If AI fails healthcare, it won’t be the technology’s fault

5 minute read


So say experts calling for more government involvement in the regulation and guidance of the management and use of AI in the health system.


The Australian healthcare system must embrace AI but experts have called on the government to be more involved in its regulation, governance and data collection. 

Speaking at the Healthcare 2040 Expo, part of Australian Healthcare Week in Sydney, a range of experts hammered home a common message on day one – long-term AI success depends on strong data foundations and governance, not just new tools.  

Governance was framed as part of the architecture of AI, not an afterthought.   

Dr Suzie Miller, director of strategy, quality, and technology of the Victorian Virtual Emergency Department, highlighted that an MIT Media Lab study did a study of American companies using AI.  

“Guess what percentage of those AI projects failed to add any value to the company? It was 95% failed,” she told conference delegates.  

“Was it the technology? Nope. It was humans and lack of governance.” 

Many presenters admitted that no one has their data perfectly accurate, something else government could help with.  

“Federal and state governments have a role, to step in and demand the way that data is presented and gathered,” said Dr Miller. 

“We need to be all digital really, really soon and off paper. That needs to be mandated … we can’t be scanning and faxing anymore.”  

Zoran Bolevich, CEO of the Australian Institute of Health and Welfare, agreed that while the benefits of AI were undoubtedly exciting, there was still a lot of work to do to fulfill the technology’s potential. 

“There will be some significant governance, regulatory, risk management issues that we are still only beginning to understand and haven’t really got appropriate frameworks in place,” he said.  

“We’ve got to embrace this technology.  

“Clearly, it’s offering significant benefits, but there is a lot of work to do, to do it effectively and safely and in a way that really adds value.” 

Governance was also on the mind of Dr Brian Tan, chief health officer for Salesforce. 

“Governance is not a process and it’s not a structure – it’s a culture,” he said.  

“That ability to adapt and to flex and to understand that change is constant in healthcare is going to show up those organisations that are particularly good at governance culture.  

“It’s also going to allow us to move at a greater pace of change,” he said.  

AI governance concerns included bias in models and data sets, particularly when it’s trained on non-representative populations. There were also discussions about the practical elements, including needing to have honest conversations with tech partners about where data is stored, how it’s managed and how outcomes are audited.  

For Dr Tan, it’s all about grounding your data.  

“Grounding your data will cut hallucinations by at least a half in AI, and it will move your AI agent from being a creative storyteller to a trusted partner,” he said.  

There was a lot of optimism on show at the conference. 

Richard Taggart, CEO of eHealth NSW, said there had been significant improvements in diagnostics thanks to AI.  

“We can see a significant improvement in the accuracy of diagnosis and decision making by clinicians when they’ve got a tool to augment their practice,” he told delegates. 

“It is helping to develop new models of care.  

“One model in experimental phase is using AI to analyse a CT scan result in emergency then making a call about whether it’s urgent or whether the patient can go home and await virtual care follow-up from a clinician.  

“These are the sorts of opportunities that AI is now bringing us that could redefine the journey for the consumer and help make the health system more effective,” Mr Taggart said. 

Andy Canning, chief technology officer and managing director of Equal Experts said: 

“When we’re talking about AI moving forward, governance has to become that hygiene part as well.  

“The important bit is, do I understand where my data is failing … understanding the data lineage, where does data come from … if I’ve got two similar pieces of data, but they actually give me different answers, what’s the source of truth?” he asked.  

Jayne Barclay, the national director of healthcare platforms at St Vincent’s Health Australia, said SVHA was constantly experimenting with using its data to integrate with other systems within the hospital to improve patient outcomes.  

“We definitely have a plan that we’re working towards, but we haven’t let perfect get in the way of good,” she said.  

The conference continues on Thursday. 

End of content

No more pages to load

Log In Register ×