DataKind’s Approach, Weaving Together Disparate Worlds, More Needed Than Ever

As its first effort under the new strategy, DataKind began last year focused on community health care in line with the collaborative’s goal of capacity building within critical existing systems.

Frontline health workers play a critical role in administering vaccinations, delivering healthy babies and tamping down pandemics. Data science and machine learning can help them achieve their missions, but it can be difficult to know where to start.

“Most organizations can’t articulate their AI needs,” Porway says. “So we talk to folks, hear their challenges.” DataKind spoke to some 50 organizations and selected about 20 where data science, AI and machine learning could help not only the organizations but also deliver insights to drive sector-wide improvements, potentially impacting health outcomes for millions of people.

“We’ve teamed up with amazing partners,” Porway says. Initial work began last year with three organizations in Africa, Jacaranda Health, Medic Mobile and Riders for Health, all delivering life-saving primary healthcare to the hardest-to-reach communities and most vulnerable populations.

For example, Riders for Health is an international nonprofit that serves rural villagers in five African countries by using motorcycles to deliver critical care. One of their key operations is transporting medical samples of infectious diseases from rural health care centers to labs. To maintain quality and integrity of samples from the point of collection through transportation, storage and analysis, Riders for Health couriers capture details by hand in a logbook.

Some couriers transport over a thousand samples a day. The laborious and time-consuming process of log entry deeply impacts their ability to reach more health facilities more frequently, causing degradation of samples and delays in diagnosis and treatment. Currently, the average time to digitize logbook entries and sample details is 30 to 60 days.

DataKind’s data scientists teamed up with Riders for Health to develop a prototype tool that will automate key handwritten entries of sample collection forms so that downstream workflows are dramatically sped up and lives are saved. In addition to building these character recognition systems to quickly digitize written records, volunteer data scientists are also improving predictive models for better maternal and child health outcomes, assessing frontline health worker training, and developing data quality and assurance algorithms for better patient tracking and care.

“So, knowing that isolation is probably coming to African countries within the next month, some groups are thinking about tech solutions to help provide remote care. Could community health workers deliver medically needed supplies? How do we get good health information into rural communities? In other words, how do we repurpose these existing networks for the new COVID-19 challenges?”

The Foundation/Mastercard collaborative is founded on the belief that improved and targeted data science capabilities of non-profit, civic and government organizations can help local leaders uncover new insights and trends from their data and build more impactful programs for the communities they serve.

Porway agrees. “There is an urgency now to strengthen the social safety net. There are lots of things that are huge systemic problems, not data machine learning problems, so we need to be humble about where these data-driven tools are useful. We have to find the right problem, given the new reality. But decision makers need data to make decisions in game-time situations, and data can also make folks more efficient, and make resources go further.”

Porway may be at home like many of us these days, but he’s definitely working, trying to figure out what’s next for DataKind. He once thought he had to backpack through Europe in search of a revolution, but instead he found his revolution here among APIs and a like-minded community of data scientists.