Applying an Equity Lens to Data-Driven Health Solutions
By Claudia Juech and Rebecca Distler
Data and AI-enabled tools hold great promise for improving health. But to realize the benefits of digital innovation in healthcare, including AI, we need to apply an equity lens to every stage of innovation — from conceptualizing solutions to data management to developing, deploying, and monitoring these tools in real-world contexts.
Through our work, we’re fortunate to partner with a wide range of organizations at various stages of their data journeys. Here are three things we’re learning along the way, and where we see opportunities to strengthen our collective commitment to ensuring innovative technologies are deployed to improve healthcare.
Center people in data-driven decision-making
Ultimately, health systems are powered by people. From policymakers and program managers to providers and patients themselves, each of these stakeholders seek to use different data in different ways to make decisions that impact health outcomes.
For any data-driven intervention to have a meaningful impact, that data must be meaningful to the person using it. By centering people in the generation, analysis, and use of data-driven solutions, we can support the design of interventions that reflect the actual needs of communities and the reality of health systems.
Nexleaf Analytics has long been a champion of this approach. The nonprofit technology company deploys low-cost sensors and data analytics to provide partners with real-time visibility around the vaccine supply chain. In Malawi, Nexleaf is now working with immunization program staff in the Ministry of Health to understand what data they need — and how they want to use it — to help protect vaccines.
By supporting the immunization program in having the autonomy to access, analyze, and act on supply chain data in real-time, they are demonstrating how a culture of data use can help ensure life-saving vaccines reach those who need them most.
Embed equity into datasets — and into data lifecycle management
Quality and bias are pervasive challenges across health datasets, which are often missing, incomplete, or not representative of underserved populations. As an example, a recent review of genome-wide association studies — which help scientists identify genes associated with a particular disease — found that nearly 80% of individuals included were European, compared to 10% Asian, 2% African, and 1% Hispanic. This could translate to poor or inadequate diagnosis and treatment for individuals already at risk of exclusion from healthcare systems.
These dataset challenges are among the ones the Lacuna Fund aims to address. Alongside a coalition of funders, we are supporting the health equity track of the Lacuna Fund to provide data scientists, researchers, and social entrepreneurs in low- and middle-income contexts with the resources to produce new datasets to address underserved populations or problems, augment existing datasets to be more representative, or update old datasets to be more sustainable. The Fund places specific emphasis on datasets that are locally developed and owned, and are open source.
Yet considerations of equity and fairness do not begin or end with the datasets themselves. Rather, we need to embed them into every step of continuous data lifecycle management. Our Data and Society team has been exploring this in practice — the data lifecycle diagram is a snapshot of a process to do this well:
Evaluate and monitor data systems through an equity lens
Even when data products are co-created with communities and equity considerations are applied to their development, we need to rigorously monitor the performance of these solutions when integrated into healthcare settings. As models migrate from labs to exam rooms, we have seen examples of AI that exacerbate existing biases in who receives care or could be found to cause harm when applied to new populations. This underscores the importance of equipping public health and medical professionals with tools to understand how these solutions behave in real-world settings — and react in real-time.
That’s the driving force behind a new collaboration we’re supporting between the Duke Institute for Health Innovation and the Aga Khan University in Pakistan. Working together, these teams are developing analytical tools and documentation frameworks for auditing, evaluating, and monitoring AI software in clinical settings.
The tools are being designed for a range of decision-makers, including front-line clinicians, operational leaders who procure these systems, and technical leaders responsible for their maintenance. The international collaboration was born with the recognition that these tools must be globally extensible and consider diverse health system needs.
Let’s learn together
As AI research continues to evolve, particularly in healthcare, we need to ensure that we’re building intentional bridges between innovation and impact. No matter where an organization or partner is in their data journey, we can all work to embed equity into our data-driven work.
What are you learning about equity in your own data and innovation projects? We look forward to hearing your thoughts in the comments section.
Claudia Juech leads the Data and Society program, which builds technical capacity and data-use cultures at nonprofits and social impact organizations around the world. This post was initially sparked by Claudia’s participation in the “Building bridges between innovation and impact on AI in healthcare” event, organized by the Geneva Hub for Global Digital Health (gdhub) on March 21st, 2022. You can watch the event here.
Rebecca Distler leads the Foundation’s work to advance digital health equity in collaboration with Foundation leadership, programmatic teams, and external partners.
