AI in infectious disease care: Trust, adoption and barriers in remote and underserved settings

That is where I come in. My research looks at the policy, program, and systems factors, right through to the individual behavioral drivers, that influence whether people engage with healthcare recommendations.

Q. What barriers need to be addressed for AI-enabled diagnostics to gain trust and adoption in remote and underserved communities?

There have been many excellent presentations at this conference about the potential of AI. I attended the sepsis session yesterday, for example, where speakers discussed how AI could be used across the patient pathway.

The challenge is that many of these studies are conducted in very high-resource settings—places with funding, staffing, infrastructure, and time to support implementation. But when we step outside those environments and into lower-resource settings, the realities can be very different.

My current work includes case studies in the Solomon Islands and rural Bangladesh. In the Solomon Islands, many communities live on small remote islands far from the capital. Healthcare is often delivered through very small local clinics with minimal equipment, limited staff, and sometimes no reliable electricity. Patients may walk for hours to reach care.

In rural Bangladesh, healthcare may be provided by informal drug vendors or village doctors with little formal training or regulation. Treatment decisions are often based on past experience and what medicines happen to be available.

So before we talk about implementing AI in these settings, we need to understand the foundations: access to electricity, digital infrastructure, digital literacy, workforce capacity, and whether AI is even a local priority.

AI may eventually play an important role—but in many places, there are more immediate system barriers that need to be addressed first.

Q. How can AI tools be designed to strengthen biosecurity and surveillance while still supporting patient-centered care and community engagement?

I think AI absolutely has future potential in this area. We have already seen mobile health technologies used successfully in many low- and middle-income countries, so it is not unrealistic to think AI systems could also become part of community care.

But they need to be designed for the realities of those settings. That means tools that can function with limited staffing, tools that can be used by people with different levels of training, and tools that can operate offline rather than relying on constant internet access. They also need to be informed by local data and local community knowledge. That is critical.

If you build an AI system in a high-income country hospital, it may draw on years of electronic medical records and highly structured datasets. In many of the settings where I work, records may still be paper-based—or not documented at all.

For example, in some of our antibiotic stewardship work in Bangladesh, antibiotic use is not systematically recorded. So yes, AI may eventually help with surveillance and monitoring, but building those systems will take time, investment, and local collaboration.

Q. What role do behavioral, cultural, and health equity factors play when implementing AI-enabled diagnostics for respiratory infections in remote populations?

They all play a central role. I recently read a quote saying that AI has the potential to support equity, but if poorly implemented, it could also deepen existing inequities. I think that is exactly right.

Historically, we have seen many examples where a research group enters a setting, pilots a promising technology, and then leaves. Once they leave, the system falls over because no one is there to maintain the software, support the hardware, train users, or sustain the program. So implementation is not just about the tool itself. It is about the package around it: training, maintenance, trust-building, governance, and long-term support.

We also need to think carefully about ethics, data sovereignty, and ownership. Where is the data coming from? Who controls it? How is it being used to inform decision-making?

Those are major issues, and in many cases they need to be addressed before we even begin discussing large-scale rollout.

Q. What are the next steps for your own work in this area?

For me, it always starts with the end users. We need to ask communities and healthcare workers what their priorities are. If a new system is not solving a problem they see as urgent, then adoption is unlikely, no matter how advanced the technology may be.

We need to listen to context, build with users, and co-design solutions that fit the local environment. Across public health, we know interventions are more successful when the people using them help shape them.

I am also very interested in who is designing these technologies. Are tech experts from low- and middle-income countries involved? Are local AI skills being developed? Or are systems still being designed in high-income countries and then pushed downward?

My sense is that too much is still top-down. Ideally, we should be investing in local expertise and enabling countries to build the systems they actually need.