That DOHDA and the ADHA are rabbits in the headlights when it comes to the speed with which healthcare AI is presenting them with problems that they can’t yet solve is, apparently, par for the course of this trend internationally. We need to change quickly.
At a briefing this week by one of the most accomplished and long-serving digital health government policy developers and regulators in the world, we learnt something that is in some ways a relief, but also just a little terrifying.
Steve Posnak, principal deputy national coordinator for health IT for the Office of the National Coordinator for Health Information Technology in the US, indicated that healthcare AI is running at a speed of change that is impossible for any government policy or regulatory body to keep up with.
And we aren’t just talking about the fast-moving and well-funded entrepreneur class scaring us all with the sort of disregard they are giving traditional regulatory healthcare frameworks or even legal frameworks such as respecting the Ts and Cs of incumbent software vendors.
Mr Posnak outlined a broad ecosystem in the US where there is rapid and massive experimentation and development by both the software vendor community and the smart provider institutions – the hospital networks, the HMOs and others – and by individual states.
Vendors and provider institutions, even the public ones are realising that this technology represents a generational step change in healthcare delivery and the well led organisations are learning that waiting for government to provide any sort of guidance is wasting precious opportunity.
For the private sector it’s commercial opportunity. For the public and not-for-profit provider institutions, who would normally wait and see, the opportunity for productivity, workforce optimisation and satisfaction and patient wellbeing, is too obvious to sit and wait.
When asked what the ONC – which is the closest thing in the US to our Australian Digital Health Agency – was doing to try to manage such an eclectic and disruptive situation from a regulatory standpoint, Mr Posnak indicated that any attempt to formulate new policy or regulation to reign in explosion of experimentation was neither feasible nor probably the right approach at the moment.
He outlined a few reasons for his logic:
- The ONC regulatory framework, although not addressing the detail of a lot of what is unfolding, was at a high level, still very relevant in terms of how both vendors and providers behave in terms of safety and risk, and that the organisation was monitoring as much as they can very closely against this existing framework;
- That the upside of what is going on is clearly very big along with whatever downside that might emerge, but it was in the better interests of the system that the innovation going on was allowed to play out, within certain limits of risk and safety of course;
- Things are changing too rapidly to attempt any sort of regulatory framework in the traditional sense, at least at this point of time.
The ONC of Health IT first moved on trying to understand AI better as far back as 24 July when it established a role of Chief AI officer alongside a chief data and information officer. These roles were then transferred into the US Department of Health and Human Services at the beginning of the second Trump administration to give the role of AI even more focus and to try to understand how to approach whole of system governance.
Australia and the ADHA are so far fairly notable for what they haven’t done in terms of taking the AI revolution in health a little more seriously.
The most substantive change made re AI and the agency was to appoint a dedicated expert advisory group on the safe use of AI in healthcare as a part of the National Clinical Governance Committee for Digital Health (NCGC-DH). Internally the agency has made AI the job of one of its key IT managers, but this is a bureaucratic move to align the agency with Australian government’s responsible AI use policy for government agencies, not the creation of a strategic leadership role.
In other words, so far in the federal government we have established a committee reporting to a committee with a remit to advise not do anything, and no leadership roles in AI to speak of either in the DoHDA or the ADHA.
By comparison to the US Australia’s central health IT agency is well behind on understanding structural AI changes and governance and you’d have to ask why.
It might have something to do with that pesky $2 billion sunk cost My Health Record thing.
Although we talk of the ONC as an equivalent to the ADHA, the ADHA does both policy and regulation and infrastructure build – eg, the My Health Record. The ONC does not build anything and in a manner that leaves the organisation a lot less conflicted.
Australia has ignored much of the good work done in the US around interoperability and standards and decided to persist with the My Health Record as the centrepiece of our interoperability strategy.
The ADHA built the My Health Record, has to maintain it, and unsurprisingly, does not want to push it to the side in any way moving forward. Hence, we now have a giant contract out to Telstra Health to somehow make the thing functional for the system – because until now, as a largely centralised store of PDFs and other clutter which isn’t accessible, it’s been in relative terms to the sort of goals in interoperability kicked in the US, useless.
Saying this gets a lot of people upset of course, even now some clinicians who say its become quite useful to them in certain circumstances.
It has. But you don’t know how good something is if you don’t have it. Everything is relative and what they have in the US is generationally better than what we have.
Current ADHA and DoDHA strategy is a plan for us to have both: a centralised government held and maintained database of everyone’s record – the My Health Record – and in the future, something equivalent to the US, which is a distributed point to point network where every point of the system is discoverable and accessible to every other point so meaningful patient data can be accessed in real time when it is needed.
The US system is vastly complex and far from perfect. But is managing to handle billions of data transactions daily and its working. If Betty BBQ has a record in a hospital in Boston that is needed for a GP visit in Florida, you can usually – via a vast networked universe of HIE nodes – get it.
Related
The beauty of this system is obvious – although maybe not so much to our health department and ADHA: data is best curated by those who its most important to, so leave it at its various spots in hospitals, community hubs, path labs, GPs and the like, where they need to maintain it very well for their operational excellence, and grab what you need, where you need it when you need it.
That’s a system that is the near opposite of what we currently have: get everyone to waste their time and money pushing bits of data to a master central record, make that record in a manner that is hard to access for nearly everyone, and a cyber security nightmare, and then attempt in real time to get that information back out to either a patient or provider when they need it.
This explanation feels like very old ground for this publication.
We spent a lot of time and words over the past decade trying to explain this very basic problem with our approach but eventually gave up when it appeared that both the DoHDA and the ADHA were going to spread their wings and move to the US system and incorporate what was useful in the My Health Record (it does have some good uses, including the legislation around it) into this new way of doing things.
So we all waited, mainly for the Department and the ADHA to do what the US did, and legislate for all vendors and providers to standardise on a common set of FHIR and open AI protocols and in time all be able to share meaningful patient data in real time with anyone.
We are still waiting. The DoDHA points to pathology uploading to the My Health Record and tells us we are on a journey and will get there.
But there is something very smelly about the time we are taking to get there when the US actually arrived there nearly a decade ago now.
A lot of industry observers think it might be that Services Australia, which is the data controlling juggernaut of data and payments on Medicare, Centrelink, My Health Record, DVA, Tax, Super and the like, has a big problem with letting go of control of patient data – for reasons of control, but also around security and cost.
Apparently the DoHDA doesn’t agree with this position and that is the reason we are being so slow in moving to a distributed system of sharing data by default as they can in the US. But Services Australia is an IT and data juggernaut with a lot more power than Health.
There you have it. Australia is decades behind healthcare data sharing because one very powerful federal department isn’t happy to go along with our health department and attempt to make healthcare far more efficient in Australia and actually let patients have access to their data directly from providers like now occurs in the US.
That’s a reasonably big diversion from what’s with AI and health in Australia but it was a necessary diversion to explain perhaps why the ADHA and the DoHDA aren’t all over AI like they should be.
They have their hands tied behind their back on data sharing by Services Australia.
Is that still a good reason to not appoint some sort of AI flying squad with real teeth to oversee in detail what is changing beneath them though?
Probably not.
The ADHA and the DoDHA are so far sticking with its vast set of strategic plans, all of which do not contemplate how rapidly AI is changing the rules of engagement.
What else might be wrong?
This is a revolution in healthcare that is going to require a particularly savvy and wise set of government leaders to make their way to, and somehow come to a form of understanding of and management framework, over time.
Who are those leaders?
The ADHA just lost its leader of the last five years quite suddenly.
The next leader of the ADHA is going to tell us a lot about whether the government is serious in engaging in the AI health revolution.
For now, for many governments of the world it’s chaos because every well laid plan – our National Digital Health Strategy as one of many examples locally – is now up in air, and potentially even irrelevant.
Our Australian Digital Health Agency has around 17 separate plans running up into the 2030s, has spent already about $3 billion on infrastructure, policy development and plans, since its formation and has probably about another $2 billion or so in infrastructure plans – the national Health Information Exchange may take that all by itself if we went ahead with it – that may now be not fit for purpose.
The good news?
Some much larger and better organised governments are struggling in much the same way we are.
The bad news?
Despite what our federal government would have you believe, we have not got a good technical foundation in place to manage this crisis, and, more critically, we still don’t have in place the right core principles around how we should share health data and who we should share it with.
Outside of the major problem we have between the DoHDA and Services Australia, we’ve come to this position through a lack of qualified leadership in key departments and agencies and as a result we’ve allowed policy and digital infrastructure to be run by bureaucrats.
Bureaucrats aren’t a bad thing per se. We have some great ones and in the right job, they are transformative.
But the government would do well now to find someone who is a leader, who understands the technology intimately, and if at all possible, who has coalface experience of the system as a clinician.
Someone perhaps like the CDC recently appointed: Professor Zoe Wainer – a clinician with intimate coalface experience of the system, public and private, roles running big provider organisations, and lots of leadership credentials.
It’s time for someone to run the ADHA who has felt and still feels the pain.