Artificial intelligence could transform healthcare from simply applying medical evidence to continuously generating new insights and improving patient outcomes.
A leading digital health expert has claimed it could be unethical for healthcare professionals to deliver care without continuously monitoring patient outcomes – but artificial intelligence will transform how medical evidence is generated and applied.
Professor Claire Sullivan, Director of the Queensland Digital Health Centre at the University of Queensland, told the Public Health Association of Australia’s AI and Public Health Conference in Brisbane that healthcare needed to move beyond simply applying the findings of clinical trials.
“I’m asking you to shift from evidence-based medicine to evidence-generating medicine. So not just taking evidence from wherever and translating into your own practice, but generating your own evidence as you practice.”
She said it will be the natural evolution of randomised controlled trials where the results was given to practitioners to translate into practice. However the outcomes of applying those findings in practice were not routinely monitored, which she said would never have been accepted in other industries.
“So why are we allowed to do things without constant and continuing monitoring of our outcomes? The answer to that question isn’t really clear to me… but it was likely because we didn’t have the digital systems to do so,” Dr Sullivan said.
She said a learning healthcare system is one in which data is collected from individuals and providers, from EMRs, wearables, and other consumer-generated data to help create better care for every patient.
“I’m not talking about snapshots. I’m not talking about audits. I’m not talking about trials. I’m not talking about a consented cohort. I’m talking about every patient, every time, every day, and this could potentially allow us to monitor people’s outcomes from the cradle to the grave and do better,” Dr Sullivan said.
She gave an example of a heatmap of BMI across Queensland. The data was hosted by Health and Wellbeing Queensland which shows where the hotspots of obesity are but also can help target interventions at a micro, meso, or a macro level, and then see if they work.
“This is not about taking a randomised controlled trial in Massachusetts and translating it across Queensland. This is about generating our own evidence in Queensland about what does work, what doesn’t work, and continuously and iteratively improving,” Dr Sullivan said.
Building the national infrastructure
That vision is also beginning to influence national digital health policy, with the Australian Digital Health Agency establishing a committee to advise on how routinely collected health information could support a learning healthcare system.
Dr Amandeep Hansra, ADHA’s Chief Clinical Adviser – Medicine said that My Health Record isn’t currently being used for research or public health purposes but might be in the future.
“Future use is still subject to decision of government, and we’re working closely with AIHW around that.
“But as digital health initiatives continue to improve the quality and quantity of information, the potential is enormous,” Dr Hansra said.
She said one of the key roles of the committee will be how to build trust in the system.
“Because we need to have trust from consumers if we’re going to be using this data to improve care, and we need a system that learns from everyday clinical data,” she said.
Making the data usable
However, as Marketa Reeves, Program Director, Australian Child and Youth Wellbeing Atlas at the University of Western Australia, said, even when governments and organisations have access to enormous amounts of data, people can’t necessarily work out how to use it properly.
“We know that the next challenge for us is to help people find the information, understand it, and use it in their work,” she said.
She gave the example of how they’ve used generative AI to democratise access and understanding of population-level children and young people data with the Australian Child and Youth Wellbeing Atlas.
The Atlas combines data from more than 90 collections into one public-facing, freely available national platform. The team is now introducing generative AI capabilities to make the information easier to find, understand and use.
Semantic search is already available, while an AI-powered chat function is being tested.
Mrs Reeves explained how a user could eventually ask for childhood immunisation data for Armadale in 2025. The system would clarify whether the user meant Armadale in Western Australia or Victoria before explaining that the latest available data was from 2024.
She said that making the data easier to access and understand is about helping users locate relevant information by asking questions in everyday language if they choose to do so.
“Gen AI doesn’t replace the evidence, the governance, or the human judgement. It simply provides another way for people to discover and find that evidence,” Mrs Reeves said.
Technology is the easy part. Changing the workforce is harder.
Dr Sullivan said the hardest part of a learning healthcare system isn’t collecting the data or building the dashboard.
“None of us in this room have been trained in digital health or artificial intelligence from the start of our career, so we’re going through an awkward moment, an industrial revolution
“We suddenly have to pivot mid-career and embrace new ways of thinking that the crusty old guys that taught me never even imagined,” she said.
She said when you think about the health system, there’s three types of change: point application change, system change, and transformational change.
A point application change could be how a patients use a call button instead of ringing a bell. System changes such as the introduction of online banking needs some education and effort. But transformational change is another level entirely.
She gave an example of screening for diabetic retinopathy which used to involve looking at the backs of patients’ eyes to try to see if there were any dots or blots or proliferation. Now there are AI algorithms that can detect retinopathy instead.
“This is the one that hits the paper. This is the one where the unions get upset. This is where the politicians get involved because it’s true transformational change. And you know that you have transformational change where there are completely new ways of working, new tasks, new skills, and old ways of working become completely redundant,” Dr Sullivan continued.
Related
She said transformational change can result in emotional reactions from people, partly because we haven’t addressed the educational deficit said change is creating.
“We’re just expecting people to adapt and embrace it without actually educating people, and that I think is somewhere that Australia needs to step up and accept that this is transformational change,” she said.
Dr Hansra agreed that there needs to be more AI in training curriculums, and it shouldn’t be a separate subject.
“Digital health impacts every single part of the system, and I think we need to find a way to integrate that,” she said.
Dr Sullivan said digital health will change the way doctors use their skills.
“What digital health does is it allows us to shift from being recallers of facts to being architects of care.
The risks of getting it wrong
There was understandable fear about where AI is going. One audience member questioned Dr Sullivan’s comparison of AI with the Industrial Revolution, pointing to the historic consequences for inequality, public health, and the environment.
“I don’t think we can oppose it. I think the genie’s out of the bottle. But what we have got is the ability to understand there’s unintended negative consequences,” she said.
She said there was currently a profound inequality in the healthcare system and argued that healthcare couldn’t afford to maintain the status quo.
“We need to have a very clear idea of what it is we’re trying to solve. If we’re trying to solve reducing disease, improving treatment, and improving the quality of healthcare aim. Let’s aim for that. I think we’ll be okay. If we are aiming for more money, profit, and ongoing late-stage capitalism, I have no doubt we’re going to end up in trouble,” she said,
She’s optimistic but is cautious about AI healthcare being hijacked by finance or tech.
“It needs to be led by health with the correct values and with constant monitoring for unintended negative consequences and the agility to address them.”