Amid surging levels of child malnutrition, Kenya has turned to a newly minted AI forecasting tool to stay ahead of the curve. Adopting the tool is Kenya’s biggest leap toward data-driven humanitarian responses, with nearly 800,000 children in the country requiring treatment for acute malnutrition.
Predicting child malnutrition with AI in Kenya
According to a press release, Kenya has integrated a new AI forecasting tool for child malnutrition into its national drought early warning system.
Announced at the One Health Summit in Lyon, the tool generates predictions of child malnutrition while visualizing them in an interactive dashboard. Researchers from the University of California and Cornell University teamed up to develop the tool, with the Jameel Observatory leading real-world testing efforts in Kenya.
In addition to the formal integration with Kenya’s National Drought Management Authority (NDMA), the AI forecasting tool has recorded a raft of use cases. Already, predictions from the dashboard have found utility in the Kenya Food Security Steering Group meetings, with the UN and donors leaning on the data for hunger prevention.
At the moment, child malnutrition in Kenya remains a major public issue. According to the latest UNICEF data, 2 million children are too short for their age due to chronic malnutrition.
Meanwhile, 11% are underweight, while 4% suffer from acute malnutrition. Malnutrition in Kenya is uneven, with the arid regions of Northern and Eastern parts the hardest hit.
“This is the problem being tackled head-on by UC Berkeley, the Jameel Observatory, and the NDMA,” said George Richards, Director of Community Jameel. “Harnessing machine learning to forecast when children in Kenya are at risk of acute malnutrition.”
Here’s how the AI forecasting tool works
The AI-powered tool leans on the machine-learning approach developed by Susana Constenla-Villoslada, a doctoral researcher at UC Berkeley. A closer look under the hood reveals the internal workings, anticipating nutritional risk based on observable trends, including child measurements.
The tool combines historical global acute malnutrition (GAM) rates from NDMA with weather data, conflict, and food prices to measure risk over a period of time. By relying on high-frequency longitudinal data, researchers confirmed that the tool outperforms previous approaches by a country mile.
The team noted that since malnutrition changes gradually and underlying drivers tend to persist, past GAM levels are predictive of future ones. A key part of its success is its implementation on the field with Jameel Observatory’s partners, including the International Livestock Research Institute (ILRI) and the University of Edinburgh, leading the push.
To ensure accuracy, the AI forecasting tool is updated monthly, with the NDMA operating it independently. Constenla-Villoslada told Charity Journal that the NDMA and Kenyan authorities will determine the specific protocol for triggering financial aid or food distribution once the dashboard predicts a spike in malnutrition.
“Our collaboration (UC Berkeley) with the NDMA is for the generation of the early warning and triggering system,” said Constenla-Villoslada. “The NDMA and other stakeholders decide what interventions are held in their territories.”
While not expressly stated, Kenya’s NDMA can issue automatic alerts to the Ministry of Health and aid agencies. Furthermore, a rapid nutrition response involving door-to-door or clinic-based targeting of children under five in drought-hit regions is tipped to record impressive success levels.