Fly on the Wall: The future of AI in health care

Artificial intelligence has seen a meteoric rise in progress over a short period of time, with just about every profession looking for ways to integrate its potential into workflows. The health-care sector is no different, with AI showing promise in making new discoveries and streamlining administrative work. But as is the case with any new technology, experts are approaching it with cautious optimism.

The past few years have seen the advancement of AI tools like OpenAI’s ChatGPT, Google Bard and Microsoft Copilot. These tools, like it or not, are here to stay, driving educators at McMaster University towards preparing medical and other learners about this new reality awaiting them.

Cynthia Lokker, an assistant professor with McMaster University’s Department of Health Research Methods, Evidence and Impact, teaches in the eHealth program – focusing on various digital tools that can be used by future health-care professionals. She has collaborated with Puneet Seth, an adjunct professor with the Department of Family Medicine, who helped build a digital health company (InputHealth) and uses digital tools in his own practice, on a guest lecture series in Lokker’s eHealth class on AI. Seth is also a supervisor on a thesis project for masters’ students focused on advancements in electronic medical records and routinely writes about the advancement of medicine on his blog.

We met with them both to discuss the potential for AI in health care, their concerns, and the importance of training students on how to use it.

How transformative is AI in health care?

Seth: If you look at the entire spectrum of care, I think there’s interventions that can happen all along the way where we’re able to glean insights from AI and make sense of data in ways that human beings cant do alone . I’m blown away by the rate of development of new tools that accelerate how drug development happens.

Let’s say, for example, you want something that blocks the thyroid hormone receptor, there are AI models that can design a molecule that would be able to do that, with the desired shape and the desired minimization of potential side-effects based on what we know about protein synthesis. The AI will tell you how to make that. It arguably saves years of experimenting. That’s just one example of so many that just completely revolutionize how we think about clinical practice and personalized medicine.

Lokker: I can speak from my experience in the eHealth program. We’re getting more and more applicants who are interested in the aspects of data analytics and how can we use the data we have in health care to drive decision making and increase our knowledge.

A growing and evolving aspect of my research is applying machine learning in natural language processing to help support the curation of high-quality evidence to make it easier to access. I do think we’ll get to a place where we will be able to have more precision medicine, where we can identify patients for treatments or risk for diseases.

What concerns do you have about the use of AI?

Lokker: A major concern I have is with representation in data – who’s missing from our data? Another concern is ensuring the factuality of what the AI outputs are for models. We already don’t capture data consistently across systems. We still have problems where data from different systems can’t be transmitted or populated easily. Those are foundational challenges. I think there’s also a culture aspect, too. We have to teach AI literacy so that people can understand what’s going on.

Seth: One hundred per cent, those are some pretty significant ones and I think that the kind of amplification gaps in the data and kind of entrenching the biases can be a really big problem.

Whether we like it or not, the organizations leading this race are very much private, for-profit corporations. And one of the inherently unique characteristics of AI is its dependency on immense computing power, beyond what’s available to public or academic organizations. Additionally, it is an arms race of sorts. Companies and countries are moving full-speed ahead in trying to position themselves as dominant forces in AI and to be at the bleeding edge. However, with speed there is a sacrifice to the inherent impact of the technology on existing structures, like the practice of medicine, and a sacrifice in correcting for things such as bias.

How important is AI literacy and education?

Seth: This is something I’ve become passionate about over the last few years. I co-authored a paper that was published last summer in JMIR Medical Education about the urgency in making data science and AI education a core competency of undergraduate medical education. A considerable amount of my time is being spent trying to push that conversation forward and trying to figure out how to integrate that into curriculum.

The incredible benefits of having something like the eHealth program is that it allows students to immerse themselves in the subject matter to a greater extent. But for the general medical curriculum, there’s not enough being taught. This is resulting in people coming out and scratching their heads trying to figure out what to make of AI. I think that can be a daunting feeling.

Lokker: AI literacy is incredibly important. Interest is currently driven by the students’ desire to want to learn more and to engage either in other coursework or through online courses.

But as Puneet mentioned, there’s ingrained biases in our data sets based on how it’s gathered and by whom. There’s certainly a whole area of responsible AI and what approaches we take to be transparent about what your data is and what the limitations of your data are. We’re teaching students that you don’t blindly believe everything you read. You need to do an assessment of the validity and reliability. For example, I have my students do appraisal of an article and then I ask them to have ChatGPT create a summary of the abstract. This allows them to think critically and assess how good the AI performed in creating the summary. Critical, evidence-based thinking, and by extension, AI literacy, are more important than ever for medical students, and it’s our responsibility as educators to ensure they have all the tools they need to prepare themselves for the real world.