Q&A

PATH’s first chief AI officer on how the technology will shape the global health ecosystem

Exemplars News spoke with Dr. Bilal Mateen, who was previously Executive Director of Digital Square, about how artificial intelligence could be incorporated into product development, diagnostics, medical devices, and health program areas


Tags
Dr Bilal Mateen, center, PATH's chief AI officer.
Dr Bilal Mateen, center, PATH's chief AI officer.
©PATH

Dr. Bilal Mateen, the newly appointed Chief AI Officer at PATH, thinks his new job is "the best kind of job because I get to define it." The appointment of Dr. Mateen, a physician by training with an academic background in health-related applications of data science and machine learning, comes at a time when PATH, like other global health organizations, is increasingly focused on the use of artificial intelligence to achieve its goals.

PATH has said it believes Dr. Mateen is the first chief AI officer "in the health NGO-space." Dr. Mateen's mission, according to PATH, is on "developing and implementing PATH’s AI strategy, which will focus on using AI to improve the efficiency and effectiveness of PATH’s programs, as well as to develop new health solutions." He will also work to build partnerships with other organizations in the AI and health sectors.

Exemplars News spoke with Dr. Mateen, who was previously executive director of Digital Square, a PATH-led initiative, and clinical technology lead and senior manager for digital technology at the Wellcome Trust, about digital transformation and how AI could be incorporated into product development, diagnostics, medical devices, and health program areas.

Congratulations on being named PATH's inaugural Chief AI officer. When we last spoke in 2022, you were working with Wellcome Trust's Data for Science and Health team. Could you tell us about what you've been working on for the past couple of years?

Dr Mateen: When I left the Wellcome Trust, I became Executive Director of Digital Square, which was founded by the Gates Foundation and the U.S. government, specifically USAID, almost eight years ago. It was established in the aftermath of the West African Ebola outbreak based on the reflections of a series of funders, implementers, and other partners. They recognized that the fragmented digital health ecosystem had contributed to the sub-optimal response to that crisis. Many lives could have been saved if decision-makers had timely access to the right data, enabling them to make more informed, evidence-based decisions.

Digital Square was created as a market-shaping entity to bring together dozens of partners to rethink what digital transformation of health systems in low- and middle-income countries could look like. Over the past eight years, we’ve raised almost US$170 million in investment across 33 African and Asian countries. We do everything from providing central governments with technical assistance, helping partners develop digital health strategies, scaling pilot projects to national-level infrastructure in their digital health ecosystems, as well as supporting the coordination of donors within countries so that the ecosystem is less fragmented.

But it is what Digital Square does at the global level which arguably makes it unique. The U.S. government’s digital health strategy introduced the idea of a 'global goods for health,' and we’ve been the stewards of that concept for almost a decade. There are many digital public goods out there, but not all of them are mature, scalable, or sustainable. We’ve curated a list of global goods for health – we’re now just over 40 – selected from the tens of thousands of digital public goods that exist. I had the privilege of doing that for about a year, helping expand the strategy to include new ideas around digital public infrastructure, like digital financial services. We recognized that incentivizing care-seeking behavior often comes down to access to money or insurance, and that incentivizing care deliverers like community health workers to show up requires actually paying them for the work they do. All of this requires a digitized financial system.

We also began thinking about the impacts of climate change. What does a climate-resilient health system look like? And more importantly, how do we take what we’ve learned about ‘global goods for health’ and apply that to ‘climate and health.’ We’re very lucky to have been entrusted by both the Rockefeller Foundation and Wellcome to do that thinking, and the team will be publishing the results over the course of the next year.

And finally, the best part of the executive director role was that I got to chart new territories and shape the efforts to transform global health through AI; it’s where I’ve spent my life as an academic, and to be able to finally make it my day job (as Chief AI Officer) is so exciting. We moved from digitization to digital transformation not too long ago, and now I think we’re rapidly approaching the next paradigm shift – leveraging AI to accelerate product development, diagnostics, medical devices, and the core health programs we’re focused on. To be at the bleeding edge of that conversation, to get to contribute to the UN high-level expert panel on AI’s work, to do a keynote at the G20 on the regulation of AI in health, was an enormous privilege.

What are some of the lessons you learned in your past roles at Wellcome Trust and Digital Square that you're bringing to PATH?

Dr Mateen: I’ll share with you two lessons. The first is the nature of different types of capital. At Wellcome, an independent philanthropic organization, we talked a lot about philanthropic capital being catalytic. Whereas, Digital Square is largely supported by bilateral and multilateral donors, which was something I hadn’t fully experienced before. It’s a very different type of money – it’s focused on building systems-level change that’s trying to be more sustainable than the catalytic nature of philanthropy.

What I learned from Digital Square is that while donors each play crucial roles, we often fail to ensure that everyone is brought along for the full journey. Many times, you’ll see a philanthropic grant kick start a project, but there’s no clear plan from the outset about who will sustain it in the long term – whether it’s a government, bilateral, or multilateral organization. As a result, when the initial funding period of one to three years ends, the project often stalls because no one is prepared to take it forward. This is still a common issue in the digital health ecosystem. There’s a tendency to sometimes 'move fast and break often,' which works well in tech development but doesn’t necessarily set us up for lasting success. Now, whenever I take on a new project, I think not only about its immediate impact but also about who will fund it in three years and how we can structure it in a way that builds the evidence needed to secure that future investment.

The second lesson I learned is about navigating the complexities of working at the national level. The dynamics, incentives, and challenges are different from what I initially expected. When I first started, I had a lot to learn. Before my time with Digital Square, I hadn’t spent much time engaging with senior technical decision-makers in ministries of health; my constituents at Wellcome were academics. My earlier experiences gave me exposure to the financial side of the ecosystem, but Digital Square helped me gain a deeper understanding of how decision-making processes work and the importance of multilateralism – which even to the most nuanced outsider can sometimes look like a glorified talking shop. I can’t explain how important those two lessons have been in helping me navigate this new role, and for articulating why I’m driving our strategy at PATH in a very specific direction.

How do you plan to integrate AI into PATH's programs and work?

Dr Mateen: That’s a great question but I want to flip it around. Instead of focusing on how PATH is going to integrate AI into its work, I think we should be talking about how we want to see AI shape the ecosystem. Rather than laying out eight specific projects we have worked on, I’d like to talk about what the big challenges are that would be useful litmus tests for whether we’re making progress.

For example, how do we get 100 randomized control trials (RCTs) of AI-based tools published, based on work done in Africa? A review published earlier this year showed only two RCTs of AI-based tools used in Africa had been published, with a third published a few months ago. It’s not because I’m an RCT zealot, instead, they provide a useful litmus test for evaluating the capacity of clinical trial infrastructure across Africa. A target of 100 is arbitrary, but if we can do that many, it suggests we’ve built the capacity we need, whilst at the same time generating more evidence for local decision-making rather than relying on RCTs conducted in the US, which operates in a completely different health care context.

Another challenge could be enabling small- and medium-sized enterprises (SMEs) in Africa to generate a billion dollars in revenue from selling their AI-based tools to the local health care systems. Again, it’s an arbitrary number, but if we can get 1,000 SMEs to exchange a million dollars each using common digital rails – meaning there’s an effective coding infrastructure, contracting procedures, etc. – we’ll have created a functioning marketplace. This is why the US has over 950 tools cleared by the FDA – the tools that smooth the path for innovators to get paid for their work exist, at scale! How do we create those same enabling features across the African continent (and other low- and middle-income contexts)?

One more challenge: how do we move from the current regulatory environment in Africa, where we rely on evaluations from high-income countries, to one where there’s 100% regulatory coverage for AI as a medical device? The African Medicines Agency or national regulators in Ghana, South Africa, or Kenya could be setting policy, rather than relying on evaluations from the US or UK, which don’t always generalize well to African contexts. Ghanaian regulators need to be able to say, 'Before you deploy this tool in our health system, let’s make sure the data you used to evaluate it applies here.'

In essence, I wish we spent more time talking about the changes we want to see in the world. My job is to convince donors that these and the handful of other ‘grand challenges’ I have swirling around my head are the right ones for us to anchor on as a community. Even if they don’t fund PATH to deliver on them, if it gets done, that’s still success as far as I’m concerned.

PATH's three main focus areas are product development and access, health and disease management, and health systems strengthening. Broadly, how do you think AI could improve each of these in the coming years?

Dr Mateen: Let me give you an example for each, and I’ll try not to go for the obvious ones. In product development, a clear example would be AI for immunogen design, which is a hot topic right now due to the advancements in protein folding prediction. AI can really accelerate drug development pipelines, and it’s exciting in the global health context because it has the potential to drive down the cost of drug discovery and vaccine development for neglected tropical diseases. These are areas where the private sector doesn’t usually engage because there isn’t a large market. If we can make drug discovery cheaper and less risky, we can encourage private-sector investment.

But instead of talking about that, let’s discuss regulatory system strengthening. How do we help regulators start using AI in a way that’s safe and effective? Could we train an algorithm to review dossiers submitted by private companies and flag areas where there’s insufficient information or unjustified claims? The AI wouldn’t replace regulators but would augment their capacity, helping them review submissions more efficiently and accurately, where currently their inability to move at pace has resulted in bottlenecks and delayed access to critical medicines. However, we’re still in the phase of generating evidence about the value of generative AI in regulatory work. Rather than jumping in and deploying it everywhere, we’re thinking about setting up experiments, similar to clinical trials for drugs. For example, we could work with a regulator and run a mini randomized control trial within their organization. Some staff would have access to an AI tool, and some wouldn’t, and we’d measure how long it takes them to complete their work and whether the quality of their work improves. That way, we’ll better understand the costs, risks, and benefits before scaling up.

For health system strengthening, I’ll give you an example from supply chain work. Traditionally, the approach has been to collect information from the last mile and aggregate it up through district and national levels, where data analytics is then used to inform policymaking. But what if we flipped that around? What if we made this information available at the last mile? A community health worker could use a chatbot to ask, 'Where’s the nearest pharmacy with anti-malarial drugs in stock?' By using consumption modeling and national data, we could help health workers direct patients to the right pharmacy, reducing their risk of going to a pharmacy with a stock-out and ensuring that they get the right drugs, at the right time.

These are the kinds of things we could do to improve operational efficiency in health systems without necessarily needing to dive into the more complex regulatory space of AI for diagnosis or risk prediction, which comes with its own challenges.

How do you think artificial intelligence will impact global health overall in the coming years? And how do you think the technology can be applied ethically and inclusively in global health?

Dr Mateen: I wrote about this recently; we’ve been talking about the potential of AI for a long time. The first FDA-approved AI-based tool was a computer-aided detection algorithm for breast cancer, approved in the mid-1990s. That’s around the time I was born. Has AI had the impact we hoped for? Most people would argue no. We keep going through these boom and bust cycles where excitement around a critical advancement leads to a huge influx of investment, and then our failure to hone in on the value proposition and deliver impact results in the bubble bursting.

There are so many vital challenges we haven’t addressed, which means that we’re at risk of another 'AI winter.' For example, we haven’t addressed the infrastructural or market-specific challenges that frustrate the scaling of effective AI- based tools in health care – all the work around interoperable health systems that I worked on as executive director of Digital Square. We haven’t built enough evidence to justify long-term investment, whether from governments or other stakeholders. I’ve said it before, and it might be a bit provocative, but philanthropy is a fickle friend. It’s useful, but trying to build an entire field on the back of philanthropic capital is dangerous. So, how will AI impact global health in the next few years? That depends on the decisions we make now. We could waste another important opportunity, or we could make a profound impact. It’s up to us.

As for applying AI ethically and inclusively, I think one of the most promising areas is filling the gap in empathetic conversations – those areas where busy health care workers don’t have the time to engage with patients in the way they’d like. There’s a great example funded by the Gates Foundation – the company leading the project is called Audere – where an empathetic chatbot is used to encourage people with HIV to seek care, stay in care, and understand their treatment options. It’s not a medical device or diagnostic tool, but it addresses the problem of treatment retention and care-seeking behavior. These kinds of AI-driven tools could solve some of the critical challenges we face in health care, like getting patients to seek care at the right time and keeping them engaged in care.

What in your current or upcoming work is most exciting or energizing for you?

Dr Mateen: One of the most exciting things we’re working on right now, with support from the Gates Foundation, is what we believe will be the world’s first randomized control trial of a large language model-based clinical decision support system in Africa. We’ll be submitting it for registration in the next few months. The project involves a series of African partners – a local academic group leading the research (KEPRECON), a local primary health care provider (Penda Health) working with OpenAI to develop the product, and a range of global partners providing support (e.g., the University of Birmingham Centre of Excellence in Regulatory Science and Innovation).

It’s exciting because it shows that there’s capacity on the continent to pioneer cutting-edge research and generate evidence. Assuming someone doesn’t beat us to it, we could be the first to register a trial like this in the world.

Another project I’m excited about is testing how AI can be useful in eliminating cervical cancer. One idea we’ve been dying to explore is the use of an empathetic chatbot to guide women through the care continuum of prevention and screening for cervical cancer. The chatbot could assess their risk factors, encourage self-testing for HPV (a major risk factor for cervical cancer), guide them through the test, and help them interpret the results. It could nudge them towards vaccination against HPV if appropriate. If they’re at risk but not in a screening program, it could nudge them to see a gynecologist. There’s a way to tie these different interventions together into a complex AI-supervised system that could take us one step closer to eradicating cervical cancer, not just in low- and middle-income countries, but globally. This kind of work gets me out of bed in the morning. I’m curious to see whether people think it’s crazy or inspired, but either way, I think it’s worth pursuing.