Rumour has it Heidi is quietly building its own EMR. What would that look like?
As far as medical software platform plays go, Heidi isn’t short of ambition, bluster or money.
According to the company’s own marketing it’s building an “AI clinical operating system”. That sounds like just about everything a doctor wants and needs to go about his or her daily business.
To realise such a bold vision, Heidi will at some point of time need to make a decision to combine an electronic medical record, all existing AI clinical workflow it’s built, and maybe even reach out into the realm of patient administration and workflow.
Which means that Heidi, with its AI focus and capital backing, is either coming for existing EMR platform players and represents an existential threat to many of them, or it will partner up with existing platform players in a manner that allows it to be all things to all doctors in its various markets.
But they can’t do it with their current “light” EMR integration strategy.
Heidi says it integrates with a substantial list of EMRs including Epic, Best Practice, MedicalDirector, Gentu, MediRecords, Athenahealth, Zedmed, eClinicalWorks, and EMIS – this is most of the big EMR players covering Australian GPs and the big US enterprise players.
The eClinicalWorks integration, for example, runs via a middleware platform called Vim, embedding Heidi as a floating widget inside the EMR so clinicians push notes into the patient chart with one click without leaving their existing system.
It has also built two integration models: a lightweight Heidi Widget for web-based EMRs requiring minimal engineering effort, and a more comprehensive Heidi API for deeper configuration – aimed at telehealth platforms and partners who want to embed Heidi’s capabilities directly into their own product.
But nearly all of these integration models are “light” – they don’t access the patient’s longitudinal clinical record. There’s a reason for that. The EMRs know that once they give up that gold, the AI players will eat them from the outside in. They will end up as a database layer only, and all the action will migrate to the AI clinical workflow.
Even in those cases where Heidi has been able to access a clinical record via a user-consented API, the vendors are wising up.
MediRecords, the only real native cloud-based patient management offering in Australia, which has the biggest share of virtual providers in Australia for one vendor, and is at the epicentre of JP2060 – the new cloud and on-premise health ecosystem for defence – did have that user-driven API arrangement with Heidi.
But their initial integration contract is up now and they’re giving the whole situation a lot more thought.
Best Practice gave the situation with Heidi a lot more thought some time ago when they made a strategic investment in Lyrebird and granted this AI provider exclusive access to the clinical data layer in its platform – a deep integration as opposed to all the light ones listed above by Heidi.
If you use Heidi and Best Practice, over time it makes no sense not to switch to Lyrebird, unless you intend getting off Best Practice, and most Best Practice users are pretty loyal.
No matter how you look at it, that was Best Practice drawing a line in the sand. It sees the existential threat that Heidi represents to its future, 80% market share of GPs in Australia or not. And it’s teaming up with Lyrebird to develop its own combined EMR/AI layer clinical workflow platform.
This more or less marks the end of the first phase of the AI healthcare vendor platform revolution.
Up until now, Heidi and all the others have been systematically surrounding the EMR incumbents by trying to own the layer that clinicians actually trust and interact with most: the consultation workflow itself.
Some clinicians literally weep with just how useful a scribe with a summary and AI knowledge layer is to their consult.
It’s the first time in history that doctors actually love their software and given the terrible brand that the likes of Epic and Cerner have created for clinicians in the past few decades, adding to workload rather than helping it, that makes Heidi a standout brand (other AI players too) with a lot of cache – cache which Heidi means to parlay into bigger and better things, including the addition of EMR functionality.
The EMR problem for Heidi
Without the longitudinal record as a reference when you start an AI consult, a doctor is more or less flying blind. They have to either go into their EMR separately first to pull the patient record manually, or use their memory.
One action is very inefficient, the other is just unsafe.
It’s the reason why deep integration is an inevitable strategic must for Heidi’s ambition to rule the clinical workflow world. And ultimately for any other scribe-plus AI workflow plays as well.
Deep EMR integration – proper bidirectional write-back to the patient clinical record, not copy-paste – is genuinely hard.
EMR data formats vary widely across vendors, with different standards and no consistent protocols, making integration slow or inconsistent. Compliance with HIPAA, SOC 2, GDPR and region-specific frameworks adds further complexity.
And Heidi already has an interesting recent emerging PR problem – amplified gleefully by some competitors – for occasionally being flaky and losing whole consult sessions, something that would be unforgivable in an EMR.
Heidi has nowhere to go but to somehow address their EMR issue. It has been playing nice so far, but it can’t keep that front up much longer.
But, from where it is now to where it’s trying to go is Alice down a particularly deep and twisted rabbit hole for degree of strategic difficulty.
Related
Why is Heidi raising again so soon?
Would you back into the next round of Heidi funding based on what you see so far, including its apparent burn rate?
Before you did, you’d probably want to know why Heidi is currently building a new funding round, not even eight months after it raised a cool $100 million on top of its prior raisings.
It could be one of two things: either Heidi is spending extravagantly and is already seeing the bottom of the last barrel of money – ouch, hope not but possible given how fast it is expanding – or it is building a war chest to take on its significant EMR platform problem.
Building an EMR isn’t anything like building AI workflow functionality – scribes, receptionist, billing and analytics et al. The reason we are seeing so many players is because the build is reasonably cheap and the degree of difficulty low.
Building a robust properly structured web-based EMR, even for GPs and specialists, though probably a lot less expensive than it once was, is architecturally far more complex than any of the AI workflow layer work, with lots of traps for young players.
That isn’t stopping some people from trying, of course. There is a lot of trial, error and learning going on in this new AI world. One GP has vibe-coded his own practice EMR (we’ll get back to you on that). Claude obviously has a lot to answer for here.
Where and how do they start
If Heidi is contemplating building an EMR what and who it is building it for – there’s the entry to that rabbit hole. One entry maybe but once you’re in, there are lots of different tunnel choices, some very deep with no apparent way back.
The most lucrative for Heidi market if it could get there is the hospital enterprise market and then in big developed countries – the UK, the US, the Middle East.
But the incumbency and strong capital and customer position of the major enterprise hospital EMR incumbents makes that sort of play seem way out of bounds, even for Heidi.
It isn’t going to try to take on Epic or Oracle in the enterprise market for hospitals.
Not yet anyway.
But it might be thinking it can take on a group like Solventum – a global group that makes AI coding and analytics middleware that talks to the big EMRs and has its own scribe.
Solventum apparently has the contract to be the AI coding and scribe solution above the NSW statewide Epic installation. If that’s true, Heidi is out of NSW public hospitals altogether and probably Queensland in time too.
But if Heidi could replicate the sort of IP that Solventum has, and combine that with its “most loved” brand status, and its mass of existing hospital doctor users that love the brand, maybe there is a way I in.
Maybe.
Would it be planning something more simple, more generic and less expensive for the non-hospital markets, in just one market to start, perhaps?
Start small, get your sea legs, before you take on the bigger overseas markets in some way.
Take on the 80% GP market share PMS gorilla in Australia, Best Practice and its alliance with Lyrebird, for example?
You’re on home turf too where your brand is strongest.
Even this is going to be pretty hard.
Heidi is in 116 countries and even if it just picks the big ones and tries to include Australia because it has a big footprint here, it’s going to need to be a pretty generic attempt at something to work in more than one country over time.
The functionality of an EMR varies quite significantly between countries based on all sorts of local health system variables like payments, system structure (such as federation), coding standards, security protocols and local laws and regulations.
And you while you might prove a point about AI and a custom integrated cloud EMR as a precursor to taking on bigger markets, you’re not going to make the sort of scale money that your VC masters are voraciously chasing.
How hard to take out Best Practice if you did try?
Best Practice has about 130 “partner solutions” integrated to its core EMR product.
Some of them are duplicated applications fighting for distribution within the BP ecosystem – like say Healthengine vs HotDoc – but even paired down for the ones people don’t use much and the duplications there are 40-plus core applications that combine to make Best Practice’s functional integration suite a pretty impressive competitive moat.
And although people have whinged about Best Practice (and Medical Director) over time, it is a much-loved local brand, renowned for its customer focus and its cranky but likeable founder.
It’s got a strong brand too.
But would Best Practice, even with Halo Connect surrounding it – a great catch up to the cloud strategy for a tech stack with a lot of old bits in it – be as agile as a native cloud-architected EMR purpose-built today to talk to a set of AI workflow layers.
Probably not.
Best Practice is a series of modules stitched together over time to compromise between the old server world where it started and the evolution of the cloud where it now finds itself.
Even today, most of its users are operating off their own servers in practice.
As a result, the innermost operating module has some old code and data structures inside. The architecture of such a long-serving and evolving set-up has lots of component parts stitched together which, though now being cloud-enabled in many ways, makes it a bit clunky to exchange data and to integrate with. That Halo Connect exists is testament to that.
This does make it at least in part potentially vulnerable if someone really wants to go at it with a much more modern native cloud-architected EMR play. One with all the cool additional AI workflow bits.
Such a product would offer more speed, more seamless data sharing across and within locations, superior mobile and virtual utility, fast and inexpensive integrations, no more hardware or maintenance costs, superior security, much easier installation protocols, and relatively seamless integration to patient aligned applications.
This sort of functionality would make it immediately attractive to large practices and corporates because of data management and sharing capability and agility and efficiency, especially for practices with evolving multi-channel care delivery models.
The big private health insurers both now have large practice networks whose technology is pretty much legacy and not fit for purpose for their future multichannel care models, which will include not just GP services but connections to virtual care services, hospital in the home, aged care and even the NDIS.
If they want to optimise their networks for modern multi-channel care delivery and realise the potential of the patient data they are sitting they all will need to replatform soon.
This all sounds like a pretty interesting “new” thing for Heidi to build, so maybe it would be an okay play.
Except, this sort of cloud EMR offering already exists in Australia.
MediRecords is most of the cloud product described above – less the AI workflow layers, so, less of Heidi AI cool stuff at this point of time – and it’s pretty bulletproof. It makes up the core of the entire Defence Department’s new cloud and on-premise-based digital interoperability health ecosystem.
MediRecords doesn’t have the scribe, knowledge layer and all the upstream admin workflow layers that we are starting to see from the likes of Heidi, MBSPro and Lyrebird.
But it has quite a bit of AI going on it already, and given what is going on, how long do you think before it will be adding most of the other cool stuff? After all, it’s the easier bit to do.
No one should count out Telstra Health and the remnants of Medical Director and Helix yet either. There’s a pretty good team over there and they have their heads entirely around modern interoperability and the technology partners to come back to market in some way.
And likely Gentu, from Magentus, which dominates the specialist PMS market, is thinking much the same way.
The AI stuff is cool, but easier to build, especially if you have an agile cloud data platform as your working base.
All of these folk – Heidi, MediRecords, Medical Director, ZedMed, GENTU – are deciding now to either build out their own AI workflow layers or whether they can pick an AI dance partner as Best Practice seems to have done with Lyrebird.
Heidi has a brand advantage but it needs to use it quickly
If you are a Heidi lover, or follower, if it did build an EMR in competition with Best Practice, and put all its groovy AI stuff in one integrated solution, how tempted would you be to trust its innovative and forward-leaning brand proposition and jump ship from Best Practice?
You can answer that in our survey HERE if you like (we’re asking other questions in the same survey).
A big factor in play for Heidi is brand.
Heidi is a new generation of doctor software brand.
Traditionally doctors hate their software. Not with Heidi.
Doctor’s love and attachment to it is on another plane to anything ever seen before in the medical software sector. I’ve heard of doctors weeping over how much difference it makes to their consult.
But I’ve heard similar for Lyrebird and MBSPro.
Which suggests Heidi is much more a cool new brand and clever early mover than a unique piece of AI wizardry and coding.
Notwithstanding, it’s brand promise is a pretty good one if it can deliver it: it’s an idea for how you could work as a doctor in the future, with much more time, agility and freedom, and a promise of continual innovation of your workflow tool.
According to the recent HealthEd AI scribe survey (sample 1535 GPs) about 20-30% of GPs use scribes day to day in Australia and of this 20-30%, 45% use Heidi and 45% use Lyrebird.
Technically that’s only 15% of GPs using either Lyrebird or Heidi regularly so far, which is maybe 5000 individual GPs using each (all the scribes put out marketing suggesting usage is much higher).
That suggests that the AI/EMR land grab has a way to go yet and you will almost certainly see other EMRs like GENTU, MediRecords and Medical Director either building out their own AI workflow layers quickly or trying to partner with Heidi, MBSPro, or one of the other established scribes.
The maths don’t seem to make the idea of developing an Australian EMR for GPs worth it for Heidi, when you consider the US and UK market potential.
But one thing now seems clear. In any of the scenarios here, Heidi is rapidly approaching its lack-of-deep-EMR-integration problem.
It could be why Heidi appears to be on its next funding round already only eight months after its last raising of $100 million.
Either it is burning money at a very extravagant rate, or it’s worked out it needs an EMR integration strategy and it understands it’s not going to be cheap so it’s lining up that money now to get it done.
It had better get its next round before SpaceX shares start dropping in the next few weeks (predicted by pundits), and certainly before the whole ChatGPT, Claude and SpaceX AI stockmarket grift comes apart at the seams, which seems likely in the next six to 12 months.
When that happens the money for AI application layer plays like Heidi will almost certainly dry up for a while.
What’s going to happen then?
My bet?
England 3-2 over France in extra time? I have no idea.
Declaration of Interest: The author is a non executive director of MediRecords
