Leveraging DHIS2 data to create AI-driven health facility forecasts in Cote d’Ivoire

This abstract has been accepted at the 2024 DHIS2 Annual Conference

Leveraging DHIS2 data to create AI-driven health facility forecasts in Cote d’Ivoire

Since 2021, the Ministry of Health of Côte d’Ivoire, in partnership with Pendulum and supported by USAID and BMGF, has been an early adopter of AI for public health supply chain optimization. This partnership has gradually grown from a theoretical competition to a field based proof of concept to a demonstrated solution that is scaling. In 2021 USAID launched an Intelligent Forecasting Competition that invited global technologists to develop AI models to improve Côte d’Ivoire’s family planning forecasting. The Ministry then invited the winner, Pendulum (then known as Macro Eyes), to fully train and deploy its model in country, with USAID’s support. Cote d’Ivoire’s development and implementation of AI based forecasting has involved a broad set of partners. The government created a technical working group including multiple programmatic, supply chain, and technical agencies to support Pendulum as it accessed, interpreted, and generated new insights with DHIS2 and other government data for the first time. mSupply, the country’s new LMIS platform partner, also worked closely with Pendulum to seamlessly integrate its forecasts into mSupply Mobile, so that health facility workers can reference them as they place their monthly product orders via the app. In the partnership’s first year, Pendulum developed a comprehensive forecasting model for 11 family planning commodities using data from 2500+ health facilities that delivers forecasts 38% more accurate than the traditional methods. To generate its predictions, Pendulum accessed, cleaned, processed, and adapted its machine learning models to data from Côte d’Ivoire’s various health data systems, including DHIS2 as well as eSIGL and mSupply. The Ministry selected 52 health facilities to receive AI driven forecasts, which Pendulum has delivered each month via mSupply Mobile. MSHP and its agencies have also tested the use of AI to improve forecasting of other health products, starting with malaria. In 2023, Côte d’Ivoire’s national malaria control program (PNLP) conducted a proof of concept for 36 malaria commodities that showed that Pendulum’s forecasting model reduced forecasting error by an average of 37%. Cote d’Ivoire is now leaning further into the use of AI tools in 2024. Pendulum will deliver its family planning forecasts to all ~1000 health facilities that have access to mSupply Mobile, deploy malaria forecasts to facilities for the first time, and potentially begin deploying the next piece of its supply chain optimization solution–AI driven allocation–to holistically optimize inventory and allocation of family planning commodities to each warehouse and facility each month.

Primary Author: Brittany Hume Charm

AI ML machine learning integration interoperability forecast forecasting