AI is playing an increasingly important role in the advancement of healthcare, with Australia's health scene taking slow but steady steps to leverage this powerful new technology. 

In 2021, the Australian Alliance for AI in Healthcare published a roadmap to provide guidance on how industry could leverage AI to be resilient and deliver safe, cost-effective healthcare to Australians. That roadmap has since been updated, and a panel discussion at the Digital Health Festival (DHF) touched on what the way forward looks like. 

The road ahead

"At the moment, there's not enough knowledge in digital health to begin with in AI, specifically in healthcare settings. And so the recommendations in the roadmap are around assisting professional bodies to really develop codes of practice and eventually do some training as well," said Karin Vespoor, professor and executive dean of the School of Computing Technologies at RMIT University.

She added that Australians need support in understanding the implications of the use of AI in the context of their care, and also be able to make decisions around the use of AI in their care. 

"In the solutions space, want to support the development of local AI companies to become globally competitive. And we really need to think about something that we call algorithmic sovereignty – having the ability to develop technology here in Australia so that we can mitigate some of the risks associated with foreign services.

"For instance, sending our very sensitive healthcare data overseas. We need to retain some amount of local capability in order to use these tools effectively."

Ensuring the careful adoption of AI

Andrew Pearce, VP of Analytics and Global Advisory Lead at HIMSS, drew the conversation to how large vendors like Epic were embedding AI into their products – and whether they would be leading the charge in the safe adoption of AI. 

"I think the most important thing is when [smaller vendors] are doing this we have a high frequency of update releases which enable us to enact a lot of change [versus] dropping a huge change that has been in development for many years," said Kai Van Lieshout, founder of Lyrebird Health.

"The problem is where do you get that data about the indigenous consults and the slight biases – you're not going to get it in a 'sandbox' environment as I like to call it.

"It's so critical that these tools are built out in real time in slow and steady increments [but] with deep collaboration with community and practitioners for it to be safe."

Prof Vespoor added: "If we have systems that are based on Epic's data – mostly data where models have been trained to understand the language of US electronic healthcare systems – and we try to take them and just plonk them into Australia, we're going to find important differences in things like local documentation practices, local drugs lists, and medication lists."

She went on to say that these differences might not seem like a big deal but when you're building a large AI system that has integrated and synthesised all this information, the small differences start to add up.

"The ability to focus in on finely constrained problems is something we can really only do in the context of a smaller company that is willing to invest in understanding the particular details and context [of that health system's jurisdiction.]"

Exemplars around the world

Regarding countries Australia can learn from, Dr Stefan Harrer, program director at CSIRO's Data 61, explained that there is no one country that has figured it all out and doing it perfectly.

He shared: "The US has the 21st Century Cures Act which essentially [lays] the foundation around access to data and managing the patient data ecosystem that is underneath all of this.

"I would say it's a very advanced and an incubator for innovation, where we see a lot of trial and error, where we see a lot of exploration around using AI, and connecting it to those data streams. That's something that although is not AI in itself, has something that is a little bit of a make and break criteria of whether you are successful in building your AI systems and whether you have your data management, electronic health record management organised in an efficient, interoperable, scalable way."

On the regulation side of things, Dr Harrer pointed to Europe, saying that the European Union's AI Act is a forward-thinking piece of legislation governing AI. The legislation seeks not to regulate the technology but regulates AI as part of specific use cases and application scenarios, rates these based on a risk framework, and then investigates how these risks can be mitigated and managed. 

"Can we learn from all of them? We should work with all of them. I think we should talk to all of them. We have a lot to offer in Australia as well when it comes to safe exploration and an advanced healthcare system that's eager to explore this technology.

"I think it's a give and take. An environment where we all can learn from each other. And this is also the most important thing. We all need to talk to each other, right? And no one has invented the right way yet. But we all together, I think, can get there if we see and check what we do right and also what we do wrong."