In this episode, Dr. Tom Zaubler discusses how AI can improve access to behavioral healthcare and give practitioners more time to spend with their patients.

Hi, I’m Tom Zaubler. I’m a psychiatrist and chief medical officer at NeuroFlow. NeuroFlow is a technology company focused on creating an infrastructure to improve access to behavioral healthcare in various settings. We do that in a number of methods, and I’ll describe them today. We work with payers, providers, the VA, and many other clients. My background is focused on integrating psychiatry into a range of medical settings, and the last 30 years of my career have been spent trying to improve access to psychiatric care.

Here’s the thing, just to level set: we know that most people with behavioral health disorders do not receive any care at all. In fact, in any given year, about 60% of people with a mental health disorder do not get care. I always ask people to pause and think about that for a second. If you think of cancer, diabetes, or heart disease, it would be unimaginable if we said the majority of people with those problems did not get care.

There are many reasons for that. Some of it has to do with structural issues in healthcare systems. Some of it has to do with stigma and bias. Some of it has to do with supply chain issues and the shortage of providers. Where does AI come in?

AI is a way that we can help improve access and democratize care so that it is accessible to the entire population, not just parts of the population that behavioral health providers have served historically. That’s a huge undertaking.

AI provides ways to economize on humans providing care. What I mean by that is if you have a patient seeing a psychiatrist or psychotherapist monthly, every two weeks, or every three months, whatever it is, AI can provide a tremendous amount of support for patients seeking care in between those appointments so they can get better. There’s a lot of work that can go on in between sessions.

The other thing is that when you look at how behavioral healthcare is delivered, most behavioral healthcare is not measurement-based. How individuals are doing is not being tracked carefully or assiduously at all, and too often, there isn’t an adequate evidence base. AI can ensure that symptomatology is being tracked in a very systematic way by delivering assessments asynchronously; providing clinical decision support based on interpretations of those assessments; being able to go into electronic health records and pull critical data, distill that data to provide a very clear picture of how individuals are doing, and provide that support to providers. It can make behavioral health providers a lot smarter and more effective.

Then, of course, it can provide that end user support: the support for patients, for individuals seeking care, so that they can not only get care from their providers but also, as I said, in between appointments. Patients can do digital cognitive behavioral therapy, and a whole range of other modalities of treatment that can be helpful. Those are just a few examples of how AI complements care delivered by humans, standardizes that care, and serves as a forced multiplier,because there is such an enormous workforce shortage in this country.

About 70% of all counties in the United States report a significant shortage of behavioral providers. AI is a way that we can expand access. The purpose of AI is not to replace human beings. It’s not to supplant the relationship between a patient and a provider. It’s intended to strengthen that relationship.

I think often there’s a misconception that AI is going to make humans irrelevant in behavioral healthcare or maybe healthcare in general, and that’s just not true. People used to be concerned that radiologists, for example, would no longer have a place in the healthcare ecosystem because AI was getting really good at pattern recognition. We found that AI has not replaced radiologists. It’s just made them a lot more efficient, smarter, and more effective because that pattern recognition helps to facilitate better judgments and decision-making by radiologists, at times capturing things that they might have missed.

The same is true in behavioral healthcare because AI can provide all sorts of information in a short timeframe to providers. It frees providers to spend time with their patients, get to know their patients better, develop closer relationships, and develop empathy and rapport, which is so critical to ensuring that individuals get better and adhere to the treatment plans, medications, and psychotherapeutic regimen.

AI allows for a very quick distillation of past care by going into electronic health records. We do that at NeuroFlow: AI can go into electronic health records, pull critical information, distill that information, and provide summaries of care and points of reference that may speak to the level of acuity or concerns about safety. All that can be done seamlessly and almost instantly, as opposed to a provider going through an electronic health record, having to click from one note to another, and spending a lot of time trying to get this information. When all that information can be provided seamlessly and almost instantly, that saves a lot of time. That’s more time that providers can spend with patients.

Similarly, providers can get a lot of clinical decision support through AI in terms of the level of acuity and what type of medication or psychotherapeutic modality should be considered. In addition, providers can help standardize care by utilizing cognitive behavioral therapies that can be done digitally and asynchronously by patients, who then report back to the providers what’s happened between sessions.

It makes the whole process more powerful and effective, and offers a much greater economy of scale in terms of time. The bottom line is that the relationship between the patient seeking care and the provider improves because there’s just more time to spend with that patient, getting to know them, as opposed to doing all these things that AI can do.

The privacy and security are really important. There has to be adherence to HIPAA privacy laws, and that’s critical at NeuroFlow. We take great efforts to ensure that we’re in full compliance with HIPAA. All sorts of governance goes into data management. It is critical to make sure policies are in place in terms of how data is collected and where it goes. There has to be transparency so that patients know what data is being collected and what is being done with it.

You want to collect the critical information. You don’t want to collect information that is not important and doesn’t need to wind up in a database somewhere. Making sure that there are protocols in place to protect privacy is really important.

We also need to be very careful with AI. We’re getting sophisticated with creating large language models (LLM), and the ways that we can provide all sorts of behavioral care delivery through AI. We have to make sure there are not inherent prejudice or biases that get baked into the AI being rendered. That’s something else that we look into at NeuroFlow and that people in the field people are very aware of. The last thing we want to do is take existing inequities in healthcare, bake them into AI, and perpetuate them. That would be a big problem.

It is important to have an awareness and make sure that AI expands access and is culturally sensitive. In many ways, AI allows access to greater diversity in terms of who is delivering care and how care gets delivered. This can address stigma.

In behavioral health, there has not been the same level of funding for technology as in a lot of other areas of healthcare. There are many reasons for that. The margins are often slimmer in behavioral health, and reimbursements are not great, so the technology often is limited. First of all, we have to educated behavioral health providers about the benefits of technology. We have to start with the basics.

Also, many who behavioral health providers appreciate measurement-based care, but there are many who don’t practice it. Why is it important to track longitudinally, not just when patients enter treatment? Not just to see how are they doing on assessments and scales, but to track how they’re doing longitudinally? We know that when you practice measurement-based care and combine it with psychotherapeutic practices, medication, or whatever treatment is being delivered, symptomatology reduces and response and remission rates improve dramatically. Patients do much better.

AI can facilitate the longitudinal tracking of how patients are doing, as I said, through asynchronous assessments. Often, patients will complete those assessments more readily and authentically in the comfort of their homes. It’s that white coat phenomenon: when a patient’s in an office or clinical setting, they may not feel as comfortable answering assessments and scales openly.

We have to train behavioral health professionals to use technology, not to look at it as something that is going to be a burden. In fact, it will save time, which will improve quality of care. That’s one.

Two, we have to recognize that there are workforce challenges, and we have to use technology as a workforce multiplier. Technology is not here to replace humans. We believe at NeuroFlow in the paradigm of “high tech, high touch,” that AI makes behavioral health providers smarter and better. It also makes primary care docs and oncologists who are treating depression and anxiety better, not just behavioral health providers. It makes everyone smarter.

Stay tuned for next week’s episode to hear more from Dr. Zaubler.

Thanks for listening. Stay tuned for next week’s episode. To hear more, follow PeerPOV: The Pulse on Medicine on Apple Podcasts, Spotify, or Amazon Music.

This transcript has been edited for readability.