‘Significant concerns’ over AI use for Medicare compliance

The RACGP has expressed fears over the potential for unnecessary targeting of GPs if there is unchecked use of AI for Medicare compliance.
 
In a submission to the Australian National Audit Office (ANAO) about AI use by the Department of Health, Disability and Ageing (DoHDA) for Medicare non-compliance, the college called for it to be avoided where possible due to the risks involved.
 
‘If used, it should be recognised that there is a spectrum of compliance reflecting different practice arrangements, and it is dangerous to rely solely on arbitrary standards set by Government,’ the RACGP submission states.
 
‘This audit also specifically refers to “non-compliance”, which neglects the opportunities presented by AI to identify good clinical practice, reduce red tape and improve workflows in general practice.’
 
The submission, made earlier this month, warns that the use of AI for compliance without a shared understanding of how it is used and the data it is trained upon ‘risks worsening transparency and belief in DoHDA’s assessment and decision-making processes’.
 
While DoHDA states AI can be used for compliance and fraud detection, the college notes it does not outline the ‘specific uses for compliance purposes’ and that the RACGP ‘does not have visibility of how AI is currently used by DoHDA to manage health provider noncompliance’.
 
‘Retaining human oversight of Medicare claiming is essential to avoid the tragic consequences of another Robodebt scheme,’ the submission warns.
 
Dr Tony Bayliss, Chair of the RACGP’s Regulation and Compliance Working Group, said there are ‘significant concerns’ around the prospect of using AI for monitoring compliance.
 
‘Some of those concerns reflect there’s already a lack of transparency about how matters get referred by Medicare to the [Professional Services Review],’ he told newsGP.
 
‘For example, we’re not made aware of the percentiles that would trigger a referral to the PSR.
 
‘Yes, we’re aware of the 80/20 rule and the 30/20 rule. But in the context of other referrals, Medicare has been consistently failing to provide detail.
 
‘The concern would be that by adding an AI process into this, that would further obscure the details.’
 
The RACGP outlines its support for managing lower levels of potential non-compliance through education, emphasising the ‘extremely stressful’ compliance processes.  
 
Its submission also highlights the risk of inadvertently targeting GPs with specific interests whose billing patterns make them ‘potential statistical outliers’.
 
‘Adding AI is fraught with the risk of unintended consequences,’ Dr Bayliss said.  
 
‘For example, what you might see with an ageing population is an increase in longer consults, which may seem concerning to the AI tool, but there’s actually a perfectly valid clinical reason. 
 
‘It’s entirely possible that the AI will pick up on patterns of Medicare billing that are different but appropriate and then refer those to the PSR – and the PSR may not have adequate resources to respond to this increase.
 
‘That would therefore decrease the quality of their work or their ability to resolve matters in a timely manner.’
 
According to the college, the need for peer review ‘becomes paramount’ if AI is used.
 
‘Attempts to enforce compliance through AI tools, without consideration of variables such as patient demographics, could encroach on a GP’s clinical autonomy,’ the submission states.
 
‘Consultation with peak bodies such as the RACGP will therefore be key as the use of AI continues to grow.’
 
As well as probing DoHDA’s current governance and readiness to deploy AI models to manage non-compliance, the ANAO is looking into whether officials are effectively monitoring the impact of AI use among health providers and the ‘assurance arrangements’ in place.
 
On the latter topics, the RACGP is calling for more support for doctors to appraise new AI technologies for safety and efficacy.
 
‘AI tools that produce inaccurate and incorrectly interpreted information in the health sector could result in adverse patient outcomes including death,’ the submission states.
 
‘This is not an area where AI use can be allowed to expand without robust safeguards.
 
‘We strongly encourage the development and promotion of further resources specifically focused on AI and Medicare compliance, outlining benefits, risks and other considerations.’
 
Dr Bayliss also raises a further issue: that doctors could be disproportionately targeted for non-compliance due to the abundance of MBS information covering their practice compared to other groups.
 
‘There’s a lot more data for GPs, so you could see a world in which GPs get investigated relatively more,’ he said.
 
‘It’s important that all professions using Medicare are treated equally and fairly, and my concern would be that the differences in data between GPs and other groups could mean that this may not happen.’
 
The ANAO is due to table its findings from the audit in September.
 
Log in below to join the conversation.

ANAO Australian National Audit Office MBS Medicare Medicare compliance Professional Services Review PSR