In Kenya, every day, hundreds of children fall ill with acute child malnutrition. It is a silent tragedy which affects 350.000 children under five years of age, turning entire families into helpless spectators of a tragedy that seems inevitable. But what if we could know exactly where and when malnutrition will strike before it even happens?
A group of researchers from USC, Microsoft, Amref Health Africa e Kenyan Ministry of Health has developed an artificial intelligence that does just that: predicts nutritional crises six months in advance with 89% accuracy.
An algorithm that reads the future of malnutrition
The system, described in the study published in PLOS One May 14, 2025, represents a paradigm shift in the approach to child malnutrition. Girmaw Abebe Tadesse, principal scientist at the Microsoft AI for Good Lab in Nairobi, explains the importance of this project:
“Malnutrition poses a significant challenge to children in Africa, a continent facing severe food insecurity exacerbated by climate change.”
The model combines clinical data from more than 17.000 Kenyan health facilities with satellite information on crop health and agricultural productivity. This fusion of information allows it to identify emerging areas at risk with much greater accuracy than traditional methods.
When researchers test the one-month forecast, the accuracy rises to 89%. Six months later, It still maintains an impressive 86% accuracy. For comparison, methods based only on historical trends in malnutrition achieve just 73% accuracy.
How Childhood Malnutrition Prediction Works
The key to success is integrating different data sources. The system draws from the District Health Information System 2 (DHIS2), a platform that collects health data from clinics across Kenya. At the same time, it analyzes satellite imagery from NASA’s MODIS to measure Gross Primary Productivity (GPP), an indicator of how well crops are growing.
Laura Ferguson, director of research at the USC Institute on Inequalities in Global Health, notes that “malnutrition is a public health emergency in Kenya. Children are getting sick needlessly. Children are dying needlessly.” Current forecasting methods rely primarily on expert judgment and historical knowledge, approaches that struggle to anticipate new outbreaks or rapid changes.
The AI model, on the other hand, can spot hidden patterns in the data, identifying correlations between crop conditions and malnutrition rates that would otherwise escape human analysis. Regions with poor crop health often show higher rates of malnutrition in subsequent months.
Satellites That “See” Hunger Before It Arrives
The most innovative aspect concerns the use of satellite data. The images from Moderate Resolution Imaging Spectroradiometer NASA (say, say: “but what’s the point of space research?”. Say it now) provide precise measurements of crop productivity in real time. When GPP drops in a region, the algorithm can predict an increase in child malnutrition in the following months.
Come we have underlined in previous analyses, artificial intelligence is radically transforming the healthcare industry, and this application is a perfect example.
Bistra Dilkina, co-director of the USC Center for Artificial Intelligence in Society, calls the model “a game changer.” The system is proving particularly effective at predicting malnutrition in regions where prevalence fluctuates and peaks are difficult to anticipate.
From research to reality on the field
Researchers have developed a prototype dashboard that visualizes regional risk of malnutrition, enabling faster, more targeted responses. Samuel Murage, of the Kenyan Ministry of Health, explains that
“The best way to predict the future is to create it using available data for better planning and pre-positioning in developing countries.”
Ferguson and Dilkina, as mentioned, are now working with the Kenyan Ministry of Health and Amref Health Africa to integrate the model and dashboard into government systems and decision-making processes.
The goal is to create a sustainable and regularly updated public resource. As Ferguson points out, “most global health problems cannot be solved in health alone, and this is one of them.”
Zero Malnutrition: An Achievable Goal
Kenya shares the same health information system with more than 125 countries, which means similar AI tools could be deployed in many low- and middle-income nations. “If we can do it for Kenya, we can do it for other countries,” Dilkina says. “The sky is the limit when there is a genuine commitment to working in partnership.”
In a world where Every minute 35 children destined for hunger are born, this technology finally offers real hope. It is no longer a question of reacting to emergencies, but of preventing them. And when it comes to saving lives, six months of advance notice can make the difference between life and death.
