Forget ambient scribing, AI receptionists and decision support, the real seismic change in how AI is going to affect GPs is the new health AI platforms in the hands of your patients. Here are some things that could happen.
Before I get going on today’s rant there are a few things you should know about me and technology.
New technology is nearly always hyper-competitive business-wise, and very often, rapidly destructive of whole industry sectors.
I have been in “old” media most of my life.
The internet has pretty much killed most of the big global “old” media companies and even the giant ad agency networks now, and replaced them with digital platforms like Google, Facebook, YouTube, Amazon and Linked In.
Old media people like me still exist but we are like cockroaches eking a living out in a new dystopian digital media wasteland, feeding off scraps and scuttling to the dark corners as much as we can, away from the burning eye of the big global platforms that will kill you in a heartbeat if you get too exposed to their distribution power.
So new technologies, especially ones that come on suddenly, like agentic AI, nearly always result in bouts of severe anxiety for me.
Occasionally I’ll even have a panic attack.
I’m a dial-telephone-head-in-the-sand-and-hope-it-all-passes-by sort of guy.
Some people are quite the opposite.
Twelve years ago I interviewed a young bright energetic woman called Melanie Perkins about a new company she had started which had built 750,000 users in its first year. She seemed like only 16 years old at the time, albeit a smart 16-year-old (she was actually 26).
Then, she was still a humble graphic design student out of WA, with a big idea, who didn’t think a lot of Adobe’s Indesign – the main paid art design software for media and agencies at the time – but who saw how the internet was democratising better, cheaper software for consumers.
Canva is a $65 billion company today with 240 million active monthly users.
An AI-can or an AI-can’t?
In the classic Robert Rodrigez western, Once Upon a Time in Mexico, there’s an iconic scene where Johnny Depp turns to his Mexican offsider who is getting a little nervous as the situation is deteriorating way off plan and asks him:
“Are you a Mexican, or a Mexicant?”
GPs – at least the ones who choose not to put their head in the sand – might now need to be asking themselves, are they AI-cans, or AI-can’ts?
It’s going to be harder for healthcare professionals to be asking themselves this question than almost any other market which has been subject to digital transformation.
Healthcare has been a market systemically resistant to digital transformation: it has huge risk profiles, massive information disconnects and silos, is highly regulated and governed, and funded mostly by government.
Unlike media, travel, finance and other markets, it’s been stuck for a long time in one very big and messy digital bog.
In this respect healthcare providers have been somewhat shielded from significant professional digital change dynamics.
Over the years lots of government policy planners talked a big game, and attempted a few breakouts, including early iterations of the My Health Record, but progress was mostly glacial.
There has been too much complexity, governance, cost and risk.
People who came to health thinking they’d make their fortune in digital mostly regretted it after a couple of years after coming face-to-face with the complexity, regulation and the huge asymmetric information power imbalance between patients and doctors (and hospitals), which disarms the main power of digital change in other markets, the power of consumers (patients).
The biggest selling point for the big digital platforms in other markets like finance, media and travel, was democratisation of consumer access and information over the old business models in those markets.
Health sector resistance to digital transformation
Health’s glacial movement in democratisation of information power for patients began thawing as a result of covid.
Government had to let go of control in a lot of areas and relax rules and regulations to get things done faster and more efficiently, especially in respect to information sharing with patients.
Some things that got done were amazing for the patient (and some providers): our national e-script infrastructure for instance.
After covid settled down, however, so did the rate of progress in information access and sharing between providers and between providers and patients.
But the government saw a lot of opportunity in better data flow to patients, and between the Australian Digital Health Agency, the federal government and quasi-independent government constructs like Healthdirect, plans were made to put in place infrastructure that was targeted at empowering the patient a lot more from an information and access point of view.
This effort only really started with the new federal Labor government in 2022 so if you’re a GP, you may have expected some decent progress after three years.
There hasn’t been, not at least in terms of anything you’d notice meaningfully day to day.
From a digital perspective, virtually nothing has changed for GPs in this last three years – same software (with a few upgrades and now AI scribing et al), same connectivity to local specialists, hospitals and pathology (ie, pretty bad) – nothing major going on from a user perspective. There is a fair bit going on under the waterline by the vendors to upgrade for web, but it’s not noticeable, really.
Big plans but nothing much built yet
Most of what has been done by the ADHA has been big plans for big health infrastructure builds, not the builds themselves, so far.
There’s a lot of good goals, a much better understanding of the landscape and some good guesses at how to manage it better moving forward.
It’s just that there is so much that is so old, complex and wrong. So many provider and government platforms remain operating as silos of inaccessible information between the silos, and then, of course to patients.
So there’s been a lot of thinking and planning.
We’re going to atomise and make more cloud accessible the My Health Record, we want to build a giant new centralised directory type resource of all providers and connectivity protocols (Health Connect Australia) and a national Health Information Exchange (you’re going to need to Google them all if you don’t know what they are).
In the meantime, the Department of Health, Disability and Ageing has been doing some important groundwork by getting our government to start changing laws around how providers who have access to their own siloed information on the patient have to start sharing it much more openly with the government (the My Health Record) other providers and patients.
We actually have a few of these new laws in place, which is why pathology labs have to send their patient data directly to the My Health Record now, and soon, hopefully, other major silos, like the GP patient management systems and hospital EMRs will, by law, have to make their data a lot more open and accessible as well.
The DoHDA is also doing some important work in nationalising and making far more efficient laws around patient consent so data sharing between all our healthcare data silos is streamlined for everyone.
Healthdirect (now 1800MEDICARE), which isn’t a government agency, but it is owned, funded and directed in a manner by all governments, has made some massive strides in making the system more accessible to patients via its triage and information service –which is rapidly morphing into a single national digital front door for all patients, and a collector and analyser of important regional patient behaviour data.
Everyone reasonably happy then? Things finally seems to be happening?
But not much has changed … yet
We have plans, we’re changing laws for software vendors and providers so they build over time for better access, and sooner or later we are going to start building some of this major digital plumbing infrastructure – although it’s not really until 2028 that any of it will start to come online for us all.
If you’re a specialist, GP, or hospital manager you could easily be thinking, “whatever, I’ll believe it when I see anything in these plans that actually helps me and my patient in their journey through this mess of a siloed and fragmented system better than what has been happening for the last 10 years or so”.
There are some notable exceptions to this slow-moving but progressive dynamic from a few energetic (and optimistic) software vendor innovators, but, other than the new AI tools, they are only spotted randomly around the system at the moment, in part because most of these vendors are getting little to no help from the government … yet.
Some examples include:
- Consultmed – a simple cloud-based e-referral and “guidance and advice” platform which can help facilitate better connectivity from a GP platform through to a specialist into a hospital and back out into the community;
- E-scripts – an ADHA-assisted national project delivered through covid to make scripts digitally accessible everywhere;
- Personify – a patient information service that integrates with hospital in and outpatient processes to make a patient journey easier;
- Telecare – an anywhere e-script pad for specialists and hospitals in and outpatient services, and a virtual specialist service for peak period help in the cities and remote regions with low resourcing and no local specialists;
- MediRecords – a fully cloud-based GP, specialist and hospital in and outpatient admissions platform with its own patient portal (I’m a non-executive director of this group);
- Halo Connect – a Best Practice-driven connection platform for lots of applications GPs use such as booking engines, Cubiko, Consultmed and payments, but which should allow connectivity from BP to specialists and hospitals over time;
- And lots more … they’re all mostly creative local vendors funding themselves who see the utility of cloud data sharing and storage, and of new resource platforms such as FHIR.
Even with all these small private innovators running around – and putting aside for now what some of the AI tools are starting to offer in terms of scribing and admin efficiency – most GPs, specialists and hospitals aren’t seeing anything particularly transformational at this point of time.
If you are me, the dial-phone-ophile, and perhaps an AI-can’t, you’re sort of okay with all of this.
After all, change is always work, even when it’s small, but when it’s big, it can be panic attack inducing.
Okay, deep breath.
Things are about to change, fast
ChatGPT Health, which launched about a week ago on trial to selected users in Australia, and Claude Health, which is scheduled to launch within a month or so, are odds on to start changing everyone’s lives in health in all sorts of ways very soon.
By everyone I mean, government, government agencies, providers, funders, and most of all … patients.
In simple terms it feels entirely feasible that in a short amount of time, all that positional information power and governance that doctors and government have, through their regulation and control of patient data, and how it flows in our system, seems set to be thrown pretty quickly out the window.
And things are going to get messy for a while for everyone.
Related
If you’re initial response as a GP is to think, “hey, we’ve been here before, I got on top of patients and Dr Google, I’m pretty sure ChatGPT Health is the next iteration of Dr Google”.
It’s not.
The canary in the coalmine for medicine is that the consumer-side AI platforms are already significantly reshaping major business sectors that have as a common denominator historical data and information asymmetry between the consumer and the profession.
Some highly visible examples include lawyers, consultants, accountants and insurers.
Another hint is that a lot of big commercial information and data services are experiencing the biggest single step decline in business in their history. Services like the global legal data and risk group Lexis, the giant consumer-side medical database WebMD, the professional medical decision support service Up To Date, from Wolters Kluwer, and Thomson Reuters financial data.
On the flip side all these big commercial data and information groups are rushing to imbed AI into their databases to somehow retain or improve the value to their clients.
But they are all fighting a generational change in how information is being created, stored and analysed and shared on the consumer side.
They are all freaking out too, by the way, and madly redoing strategy.
Some will disappear pretty quickly, just like print classifieds and the Yellow Pages did when the internet got going.
GPs don’t need to freak out, they are here forever, but they shouldn’t do my preferred trick and bury their head in the sand either. This is something that needs to be engaged with, and as quickly and professionally as can be done.
I’m going to admit that I have no real idea of how this new giant patient-side data platform might actually end up disturbing the thus far relatively digitally peaceful lives of doctors and government, because the variables are so wide and unknown.
But here’s a few wild guesses to get us started, based on what we already know about this technology, how powerful it is, and how fast it is going.
Patients will drive you nuts soon
If Dr Google annoyed you, ChatGPT Health is going to really test your mettle, especially for the next year or so while it gets itself up to speed and keeps making dumb mistakes.
It’s already being used at record levels overseas.
But I challenge you to tell me you’ve not heard of an instance where ChatGPT has been used by a patient very effectively to create a more meaningful interaction with their doctor or other health professional.
I’ve heard a few life-changing and life-saving examples.
Yes, you will have also heard the stories where its advice is misleading and even dangerous. But these new AI models learn from their mistakes. They are getting better at speed.
Okay, maybe some patients are going to get themselves killed here, and that’s definitely not a good thing.
But, according to the Australian Institute of Health and Welfare, in 2023 there were 28,112 potentially avoidable deaths accounting for almost half (48%) of all deaths for people aged less than 75. Add to that the amount permanently disabled in some way as a result of mistakes by doctors and the health system and you are going towards 100,000 mistakes per year.
How much worse is AI going to be? Especially if doctors embrace the problem and make sure they engage with it with a view to taking the best of it and protecting their patients against the worst of it.
And if it keeps on learning, centrally, it should get better and better, right?
But it’s going to be tricky striking a balance of consumer knowledge to what you think as a doctor and have worked with in the past.
Government plans to help are upside down now
All our government’s plans to build a lot of really expensive infrastructure, both digital health plumbing, and new hospitals based on our old paradigm – that paradigm being healthcare pre-consumer-side agentic AI – are now out of date.
These plans were pretty good given the constraints all governments have politically and risk management wise. But they won’t work anymore – at least a lot of them won’t work.
Here’s are just two potential examples of what might happen:
- The government’s objective was to make healthcare data flow between the various provider-side siloes (things like GP patient management systems, hospital EMRs and pathology labs) and use that accessibility to have the data follow the patient wherever they went. The new AI platforms present an opportunity for patients to have the data, not follow it. There will of course still need to be a lot of government work to make databases more accessible to patients, but in the new paradigm, government and providers won’t be the “keepers of the data”, the patients will be. Stuff like this might make giant and complex projects like the national Health Information Exchange (a complex solution to share data between various provider data silos) redundant. The patients should be able to get the information wherever they go and have it with them for when they have a health system encounter. That’s the reverse of what happens now.
- Healthdirect or 1800Medicare could easily over time become largely redundant for what it is doing now and plans to do into the future: collect, analyse and use masses of patient-generated data and other data sources to triage patients to the right service when they don’t know what to do, and even be a patient’s chaperone in their journey around our complex system. It won’t happen immediately, but if you think about how AI collects and learns information, it might not be too long. It would be a lot shorter and probably safer if the government and Healthdirect help. The same dynamic is going to smash into our major booking engines too as they do things that Healthdirect does. They have a lot of new plans to deal with agentic AI and dovetail into it. So does Healthdirect. That should tell us all something.
Will government be an AI-can or an AI-can’t?
Here’s the problem for doctors, government and people like me who don’t like any internet technology other than Spotify (and I’m a bit down on that now, but not for technical reasons).
The money, IP, data and learning capacity of these new AI platforms is incomprehensively bigger in scale, capacity, speed and potential than anything government can do. These platforms are going to be huge, hold more data than any commercial or government silo, and rapidly be able to learn and improve.
From any technology perspective we can’t compete.
And the possibilities of such a data store, accessible to our patients, are endless.
This isn’t to say we all step back and give everything over to the AI platforms. These people have proven themselves to not be ethical, and they’re greedy.
But we do need to stop what we’re all doing for a minute and think very carefully how we take the good with the bad, and manage the bad, because we know from Facebook, Google, Amazon and all the others, there is going to be quite a bit of bad.
But there are massive aspects of what is happening here that it is a waste of time and money to fight against.
For one thing, taking a stance that medical information is the domain only of the professional provider, and then government, for risk and safety reasons, is thinking from the past.
Government and providers can’t control that aspect of healthcare soon.
Have a quick think about all the talk of governance and guiderails on AI that government has been contemplating for the last 18 months … and done nothing. They can’t even start. It’s going too fast. Government is necessarily too slow.
They – we all – need a new plan.
These AI platforms are going to be a huge part of our patient’s lives, and no one is going to stop this happening.
As much as it depresses me to say it, they win. We can’t beat them, we need to join them in an intelligent and forward-thinking manner. A manner that has the patient really at the centre of our thinking.
Because AI platforms sure as hell don’t have patients at the centre of their thinking.
