Unlocking mental health therapy with AI

“We are not here to replace therapists, we are here to empower them with human-centered AI. We are here to give them superpowers,” said Alon Rabinovich, co-founder and CTO of Eleos Health, an Israeli startup that is looking to improve mental health and talk therapy by using machine learning. “You know, the ‘state-of-the-art’ tools that therapists are using are notebooks. So the goal is to have all the interesting, all the essence of the sessions already kind of tracked for them. So we are not here to replace them. We are here to equip them with more capabilities.”

“We developed a care intelligence solution that leverages in-session data to help drive both clinical and administrative efficiency,” Rabinovich explained. “We analyze the data of the session and we help clinicians to be more efficient by auto-generating clinical notes, saving them time, and reducing burnout. And we also give clinical insights that are related to the quality of the care, such as subjects or techniques the therapist has been using. And the whole vision is to deconstruct this realm to be more personalized and predictive, with the use of technology.”

Eloes’ software relies on the therapy session’s audio, generating insights and information meant for easing the day-to-day life of the therapist and improving the patient’s treatment. “We are focused on the content and after that, on the voice,” ​​Rabinovich said. “In the literature, there are a lot of researches that say that there is a high correlation between voice or prosody or frequency in the pitch to progress in therapy, and those are the things that we are detecting here as well.”

“The session recording is transcribed, and with that and the audio, we use different models to gather insights. On the audio side, for example, there is the session’s pitch level and the anomalies spotted in it, so that is one example of acoustic features. Combine that with the content, which you can draw insights from, like what were the most discussed subjects, what were the emotions expressed, what techniques did the therapist use,” he added. “There is also the time-saving aspect, reducing the burnout from documentation, which is a really huge pain point for them. Clinicians say that documentation is the bane of their existence. They usually spend on average one day a week only on documentation.”

Eloes,which secured $6 million in its Seed round this past September led by Israel’s largest health-tech fund aMoon, was founded in 2020 when Rabinovich, CEO Alon Joffe, and COO Dror Zaide participated at Reichman University’s (formerly the IDC) Zell Entrepreneurship Program. “We knew that we wanted to work together and we researched a variety of industries that we wanted to kind of assess if we want to build a startup in,” Rabinovich recalled. However, it was each of their personal backgrounds and experiences with mental health treatment that helped shape their decision.

“Both Alon and Dror have experienced firsthand their soldiers dealing with mental health conditions, Alon in the IDF’s unit 669, the combat search and rescue unit, and Dror as a drone operator. They both have people that are dealing with really severe PTSD. And I had a very big love in my life that is dealing with a borderline personality disorder.”

“We all kind of had this patient perspective or being close to a patient that was not progressing in their mental health treatment. Something did not seem right in their process of getting well. And after digging, after doing a lot of deep dives into the market, we understood that mental health is still some kind of a black box, that is very hard to get a glimpse into what is working and what is not and for who,” he said.

The three brought in Dr. Shiri Sharvit, a clinical psychologist, to serve as the company’s Chief Clinical Officer heading a team of seven, in-house, clinical psychologists tasked with introducing and inputting the clinical knowledge used to train Eleos’ AI models, validating the outputs of the models, and ensuring the use of therapy best practices.

To date, Eleos has already deployed its solution with over a dozen leading provider organizations such as Solvista Health, Achieve, Yeshiva University, and ReachLink. “Health systems and community behavioral health clinics are the two main segments,” Rabinovich stated. “And we are really proud to have 0% churn. I mean, no organization has left us, yet. So they all receive something from the product, and they see that they benefit from it.”

When discussing potential issues or fears regarding the use of AI technology in something so personal as therapy, Rabinovich clarified, “we do need to be HIPAA compliant. HIPAA is the GDPR (general data protection regulation) of healthcare, a set of guidelines that you have in organizations, in terms of the infrastructure and the way you encrypt the data, how do you save it, for how long.”

“So it is a combination of culture and security, culture-wise, it is a matter of having people having ownership over how they deal with it in the day-to-day. No one is accessing the data if they do not need to. All the passwords are very strict and everything is accessed through two-factor authentication.” he added.

Since COVID changed our global reality, like so many other fields, mental health had to adapt to lockdowns and social distancing. However, unlike other professions, therapy also had to deal with the growing and evolving needs patients had due to the pandemic. “We started working and then COVID hit, and demand skyrocketed, there was a tsunami for mental health services. And we understood that to improve the effectiveness we have got to make the system more efficient,” Rabinovich said.

“It definitely accelerated the market in terms of being more accepting of technology. There were therapists and users that started using Zoom, started doing remote therapy. They had to adjust to a digital world. It definitely helped, business-wise or product adoption-wise, having more engaged users and it really accelerated the market. We had to make a lot of fast decisions and adjust the product to be more, remote first. Doing quick integration with Zoom, and with other telehealth solutions such as Microsoft teams or Google meet,” he added.

When asked about the company’s next goals, Rabinovich describes a future where AI can help spot specific and immediate needs. “The next step, which we are having discussions about with our customers, is creating a therapy understanding engine and more predictive models. One such example is having a suicidal risk score, a predictive score. We can take a few sessions and try to understand what is the chance that this patient has suicidal thoughts. I mean, what are the chances of this therapy working or not, trying to predict success.” However, he reaffirmed the product will only supplement human therapy and not replace it. “There are early signs that those things are doable. And we are very optimistic and confident that we have the capability because we have reached the places we reached. We really need to be careful with those things and not be very decisive unless we are very confident of this.”

“We will not replace a human in the loop. This is very important to say, we are just giving them aids and signals that they might want to take a look into,” he concluded.