AI tools to optimise billing can increase revenue, but what are the limitations…

Despite the assertions of people like Dr Margaret Faux, GPs are over seven times more likely to underbill than to overbill a consultation.

A previous Healthed survey showed two thirds of GPs underbill on a daily basis rather than risk getting it wrong. This is reflected in 2022 Health of the Nation report findings that “almost half of doctors either avoided providing some services or claiming patient rebates out of fear of Medicare compliance,” RACGP President Dr Michael Wright recently noted.

Given the complexity of MBS billing rules and the thin financial margins that many clinics are operating on, it’s no wonder several AI-powered solutions have entered the market.

Their marketing typically suggests they help save doctors time, simplify billing, and optimise revenue.

But along with the upsides, it’s important to understand the limitations and potential liabilities if you decide to implement one of these tools in your practice, experts say.

Why use AI to optimise billing?

AI tools can help better match remuneration to a GP’s workload.

“The work’s being done. It’s just essentially the data’s not getting captured appropriately,” explains GP registrar Dr Patrick Gough, medical director of AI-enabled billing optimisation software MediBetter.

“We wanted to ensure Australia’s primary care sector remained strong, and we felt the best way to help fund GPs was by optimising the current MBS infrastructure. AI has allowed us to do that,” Dr Gough says.

Chantelle Brott, a lawyer (non-practising) and the co-founder and CEO of AI-powered clinic solution mAIscribe, says optimising billing in a compliant way usually requires several steps, including educating staff about MBS item numbers and which ones you can and can’t put together.

During and after consults, doctors need to “figure out the one, two, three or more item numbers that are applicable — either in isolation or in combination — such that I actually get paid for the work I have rightfully done and documented.”

When you’re busy and time-pressured, it can be easy to “go level B and move on,” she adds.

GP Dr Paul Tescher, mAIscribe cofounder and CTO, says AI can help with these tasks — without ever feeling “rushed and tired like a human.”

For example, he recently did a new chronic condition management plan for a patient and arranged a COVID vaccination. At the end of the consult, he almost missed claiming the COVID vaccine eligibility check, but the AI helped pick it up.

“So I’d done all the work, the patient had had the vaccine, and because I was busy and maybe tired by that point of the day, I could have easily missed that if the AI hadn’t prompted me.”

How does it work?

These solutions work by listening in on consultations and suggesting item numbers based on what they hear.

They are trained on huge amounts of data, Dr Tescher explains, so they’re very good at recognising patterns. In this case, the patterns are “certain themes within a consultation that match to the 700 pages of the MBS,” he explains.

An algorithm then works out which item numbers apply — and which don’t, he says. “Because there’s no point a doctor putting in work thinking I can claim an item number only then to get a Medicare rejection.”

Some tools, like mAIscribe, can also write and organise your notes. Others, like MediBetter, do not — although you can also use it with AI scribes via a browser extension.

What are the benefits?

MediBetter performed early testing with 50 GPs who first shared how they would typically bill each of five cases, then reassessed after receiving recommendations from the app. They achieved an average 43% increase in billings using the app, Dr Gough says, although he stresses this was a small case selection and sample size. However, it showed that providing MBS prompts can work, with even a fraction of that increase equating to thousands to tens of thousands in extra revenue annually, he notes.

You can also save time and effort, he adds.

“Prior to this, people would be navigating MBS Online, which is clunky and not very user-friendly. As well as that, people would have cheat sheets they carry around—and that would only be a small selection of item numbers. Whereas by having it as a single platform, you’ve always got that information, no matter where you go,” Dr Gough says.

Ms Brott says they’ve seen billings increase by 10% per hour at an already high-performing practice.

Dr Tescher recalls reviewing a patient’s mental health plan and dealing with some acute issues in one consult.

“That needed close to 30 minutes all up, and some doctors, their instant instinct might be 36 and move on to the next patient, because that’s straightforward and easy. But that’s underutilising the MBS schedule,” Dr Tescher says.

With mAIscribe, he worked out that items 36 and 2712 were the optimal combination.

“That ended up being an extra $70 or $80 of Medicare rebates that otherwise would have been left on the table,” he says.

What are the limitations?

Dr Owen Bradfield, Chief Medical Officer and Acting Medicolegal Lead at MIPS, says it’s important to understand the limitations of any software, particularly in relation to Medicare billing.

Even when using an AI tool to support billing optimisation, knowledge of the MBS item descriptors is still required.

“And one of the key principles of Medicare billing in the Health Insurance Act and associated regulations is that what’s billed to Medicare must be clinically relevant. And that comes down primarily to the clinical judgment of the clinician,” Dr Bradfield said in a Healthed podcast exploring the topic.

“So I think it’s important to make sure there’s always a human in the loop, checking the output, checking what’s suggested against your own knowledge and understanding of the MBS, but also bearing in mind other considerations under the Health Insurance Act as well.”

Dr Gough agrees, noting AI alone will not solve all billing issues.

“This is going to be the GP’s own knowledge as well.”

He suggests double-checking factors such as patient eligibility, as well as how notes are structured and whether appropriate timeframes have been allocated to each section to ensure the correct co-billing criteria are being fulfilled.

“And I always encourage all the GPs to learn more about the MBS, to attend courses, to do some of the RACGP online modules as well.”

You are responsible for what you bill

David Leach, General Manager of Underwriting & Partnerships at medical indemnity insurance provider Tego, stresses you can’t blame the AI if billing goes awry.

“The practitioner remains ultimately responsible for what is billed under their provider number, and using an AI tool to enhance billing practices could potentially introduce or increase the risk of an illegitimate claim,” he says.

Dr Gough also stresses GPs are responsible for their MBS billings.

“The best we can do is provide that information at your fingertips,” he says. “At the end of the day, in court it’s not going to hold up for the GP to be like, ‘I didn’t know or I used it blindly’.”

Dr Bradfield stresses practitioners have both a professional ethical responsibility and a legal responsibility to ensure they are billing Medicare appropriately.

That’s the case whether AI is assisting or not, he notes, adding that in some cases a practice and a practitioner may be jointly responsible — but the software developer is unlikely to be.

Incorrect claims can trigger audits or PSR referral, though so far there have been no matters involving AI, the PSR told Healthed.

However, the PSR cautions that if errors do occur, it is the medical practitioner that would be liable.

“It remains each practitioner’s responsibility to ensure that all services billed under their provider number are billed appropriately, including ensuring that all Medicare Benefits Schedule requirements are met and that the record for the service is adequate and contemporaneous,” they said.

What are the consequences of incorrect billing?

Mr Leach says if you do come under the PSR’s radar, the consequences can be significant.

“The nature of the investigation and severity of any penalties varies depending on the extent of the alleged improper billing, and can range from increased scrutiny of future billing practices, repayment of overpaid amounts, additional penalties, and potentially criminal prosecution.”

“If a practice or billing company drives the error, the Shared Debt Recovery Scheme can split repayment between the practitioner and the organisation that manages billing.”

For context, in 2023–24, the PSR ordered the repayment of funds totalling over $31 million, and issued 54 partial MBS disqualifications, one full disqualification, and 50 reprimands.

Tips to help avoid any issues

Maintain accurate records

Dr Bradfield says for any audit, the medical record is the single source of truth.

“Ultimately, whatever the GP has billed in relation to the services rendered, Medicare’s only window into that consultation to figure out whether or not what was billed was appropriate and correct is the medical record,” he says.

He notes cases where billings may seem appropriate, but the records don’t actually support what was billed.

“For example, if a GP has billed an item 36 or an item 44 because it was a prolonged and complex consultation, if the records are too brief or they don’t adequately describe or reflect the complexity of that consultation, then that can be problematic.”

Detailed records generated with help from an AI scribe, when checked for accuracy by the GP, can potentially provide “a much more accurate reflection of what actually occurred during the consultation,” he adds, and make it easier to justify billing certain item numbers.

Make sure your notes are compliant

Dr Tescher says optimising your billing can put you “on the pointy end of the spectrum when it comes to being audited.”

“So we want to feel as confident as we can that if we were unlucky enough to come under the watchful eye of a Medicare audit, that we could satisfy ourselves and any auditor that our notes fully explain how and why we can legitimately claim the item numbers that have been claimed.”

Dr Gough agrees, stressing you need to read MBS descriptors carefully and ensure your notes reflect them.

Ensure a human is involved

With any professional use of AI, human oversight is crucial.

“I think there must always be a human checking the accuracy of the output or what’s suggested,” Dr Bradfield says.

Don’t forget patient privacy obligations

Mr Leach points out that if you use AI tools to capture or process patient data, then Privacy Act obligations come into play.

“It is important that practices/practitioners adequately disclose how and when an AI will be utilising their patients’ data and obtain consent where necessary,” he says.

For more information

Hear more on this topic from Dr Owen Bradfield and Dr Patrick Gough on this podcast: Billing smarter, not harder: Tech, AI and compliance in clinical practice – Part 1 – Healthed