Montreal’s ELNA Medical emerges as a healthcare AI powerhouse

MONTREAL – ELNA Medical Group, Canada’s largest network of integrated medical clinics, has become the sole Canadian partner of an innovative project in the U.S. that’s creating the world’s first, general-purpose LLM for healthcare. The effort is being organized by Silicon Valley-based Hippocratic AI, which has already tested its AI-driven Large Language Model (LLM) and claims that it outperforms OpenAI’s ChatGPT when it comes to medical tests.

“There has been a huge explosion in AI activity recently, with tools like ChatGPT getting a lot of attention,” said Maxime Cohen, chief AI officer at ELNA Medical. “But Hippocratic AI is the first to focus on healthcare, specifically on patient-focused applications. And for this reason, it has outperformed ChatGPT on the majority of medical exams and certifications.”

Indeed, Hippocratic AI ran tests and found that its LLM beat ChatGPT on 105 of 114 healthcare exams and certifications. It also scored higher when it came to healthcare specific vocabulary and measurements of bedside manner.

Dr. Cohen, a full professor specializing in artificial intelligence at McGill University, leads the AI work at ELNA Medical. (He earned his PhD at MIT, in Cambridge, Mass., worked at Google AI and taught for some time at New York University.)

He noted that ELNA is co-developing the LLM with Hippocratic AI, along with several other healthcare partners in the United States, including Cincinnati Children’s Hospital, Universal Health Services (UHS) and Capsule.

One of the first applications the group is working on has been dubbed the “super-nurse” by Hippocratic AI’s co-founder and CEO, Munjal Shah. The super-nurse will be able to follow up with patients who are discharged from hospital, making sure they are taking their medications, that their wounds are healing, and that they are recovering well.

If not, the computerized system, using voice or other types of communication, will be able to arrange follow-up care for the patient – thereby improving medical outcomes and reducing hospital readmissions and trips to the ER. ELNA Medical is helping to develop and test the application, making sure that it’s ready for prime time and applicable to the Canadian market.

As Dr. Cohen observed, in addition to creating the algorithms and systems to accomplish this task, ELNA’s network of I.T. professionals and clinicians are making sure that’s it’s done in a safe, responsible, and effective manner and one that protects the privacy of the patients.

Moreover, “there’s always a human in the loop,” asserted Zachary Stauber, chief strategy officer at ELNA Medical. “Nothing happens without being reviewed by a nurse or doctor.” At a time when generative AI systems have been found to make errors, the human review is there to make sure everything is accurate and done in an ethical way.

For its part, ELNA operates nearly 100 medical clinics and points of care across five Canadian provinces. It also has a home monitoring division, a cardiac monitoring company, and a virtual visits service. The company began as a medical laboratory organization and continues to operate labs.

Further to the work it’s conducting with Hippocratic AI, ELNA is developing its own AI-driven tools and applications. Indeed, it is emerging as one of Canada’s powerhouses in the development of real-world solutions that use AI to improve the delivery of healthcare.

“We’re applying AI before, during and after the appointment,” explained Stauber.

He gave examples of how the team at ELNA Medical is working on developing solutions in all these areas. Before a patient comes to a medical appointment, he noted, the AI-driven system will be actively combing through its records, checking on whether the patient is due for tests such as a mammogram or colonoscopy.

“It can look at the age of the patient and other factors, but it can also cross-reference the patient’s family history, to see if there’s a history of cancer, for example, and whether the patient should be doing certain types of screening,” said Stauber.

The result of this analysis will be, of course, brought to the attention of the physician so that it can be discussed during the appointment.

Moreover, the system will also analyze the type of appointment that’s needed for the patient, along with the patient’s preferences. “It looks at whether the appointment can be done virtually, at home, or if it must be at the clinic,” commented Stauber.

Another major component on the roadmap is a decision-support system for clinicians. It will be able to assist the doctors, advising them on best practices and the latest findings in medical science.

“With our lab, we’re at the forefront of the science,” said Stauber. “We know the latest tests that are available and the newest markers. We’re building all this into the system.”

And post-appointment, the system will be following up with the patient, to make sure they got their medication and that they’re taking it appropriately, or whether they scheduled a test that was recommended by the doctor, and more.

This is much like the “super-nurse” under development with Hippocratic AI, where the LLM is following up with discharged hospital patients.

In an interview with Hippocratic AI’s Munjal Shah earlier this year, Shah observed there’s not enough follow-up with patients, due to the severe shortage of nurses – in North America and worldwide.

However, an LLM could conceivably do the work, something that’s now under development.

Impressively, ELNA’s AI-related projects don’t stop there.

Dr. Cohen noted the company is also working on AI-driven dashboards that monitor a host of operational variables at its clinics across Canada. “We’re tracking all kinds of metrics at our clinics. Because to understand and improve, you first have to measure,” he said.

As an example, they’re tracking lead times to see a physician. This metric has been noted by the Canadian Medical Association and the Fraser Institute as one of the chief complaints of patients, as they often have trouble getting an appointment with their family doctor in a reasonable amount of time.

While in other countries, patients can obtain an appointment the next day, Canadian patients must often wait weeks.

“We’re benchmarking our results against the Fraser Institute’s rankings when it comes to lead times,” said Dr. Cohen.

AI can help clinic efficiency by predicting and spotting no-shows and late cancellations, a troubling occurrence for medical offices. “It’s a big, wasted opportunity,” said Dr. Cohen, as a no-show takes up a spot another patient could have had.

The AI system can be trained to predict which patients might not show up, based on past histories and other factors. It can send out alerts to these patients, reminding them of appointments and asking for confirmations.

Dr. Cohen mentioned that another pain point for clinics is the amount of time doctors must spend on note-taking. But today, generative AI has shown that it’s excellent at listening to the patient-physician encounter, transcribing the conversation and creating a medical summary.

“The physician can then fine-tune the summary,” said Dr. Cohen.

“It saves a lot of time for the doctor and can ultimately increase access to more patients. We’re going to be piloting this in the coming months.”