Leading research, government and industry representatives met in Sydney yesterday to develop recommendations on the safe and responsible use of artificial intelligence (AI) in healthcare, convened by the Australian Alliance for AI in Healthcare (AAAiH) and facilitated by Dr Noman Swan AM.
The National AI in Healthcare Policy Workshop, held at Macquarie University, was supported by the Australian Institute of Health Innovation (AIHI), CSIRO’s Australian e-Health Research Centre (AEHRC), RMIT, the Australasian Institute of Digital Health (AIDH), and the Digital Health Cooperative Research Centre (DHCRC).
Attendees at the workshop included representatives from the Department of Industry, Science and Resources, as well as Federal, State and Territory Departments of Health, and representatives from key stakeholder and regulatory bodies.
AAAiH is strongly supportive of the Government’s active interest in developing industry appropriate policy and governance for AI.
A Director of AIHI and founding member of AAAiH, Professor Enrico Coiera, said: “We are developing policy options to not only assist in dealing with the specific challenges of using AI safely, effectively and securely in healthcare, but also to support companies investing in bringing AI to healthcare settings.”
“A whole of system approach will be required from Government, industry and research due to the increasing opportunity for AI to be expanded into Australian healthcare settings, bringing with it not only prospects for enormous benefit, but also areas of concern,” Professor Coiera said.
While Australia regulates AI that is used in medical devices through the Therapeutic Goods Administration, clinicians and consumers need to be aware that any AI that has not gone through rigorous testing is likely not fit for use in patient facing settings, Professor Coiera said.
“Large language models such as ChatGPT are not designed for use in clinical settings and have not been certified to be used safely in any aspect of patient care,” he said.
Dean of RMIT’s School of Computing Technologies and cofounder of AAAiH, Professor Karin Verspoor, noted that privacy is also a critical issue to be considered around AI in healthcare.
“There are strict requirements to obtain consent and protect personal information in healthcare, which for example may prohibit cross-border transfer of data (e.g. to a large language model run by a foreign corporation) where its re-use is not controllable, and may breach local laws,’ she said.
AEHRC CEO and cofounder of AAAiH, Dr David Hansen, said localisation was also an important issue.
“AI models can perform very differently across clinical settings because of patient, service and workflow differences. AI model localisation will be needed for the routine use of many AI systems in healthcare and will inevitably require specific governance such as provider organisation accreditation,” he said.
DHCRC CEO, Ms Annette Schmiede, said Australia excels in healthcare delivery and can be positioned as a leader in AI, however our digital technology sector needs to be supported with appropriate policies, regulation and financial incentives.
“Pressure from larger, more well-resourced, global competitors and lack of regulatory clarity are real challenges for Australian SMEs,” Ms Schmiede said. “There needs to be targeted support to help innovative products come to market, mechanisms to provide access to clinical data to support AI development, and the creation of a National AI in Healthcare Strategy.”
The policy recommendations discussed at the workshop will be released later this year. They will build on current regulatory structures that are healthcare specific and cover three areas:
- Safety, Quality, Ethics and Security governance for healthcare AI
- Supporting the development of a healthcare AI industry in Australia
- Building the capabilities of the healthcare workforce and consumers to be effective users of AI
AAAiH is a national community of practice with over 100 member organisations drawn from industry, health service providers, academia and consumer organisations.