Projecting Malaria Risk in Africa Accurately and Sustainably
While Ethiopia has made progress in reducing malaria risk, the country’s malaria cases surged from 4.1 million in 2023 to 7.3 million in 2024. This increase was fueled by several factors, including weather variability and environmental conditions, civil conflict, and growing resistance to insecticides and treatments.
This approach ensures consistency and comparability while honoring local context, lessons learned, and country ownership. Our work supports the creation of more accurate malaria projection models and strengthens national capacity to sustain and use malaria early warning systems.
Learn more about our experience designing, implementing, and evaluating research initiatives that serve diverse populations across global contexts.
Malaria, an infectious disease spread by mosquitos, is sometimes life-threatening but is also preventable with targeted interventions, such as insecticide treated bed nets. In 2024, 95% of malaria cases and 95% of malaria deaths were in the World Health Organization (WHO) African region, according to the WHO. About 76% of deaths from malaria in the region that year were among children younger than age 5.
Many African countries are vulnerable to similar challenges as Ethiopia. How can governments, funders, and partners act now to prevent malaria health crises? Early warning systems can contribute to a solution.
Early warning systems may help decisionmakers and relief agencies plan for and respond to malaria risk, like they do for other humanitarian and public health crises. For instance, the Famine Early Warning Systems Network (FEWS NET) uses climate, market, and livelihood data to forecast acute food insecurity. (AIR manages FEWS NET’s Data, Learning, and Communications Hub.)
Sustainable, high-quality malaria risk prediction models are becoming even more critical. Governments in Africa and other low- and middle-income countries face challenges in health financing because of reduced funding from USAID and other donor agencies. Allocating resources to areas with a higher risk of malaria can protect thousands of lives. Models for predicting malaria, while initially costly, can serve as a cost-effective solution to improving health outcomes, especially in current contexts with limited resources.
Malaria Risk Varies by Location
Developing accurate predictions of malaria incidence requires the development of models and early warning systems that incorporate country-specific and contextual characteristics. Around Africa, malaria transmission varies both within and across countries. An intervention that is effective in one area may not work as well in another.
Optimizing the cost-effectiveness and sustainability of models to predict malaria incidence requires a balance among precision, adaptability, and usability.
In Kenya, for instance, malaria transmission is seasonal and concentrated in certain regions of the country. There, insecticide-treated bed nets are one of the most effective interventions for reducing malaria. While insecticide-treated bed nets have led to tremendous progress, their impact is affected by inconsistent use; not everyone who owns a net uses it, meaning that the overall effectiveness of bed net programs depends on their adoption. Compounding this challenge, mosquitos have developed resistance to the primary insecticide used to treat bed nets in some parts of Kenya.
Beyond different transmission and risk patterns, there are other country-specific variations. For instance, conflict or climate-related shocks can cause disruptions to countries’ health systems, changes in how much an intervention is used, and mass population displacement—all of which should be accounted for in malaria risk prediction models.
Countries differ in their starting conditions, health systems, and the types of malaria programs they use. Because of these differences, a single global model may not predict malaria risk very well for every place. Models that are built specifically for each country—while still fitting into a larger global framework—can give more accurate and useful estimates.
New advances in artificial intelligence (AI) offer even more ways to tailor these models. For example, AI tools that analyze news stories and other media have been used to improve predictions of food insecurity. Similar types of text data could help improve malaria prediction models as well, allowing them to better reflect what is happening on the ground in each country.
Effective Malaria Early Warning Systems Require Collaboration
To accurately predict malaria incidence risk, any model must incorporate country-specific characteristics. This means that researchers must work closely with countries’ Ministries of Health when designing malaria risk projection models. This will help to:
- Generate accurate projections of malaria incidence, by identifying key contextual characteristics and other important factors to include in the models;
- Ensure outputs are interpretable and meaningful for decision-making, as stakeholders across Africa have varying technical expertise, capacity, and resources to address malaria risk and transmission;
- Establish strong buy-in and ownership of the models and systems; and
- Create sustainable malaria projection models that do not require long-term support from funding agencies.
Ultimately, the use of malaria projection models will achieve sustainability goals only if the systems are owned by the government and not driven only by external funders.
Successful Risk Prediction Models Must Be Useful for Decisionmakers
Optimizing the cost-effectiveness and sustainability of models to predict malaria incidence requires a balance among precision, adaptability, and usability. The ideal early warning system for malaria combines AI methods with domain knowledge related to the human, biological, and environmental predictors of malaria incidence. Two models funded by the Gates Foundation show the importance of combining these methodologies:
- The Institute for Disease Modeling develops models to simulate how specific interventions (e.g., bed nets, vaccines, chemoprevention) affect the transmission of malaria under different scenarios.
- The Malaria Atlas Project uses geospatial and epidemiological modeling and analytics to map malaria transmission globally.
A framework that bridges these components and is open to decisionmakers across countries will help to:
- Ensure models are adaptable to both diverse contexts and changing circumstances, by leveraging the flexibility of AI methods with inputs and guardrails informed by domain expertise;
- Establish outcomes that are comparable across countries; and
- Support collaboration with a large network of stakeholders with varying expertise and resources.
Additionally, systems should provide actionable insights about the specific factors to be addressed to decrease malaria risk most efficiently. For example, is the forecasted risk of malaria higher because there are more mosquitoes due to lots of rain or because too few people use insecticide-treated bed nets? AI-supported decision-making can help countries move from simply understanding where the risk of malaria is high to addressing that risk by combining the identified primary contributors with existing evidence about intervention impacts and contextual factors (e.g., from media coverage and policy documents).
With tightening health budgets across Africa, ministries need tools that support evidence-informed decision-making while minimizing costs.
Working Together to Control Malaria
The future of malaria control depends on our ability to integrate data, theory, and technology into systems that deliver actionable insights, while ensuring that country governments use those insights. Co-creating malaria prediction models with national governments can help to achieve each of these goals and establish sustainable systems.
