AI tool helps get needed medicine throughout Sierra Leone

Solution

A research team including CHIBE’s Dr. Hamsa Bastani built an AI tool that helped improve patient access to medicine in Sierra Leone. This AI tool, a decision-aware machine learning framework for the allocation of essential medicines, helped increase the consumption of needed medicine by 19%. Read the paper here in Nature.

“As a mother, this work is especially close to my heart—supporting pregnant women and children means investing in the next generation and the future of Sierra Leone,” Dr. Bastani said. “It’s incredibly humbling to play a small part in bringing the benefits of AI to millions of people who have historically been underserved by technological advances.” 

Key Numbers

• 19% increase in consumption of the medicines that had been allocated to 5 treated districts in Sierra Leone
• 2 million women and children under 5 benefit from this tool, which has been scaled throughout Sierra Leone
• $30 is the cost of monthly server fees for this automated system to run  

Zeroing in on Sierra Leone

The team focused on Sierra Leone, which has a “Free Health Care Initiative” that gives free medical care and products to pregnant women and children under 5. One issue, however, is that the supplies are given once a quarter, and sometimes there is a surplus of supplies in one location and not enough in another location.

The country’s National Medical Supplies Agency manages the distribution of supplies to health facilities, and due to several difficulties with their system, 42% of facility requests go unfulfilled.

“By replacing manual effort with data-driven decision-making, our system streamlined supply chain management, reduced administrative burdens and adapted to real-time changing consumption patterns, enabling facilities to better align limited supply with patient demand,” the study authors wrote.

Sustainable impact

The team built a web application for the tool with an intuitive user interface. The system is also:

• owned by the Sierra Leone government
• integrated with government databases
• automated from data extraction to allocation results
• needs no additional workforce

The system is also a decision-support tool, meaning that the people on the ground can still override it if they see fit.

What else they learned

The research team found that previously under-served facilities saw a 32% increase in consumption of supplies, which they said suggests “that our system successfully addressed potential biases from uneven data quality and availability, leading to more equitable resource allocation.”

Helping rural facilities also did not compromise urban ones, they found.

Bottom line

“Our work demonstrates how machine learning methods can improve efficiency at very low cost in resource-constrained global health settings,” the study authors wrote.