Heidi Health gets $10 million injection for AI healthcare

Melbourne health-tech startup Heidi Health has announced the completion of a $10 million Series A capital raise. This funding round, led by Blackbird Ventures, saw participation from Hostplus, Hesta, Wormhole Capital, Archangel Ventures, Possible Ventures and Saniel Ventures.

Before this Series A round, Heidi Health had previously secured $5 million in a seed round, also led by Blackbird Ventures.

Founded in 2021 by Dr. Thomas Kelly, Waleed Mussa, and Yu Liu, Heidi Health (originally named ‘Oscer’) operates an AI-integrated platform aiming to address challenges in the primary healthcare sector.

The Australian healthcare landscape indicates potential challenges in the coming years. Recent data from the Australian Health Practitioner Regulation Agency (AHPRA) suggests that Australia will experience a shortage of up to 10,600 general practitioners (GPs) over the next decade.

The demand for GP services is also projected to rise by 58%. One of the highlighted issues facing GPs is the amount of time spent on administrative tasks, which often equals or exceeds the time spent directly with patients.

Heidi Health wants to use AI to free up doctor’s time

With a background in vascular surgery and technology, co-founder and CEO Dr Thomas Kelly identified potential time-saving applications for AI within the medical administrative field.

The Heidi Health platform provides a few different options for standalone doctors, as well as for clinics. With its standalone transcription product, doctors are able to utilise the Heidi tool to record consultations, and that information is then used to generate patient files, notes, consultation letters and more. This is quite similar to the PatientNotes platform that we reported on a couple of weeks ago.

“Every conversation [a doctor] had with their patients can sit in Heidi, and we can use that information for future visits,” Dr Kelly said to SmartCompany.

It also offers a Clinic AI solution that aims to address the practitioner shortage by offering automated pre-consults that patients can do at any time that the doctor can then review and make decisions on. For example, which patients need to be prioritsed.

It is also able to offer self-managed bookings and payments to the patient to further save time. According to Dr Kelly, this is useful for understaffed clinics, especially those is rural and regional areas.

Balancing doctor-patient confidentiality while improving the platform

At the present time, Heidi Health is using a mixture of large language models (LLMs) due to the requirements needed by the Heidi platform.

“So for the pre-consult, we actually run our own models. We have multiple different open source models like LLaMA, but there’s others as well,” Dr Kelly said

“For transcription right now we still use Microsoft — we use OpenAI’s GPT-4. But I’d say within the next two to three months we’ll probably be running our own model to do pretty much all of the standalone clinician tasks as well.

Dr Kelly says that you need to be able to swap to your own model in order to scale because third-party services aren’t stable enough to be a 100% uptime product. And that’s especially true for a product that is dealing with medical visits – accuracy is essential

Dr Kelly also confirmed that Heidi does utilise patient information to train the pre-consult model, but says it is a private and ethical way.

“We fine tune it on thousands of consults where we where we’ve taken the history beforehand, and then basically create our own model,” Dr Kelly said.

“That was part of the reason we went into GP so we spend quite a lot of time doing like all the relevant compliance to get double opt-in from both the clinician and the patients.

“So [with] patients, lots of them do opt out of training. But luckily for us, some patients are happy to try to improve the service, and same with the doctors.”

Dr Kelly says that all information is de-identified and double-encrypted, both during storage and in transit. It is also stored on local servers and will continue that practice when it pushes into the US and UK.

“Lots of security and compliance for the team, but the de-identified information is really what we use for fine-tuning and training the model to be really good at those GP-related tasks,” Dr Kelly said.

What’s next for Heidi Health?

Kelly says that the fresh cash injection will be used to expand its team of doctors, designers, and engineers. It is also looking to increase the number of clinics and GPs using the software in Australia before ultimately branching out into the UK and US.

He also spoke about some of the upcoming features that will be released in the near future. The first is a chat to patient record.

“You can talk to Heidi about every historical visit that you have to the patient. We’re trying to encourage people to record their visits and start building up this bank of clinical memories,” Kelly said.

Kelly says the benefit for patients is that their doctor will have a richer context for you as a person and won’t miss things that may have been briefly mentioned in past visits.

“If they’re busy, hungry, rushed they might not remember when you mentioned this important symptom from a little while ago.”

Heidi also confirmed that its pre-consult tool — which currently has to be set up by the company itself — will be available for clinics to set up autonomously over the next few months.

In the future, Dr Kelly wants to see Heidi use information to prompt doctors with reminders when seeing their patients.

“It could say ‘remember the last time you saw Tegan, you said you’d follow up on this thing but you didn’t mention it in the conversation. Maybe you should remind her of this thing we haven’t checked,’” Dr Kelly said.

Another example given is if a doctor doesn’t do a particular gold standard thing with patients with coronary issues, Heidi would be able to suggest doing it.

“Clinical decision support and superhuman memory — both for clinicians and patients — that’s the long-term thing we want to build.”