AI and the Future of Malaria Prevention

Malaria remains one of the defining public health challenges of our time. According to the World Health Organization, the WHO African Region carries the large majority of global cases and deaths, with Nigeria bearing the largest share of any single country. Progress made in the early 2000s has slowed, and programmes face growing threats from insecticide and drug resistance, funding constraints and a changing climate.

Against that backdrop, it is tempting to see artificial intelligence as a breakthrough technology. We think a more useful framing is narrower and more practical: AI can help malaria programmes decide where and when to act with greater precision. That is valuable precisely because the tools of malaria prevention - nets, spraying, chemoprevention, case management and, increasingly, vaccines - already work when they reach the right people at the right time.

Where AI can help

Anticipating seasonal risk

Malaria transmission is strongly seasonal in many settings, shaped by rainfall, temperature and humidity with characteristic time lags. Statistical and machine-learning models that combine surveillance history with climate data can estimate how risk is likely to evolve over the coming weeks. Even modest additional lead time can help programmes pre-position commodities and schedule campaigns.

Seeing sub-national variation

National and even state-level averages can hide large differences between districts and wards. Geospatial models can estimate risk at finer resolution, helping programmes avoid spreading resources too thinly or too uniformly.

Prioritizing under constraint

Programmes rarely have resources to cover everywhere at once. Combining forecast risk with indicators such as the proportion of young children, travel time to health facilities and recent net coverage can produce transparent priority rankings - provided the weights are visible and agreed by stakeholders.

Reducing operational friction

Microplanning, logistics and supervision consume large amounts of programme staff time. Tools that draft plans from agreed priorities, and track planned versus completed activities, can free experienced staff to focus on problems only they can solve.

Where caution is needed

Data quality is uneven

Routine malaria data can be affected by testing rates, stock-outs, reporting completeness and changes in care-seeking. A model trained on such data can learn artefacts rather than transmission. Careful data assessment, and close work with programme epidemiologists who understand local reporting, is essential.

Forecasts are uncertain

Forecasts should always be presented with prediction intervals and evaluated against simple baselines. A model that is no better than “same as last year” adds complexity without value.

Equity can be undermined

Areas with weak reporting may look lower-risk simply because fewer cases are recorded. Without explicit checks, data-driven targeting can direct resources away from the communities that most need them.

People decide

AI outputs are inputs to expert judgement, not substitutes for it. Programme managers, epidemiologists and communities must remain in charge of decisions.

What this means for Epinova

Epinova’s initial malaria focus is Nigeria, where the scale and diversity of transmission make precise targeting especially valuable. We are developing forecasting and prioritization capabilities with these principles in mind: validation before use, uncertainty made visible, transparent weighting and human decision-making.

We are at an early stage, and we will not claim results we have not yet demonstrated. What we can say is that we are building with the people who run malaria programmes, measuring ourselves against simple baselines, and committing to report what we find.