How human expertise and AI can support vector-borne disease control

As climate change, urban growth and human movement reshape the world around us, the diseases carried by mosquitoes are changing too. Malaria, dengue and other mosquito-borne illnesses are appearing in new places, putting more communities at risk than ever before.

A new Perspective paper in Environmental Change and Disease Dynamics and co-authored by WorldPop visitor Dr Zeyu Zhao, colleagues from Xiamen University and WorldPop’s Professor Shengjie Lai, explores how artificial intelligence (AI) and human expertise together could help public health teams respond more effectively to these shifting health risks.

Diseases such as malaria, dengue and chikungunya cause more than 700,000 deaths each year. Their spread is increasingly influenced by these environmental and demographic shifts, making it harder for health authorities to identify where risks are emerging and who is most vulnerable.

The authors argue that AI should be assessed by its contribution to public health action, alongside its ability to predict disease patterns. A forecast or risk map becomes useful when it helps teams decide where to investigate, when to intervene and how to allocate resources.

Their framework ties together four stages of disease control:

• Surveillance and detection – spotting the first cluster of cases.
• Risk assessment – understanding who and where is most exposed.
• Early warning – flagging the next likely hotspot before it’s too late.
• Intervention planning – deciding what to do, with whom, and how fast.

AI can help analyse health, environmental and population data, while public health experts guide the analysis, check findings against local conditions and decide what action is feasible. The paper sets out an approach for development and evaluation within public health systems.

Understanding where people are at risk

A central theme is that populations at risk change over time. Residential population estimates provide an essential starting point, but commuting, seasonal migration, tourism and displacement can alter where people encounter disease-carrying mosquitoes.

This is where WorldPop’s work at the University of Southampton steps in. Our open datasets map population numbers, age, and sex structures at an incredibly fine scale—down to roughly 100 meters. Covering 2015 to 2030, these projections blend census data, satellite imagery, and advanced modeling to show not just where people live, but how populations are shifting. When paired with anonymized mobility data and local health surveillance, these maps could help health teams track exposure in real time, across seasons and borders.

For mosquito-borne diseases, this also means considering where mosquitoes live and when they bite. Population presence alone does not establish exposure: environmental conditions, housing and protective measures also matter. Mobility data can under-represent some groups, so estimates need local validation and clear communication of uncertainty.

Connecting information with action

The authors emphasise that human involvement is essential throughout the process. Local teams need to verify surveillance signals, interpret uncertainty and assess whether proposed interventions are affordable, practical and acceptable to communities.

Transparency, privacy protection and accountability are equally central. Models must be checked as conditions change, with their performance assessed through their contribution to timely, equitable and effective disease control.

The Perspective calls for AI to be integrated into existing public health systems, with clear responsibilities and continued evaluation. This combined approach could help teams make better-informed decisions about protecting communities in a changing world.

Headline image: Sleeping under mosquito net, Chris Clogg, 2014, Public Domain.