the design and application of the systems.
The fifth paper evaluates and compares the predictive performance of Poisson temporal and Poisson spatial-temporal models for malaria prediction and forecasting in seven moderate-burden districts in Malawi, using routine surveillance data from 2015 to 2023 and climate covariates. Both models incorporated climate covariates such as rainfall and temperature and were evaluated using root mean square error estimates on 2023 hold-out data, followed by 12-month forecasts for 2024. Performance varied across districts and models.
Collectively, these studies demonstrate how geostatistical modelling, burden stratification, evidence synthesis, and predictive analytics can inform the design and implementation of fit-for-purpose EWS. This PhD contributes a practical roadmap for embedding predictive tools into national surveillance systems, aligned with the strategic goal to eliminate malaria.
| Date of Award | 11 Nov 2025 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Anja Terlouw (Supervisor), Feiko Ter Kuile (Supervisor) & Emanuele Giorgi (Supervisor) |