How a global non-profit uses AI to bridge healthcare gaps
AI in healthcare: scaling services to the neediest areas
Millions in regions around the world need surgery and other urgent healthcare, but connecting them with life-changing care is one of the world’s greatest medical challenges.
That mission is the guiding vision of the nonprofit Virtue Foundation. Formed by like-minded surgeons and other medical professionals, Virtue Foundation aims to serve resource-poor regions worldwide in an effective, ethical, and lasting way.
“The DataRobot platform and people have been a phenomenal partnership to develop, deliver, and govern AI at scale. Without them, we could not have realized our goals of scaling our global health efforts. We’ve created the whole analytics infrastructure with AI capabilities.”
Since 2001, Virtue Foundation has brought care to patients in 25 countries. In Ghana, that meant reducing child mortality rates due to respiratory failure from 14% to 4% at one hospital. In Mongolia, the nonprofit’s efforts have enabled hundreds of surgeries and thousands of hours of best-in-class clinical training for local surgeons.
The nonprofit connects volunteer surgeons, physicians, and global health actors to those in need. But with spotty data in many countries, it’s difficult to decide where their efforts will have the greatest impact.
“Global health is a supply-demand challenge. It’s hard for the people who need care the most to make themselves known and for caregivers to find them,” explains Dr. Joan LaRovere, Co-Founder and President and Assistant Professor of Pediatrics at Harvard Medical School. “How could we do this faster and more efficiently? We saw where healthcare was going and knew AI was going to be part of it.”
Expediting predictive modeling in healthcare using geospatial data
To turn data into answers, the nonprofit launched the Actionable Data Initiative as the heart of its matchmaking efforts. As they built their dream team of data scientists and technologies, DataRobot became an early partner to develop, deliver, and govern AI at scale.
“We wanted to start building predictive analytics and we knew DataRobot was going to be transformative for achieving that faster and more effectively,” LaRovere said.
Driven by their mission, data scientists from Virtue Foundation and DataRobot teamed up to tackle the challenge of finding, organizing, and analyzing datasets. Fortunately, they didn’t need extensive onboarding to use the suite, expediting the time-to-value.
“With DataRobot, we can quickly go from idea to prototype to production. We simply drop in a dataset, click a button, and the modeling process begins. We’re able to identify areas of need in a way that would otherwise be impossible at such a large scale.”
The Virtue Foundation team finds that DataRobot accelerates every step of predictive modeling, and critically, helps make sense of granular geospatial data.
“With DataRobot, we can quickly go from idea to prototype,” explained Nicolas Douard, Lead Data Scientist at the Virtue Foundation. That ability to scale AI rapidly means the Virtue Foundation can more readily bring healthcare to dozens of low- and middle-income countries.
Optimizing volunteer matching with AI
Their efforts have culminated in an innovative technology platform called VFMatch. Powered by DataRobot, Databricks, and CARTO, VFMatch’s interactive 3D map directs volunteer healthcare providers to the locations where their skills are most urgently needed.
After the soft launch of VFMatch, the Virtue Foundation received hundreds of applications from new volunteers and monthly usage grew to tens of thousands of site visitors.
Using complex datasets to identify medical deserts
Another part of the nonprofit’s efforts is finding “medical deserts” and predicting suitable locations for new facilities.
“Given limited budgets and resources from governments or global health contributors like NGOs (nongovernmental organizations), where can you maximize your impact? This is the main question we looked to answer,” Douard explains.
To meet the challenge, DataRobot helps Virtue Foundation draw from numerous data sources, such as population density, land cover, elevation, road networks, and current healthcare facilities, to identify suitable locations.
The resulting multimodal model with DataRobot uses geospatial features such as images, coordinates, and numeric variables with interactive visualization tools to explore the data. With that, Virtue Foundation generates a list of potential areas for new facilities. Their team then ranks potential locations and determines where a hospital or specialized surgery center is needed most.
“We can effectively predict suitable locations, which is the first step in maximizing the impact and reducing travel time to healthcare for as many people as possible,” Douard said. “It’s a central piece in our efforts in global health.”
The visual elements of the models also help explain to others which features drive predictions and facilitate critical conversations.
“In one meeting with a minister of health, the ability to see the data helped them understand. It became ‘Clear the calendar.’ A half-hour meeting turned into two hours,” LaRovere said.
Pinpointing last-mile medical needs with AI
Other models are unique to some resource-poor regions. For example, Virtue Foundation learned that transportation infrastructure is essential for effective medical facility placement.
“We wanted to start building predictive analytics, and we knew DataRobot was going to be transformative for achieving that faster and more effectively.”
That insight led to tricycle ambulances in Ghana. Virtue Foundation uses modeling to determine where three wheels could link people with potentially life-saving care, such as women in labor. Sixty percent of these rides are used for maternal care.
The nonprofit also turns to modeling to predict disease outbreaks and proactively prepare. In Ghana, they built a model to predict meningitis cases within granular geographical subdivisions, showing the approach in a successful proof of concept.
While data is not always available for a particular area, DataRobot enables Virtue Foundation to extrapolate from alternative data sources and geospatial features to address these gaps.
“We can get actionable insights on which regions may benefit from specific healthcare resources,” Douard said. “This is a major milestone in the global health puzzle we’re looking to solve.”
Matching caregivers with needs
These are just some of the models helping the Virtue Foundation achieve its ambitious goals and scale much-needed healthcare around the world. In a short time, they went from “three laptops in a room” to a world-class team of data scientists from various organizations.
For LaRovere, it’s rewarding to see their vision take shape and she appreciates the teamwork that made it all happen. Working together, data scientists from Virtue Foundation, DataRobot, and others published a paper about their novel approach in 2023. In 2024, a follow-up study explored the prioritization of suitable sites.
“DataRobot is more than a platform,” she says. “The technology and the people have been a phenomenal partnership. Without them, we could not have realized our goals of scaling our global health efforts. We’ve created the whole analytics infrastructure with AI capabilities.”