The demand for high-quality mental health care outstrips supply, but telehealth has democratized access to care and decreased no-show rates. Generative AI has the potential to revolutionize the field by integrating patients' data to tailor treatment models, says David Mou, CEO of Cerebral.

Cerebral is a US online therapy provider founded in 2019. In 2021, the company raised close to half a billion dollars and was valued to 4.8 billions. Even Olympic gymnast Simone Biles, who withdrew from the Olympic games in Tokyo in 2020 due to personal mental health struggles, and became a public advocate for a new attitude and public perception of how we approach mental health, struck a partnership with Cerebral, joining as an investor and its chief impact officer. 

But then in 2022, things started to shift in the opposite direction for the company due to allegations of unsafe prescribing practices. The Wall Street Journal, Insider and other media publications investigated and reported about these through the accounts of patients and former employees, and the Department of Justice launched an investigation about "possible violations" of the Controlled Substances Act. Eventually, in 2022, the company dropped prescribing controlled substances entirely.

Today, Cerebral is moving forward and is betting on quality mental health provision, with high hopes around enhancements that could be achieved with the help of AI.

Telemedicine vs. In-Person Mental Health

Cerebral already uses AI for patient-clinician pairing and detecting suicidal messages. In David Mou’s assessment, using AI for documentation could save up to 70% of clinicians' time, allowing them more face-to-face time with patients. Cerebral's telehealth platform uses generative AI to tailor messages to patients, reducing anxiety and increasing engagement in care.

Telemedicine increased access to care. In the US, during the pandemic, physicians were able to provide consultations across States. Cerebral today still provides talk therapy in all 50 states in the US and measures clinical outcomes.

AI in Mental Health in Practice

David Mou sees several positive impacts of the telemedicine approach to mental health services provision: lower no-show rates, greater accessibility since an online session is less time-consuming than visiting a mental health professional in her office, and then there’s AI. “I actually think generative AI, artificial intelligence, will completely upend the space for the better for patients. When you see the patient in person, you get a sense of what this person is like. But can you imagine the amount of data you're collecting about the patient through telehealth? It's their voice, it's the rhythm of their voice, their word choice. Did they show up for their first appointment? Did they cancel? Did they hesitate when they texted you? Did they say certain things or did they say other things when they filled out these clinical surveys? You can now integrate all of that information and tailor the treatment model to that patient going forward. And I think that's what's so exciting,” says David Mou.

In his view, using generative AI for documentation optimization is only the lowest-hanging fruit.

Even before 2023 Cerebral was leveraging AI to match patients to clinicians, for example, if patients requested someone with a similar ethnic background, the same gender. Additionally, the company also uses natural language processing to detect through messages if a patient is suicidal or not. “When they send a message like I want to end it all, our machine learning would detect, okay, this is a risky text. And we have our internal crisis specialists that call the patient within minutes. We don't wait for them to call us. We actually proactively call those patients and we're hoping to publish that data,” David Mou says.

Generative AI is Not Going to Replace Therapists Any Time Soon

While generative AI has much potential, it is still far from being used as a replacement for clinical judgment. David Mou explains: “Let me give you a scary example. What if the patient asks something simple and very straightforward, like, I want to kill myself? Maybe ChatGPT could say, okay, this is a suicidal patient here, call 9 1 1. But what if they do it a little bit more subtly? What if they say something like, would it be better for carbon emissions if I do not exist? The factual answer to that is yes. And so the GPT system would say yes. It wouldn't be able to infer that that question might be a cry for help. We're not close to being able to build that human trust yet. Maybe one day we will, but it's not coming anytime soon.”

Future Potential of Digital Therapeutics

Digital therapeutics are still an emerging category of treatments in healthcare with many innovators offering solutions for mental health issues. Many efforts are being made from the industry’s side to include DTx in reimbursement schemes, but that does not address the challenge of how can clinicians be aware of and trust these solutions enough, to recommend them to their patients.

In 2023, it seems that the market is still not ready. Pear Therapeutics, a company founded in 2013, was the pioneer in the DTx space. It created the first Prescription Digital Therapeutic (PDT) to receive authorization from FDA to improve disease outcomes in Substance use disorders. The company went public in 2021 but had to file for bankruptcy in 2023. “We always look for reasons. The biggest reason is, it's been a tough year. It's been a tough couple of years and the macroeconomics have not been cooperating, that's true for companies across the board. But in short, I’m very interested in partnerships and if anyone has an asset or a service or something that really improves the outcomes for any mental health diagnosis, I will have a conversation with them, and I'll say let's try it. Let's give it a shot, if the evidence is really there and the cost-effectiveness is there. What we can do is create a platform where those treatments that are validated can have a place to flourish and we can pick and choose the ones based on outcomes, um, going forward,” comments David Mou.

This is just an excerpt. Tune in to the full discussion.

Show notes

Comment on the last three years and introduction to Cerebral

00:52-01:05

Personal assessment of mental health treatments

05:02-05:19

Demand and supply of mental health services

05:11-06:49

Clinical outcomes and metrics

07:45-08:41

Telemedicine and video therapy

08:56-10:46

Role of AI in mental health

11:24-14:54

Tailoring treatment to individual patients

15:13-17:43

Innovations in mental health care

18:27-19:30

Engaging patients and building trust

20:52-21:46

Data and mental health care

22:24-24:21

Clinical trials and workflow optimization

37:06-38:53

AI and patient communication

40:18-41:19

AI and data generation

44:24-45:49

Workforce structure and clinical guidelines at Cerebral

45:49-47:28

Measuring outcomes and structured data

53:08-53:44