A New Kind Of Healthcare Breakthrough

David 0:00

What if the biggest breakthrough in healthcare isn't a new drug or even a new device, but someone who's always there when no clinician can be? And what if artificial intelligence doesn't replace the human side of healthcare, but finally makes it available whenever people need it most? We'll find out on this episode of Shift Shapers.

SPEAKER_01 0:21

This is the Shift Shapers Podcast, connecting benefits advisors with thought leaders and entrepreneurs who are shaping the shifts in the industry. And now, here's your host, David Saltzman.

Meet Ken Resnikau And The Goal

David 0:34

And our guest today is Dr. Ken Reznikau. Ken is one of the world's foremost behavioral scientists and a pioneer in the field of motivational interviewing. After spending decades studying what actually causes people to change their behavior, he's now helping shape one of healthcare's most intriguing frontiers using artificial intelligence, not to replace clinicians, but to extend their reach and expand access to therapeutic care. Welcome, Ken. Thanks for being here.

SPEAKER_02 1:02

Thank you, David.

AI Should Expand Care Not Replace

David 1:03

Happy to be here. Let's start with the biggest question. Everybody seems to be talking about AI replacing people. You've taken almost the opposite view. Why do you believe AI's greatest opportunity is expanding health care rather than replacing healthcare?

SPEAKER_02 1:19

There will always be a need for human-generated healthcare. However, with the advances in AI, there are certain things that it does better. So, for example, it's really good at reading x-rays because it's been trained on millions and millions, whereas a human might be trained on thousands and thousands. So too with behavioral counseling, that as the agents get better and better, they'll start to get closer and closer to the human experience, although I don't think they'll ever completely replace it because there's always going to be some people who want and benefit more from the human touch.

David 1:59

When

The Real Care Gap And 3 A.M.

David 2:00

we talked a little bit about expanding stuff, how big is the gap between the care people need and the care we can actually provide today?

SPEAKER_02 2:08

Great question. So there's both a shortage of trained counselors in many fields. For example, genetic counselors have a very long wait in most healthcare facilities to get seen. But there's also, as you alluded to in the introduction, just the timeliness of it. At 3 a.m., if someone needs counseling, there's very few companies that provide round-the-clock support. And it always amazes me that Delta Airlines manages to do it, but there's very few psychologic services that work 24-7. And this is a great opportunity for AI when the counselors are asleep. Literally and figuratively.

David 2:48

Both literally and figuratively, I'll take

Why Information Rarely Changes Behavior

David 2:50

that. You've spent your career studying behavior change. Why is information alone almost never enough?

SPEAKER_02 2:57

Most of the reasons we change our health behaviors relate more to the meaning of the change than the information about that change. So although it does pay to tell people their cardiovascular risk if their LDL is above 130, what really will drive behavior change is going to be things such as what does it mean for me to lower my heart disease risk? It means my kids will be less anxious that dad is going to have a heart attack. It means my wife will be less anxious. And then it's also around feelings such as efficacy. How confident am I that I can make these changes? Do I believe I can change my diet activity and medication behaviors? Those are not knowledge-based drivers. Those are sort of affective and cognitive drivers. And that's a huge part of the behavior change process. Knowledge alone, we generally estimate has a two or three percentage effectiveness rate. It's just a very, as we say, necessary but insufficient pathway.

David 4:00

Well, so I guess we're talking kind of broadly about motivation. Is that something we discover rather than something we're just told?

SPEAKER_02 4:08

Correct. I'm going to add a third category that a good counselor and a good AI LLM agent develop collaboratively by exploring the meaning of change for the individual. They can help the person get what we call high quality motivation. And high quality motivation includes I've made the decision myself, which is the autonomy. It's personally meaningful, and it's within my skill set. And that's what we call efficacy. So a good counselor and a good AI agent help develop that high quality motivation.

From Chatbots To Therapeutic AI

David 4:39

Today, most people think AI is a chatbot. So let's talk a little bit about the kind of get into the work that you're doing today. What's the difference between a chat bot and therapeutic AI? And why does that distinction matter?

SPEAKER_02 4:54

Ultimately, it might be an arbitrary border between the two. Typically, chatbots have been rules-based, closed-ended questions and have not interpreted free text and have instead used things like choose from the four answers ABCD, and the chatbot has a canned answer for whether or not you chose ABCD, and the chatbot delivers that canned answer matching whatever you answered on the mini-survey, maybe in your phone or on a computer. Whereas one of the most interesting things about LLM AI-generated is it's really good at handling free text and understands very well what we saw, and that's just improving more and more. And that distinction, I think, ultimately is somewhat arbitrary. So we don't refer to our products at Chronologics as bots. I think that is sort of a little bit of a Betamax.

David 5:46

So we like to think we're at least CDs. I'll take point taken. So when does a conversation become therapy instead of just a simple conversation?

SPEAKER_02 5:58

That's a great question. There are some there are some very intricate legal issues. A couple things we try to avoid, and we have put a lot of effort training the LLMs not to do these things, and that's to diagnose and prescribe and to encourage someone to come off of a treatment, particularly medication. So those we still leave currently to the providers, the human-generated providers, to docs, nurse practitioners, dietitians, et cetera. It will be interesting down the line if they start to absorb that role as well. But right now, we ask whether or not you've been diagnosed with these conditions. We ask whether under the care of a physician, and then we go through role, which is what behavior changes can we talk about that will help you control the condition that a healthcare provider diagnosed. So that's a very important area where we still generally stay clear of.

David 6:54

Can you give us an example that listeners would immediately recognize?

SPEAKER_02 6:59

Sir, sure. So one of the agents we've developed works with people of type 2 diabetes. We ask if they have been given a diagnosis. We ask them if they're on medication, and if so, which ones? And then we give them a list of six or seven behaviors they can work on. And that's where we take over. And that list includes changing your diet, changing your exercise, changing your screen time, managing your stress, taking your medication more regularly, or getting a new prescription, not us prescribing it, but getting a new prescription from your healthcare provider. And so that list, which we call the agenda setting report card, helps us focus quickly within about a minute or two of the session. What is the app going to talk about today with the user?

Motivational Interviewing Over Advice Giving

David 7:48

Now, in the open, we talked about you being your entire career being dedicated to this thing called motivational interviewing. What's the other kind of interviewing and how does it differ from motivational interviewing? Great question.

SPEAKER_02 8:02

I'll give you the four-hour answer and I'll start with the two-minute answer. We appreciate that. The key differences in motivational interviewing, the provider, and in our case, the app, the AI-generated counselor, uses a lot of reflective statements and doesn't give a lot of directive advice. So in the real world, providers use things such as you must, you should, you auto, it's important, you better, it's dangerous if you don't. And we consider that a very low-grade form of motivation. We call that controlled motivation. And instead, motivational interviewing really tries to help the person self-motivate, helps the person find their own internal reasons to change, and even allows them to draw up the change plan, what they want to do, when they want to do it, how much, how long. So the key thing is the nexus of control shifts from the provider to the patient. A synonym for MI is patient-centered counseling, because it's really meeting them where they are and evoking from them their own strengths and evoking from them their own meaning. And in that case, I like to say that in MI, we're the rudder of the boat, not the steering wheel.

Training LLMs To Ask Better Questions

David 9:14

Is that difference what makes motivational interviewing such a natural fit for AI? It's a good question.

SPEAKER_02 9:22

I think we've had to retrain the agents a little bit in its natural feral state. The main LLMs out there, the ones people know about, chat, Claude, Llama, et cetera, tend to give too much advice by our standards. And there's a long explanation, depending on what it was trained on originally. It was trained on traditional counseling. Ultimately, that's what dominates the literature out there and the and the data sphere. So it defaults to a little bit more directive care than we provide. So we retrain it every time we log on. We give it a set of very long instructions that discourage the use of unsolicited advice and directive language. So it's inside, I like to think of it as a it's a mind that has these various rich veins. And our job is to get in there and get into those rich veins. Aaron Powell So is asking great questions more important than giving great answers? Yes. That's a very nice way to put it. That the most effective things are things that are evoked from the patient, not the provider.

David 10:28

I

Why People Open Up To AI

David 10:29

think as AI grows and becomes a bigger and bigger part of our lives, I think a lot of people are surprised that patients often disclose more to AI than to another human being. Why is that?

SPEAKER_02 10:43

It's a great question. And again, there's personality types. There are people who are just inherently suspicious of technology. But I think there's a sense that the AI agent doesn't gossip, and the AI agent has no motive to disclose this. So it it I think is assumed to be less of a risk socially to tell it to this agent. And I think there's also a general sense that people aren't listening in, which is generally the case, although in some cases these encounters are being recorded.

David 11:17

So does just simply removing judgment change everything for those those people, those archetypes?

SPEAKER_02 11:25

Yes. I think the fact that it's a free space where they should feel comfortable. I don't think what's what's nice about LLMs is people don't feel they could shock it. I think there's always a sense that maybe I'm going to shock my therapist because I'm so extreme. I think there's a general sense that an LLM can't be shocked and therefore has a broader range of normal.

Between-Session Support And Just-In-Time Help

David 11:51

So let let's build on that. One of the things I know you talk about is the time between appointments. Why are those moments so important?

SPEAKER_02 11:59

Very good question. I'll give you a concrete example. If I'm going to see a counselor through my healthcare at work, I may have a limited number of sessions and they may have a protocol to meet every month. And so on July 21st, I set a goal to walk three times in the next week. I may not see this live human for a month or even two weeks. What apps are really good at is I can dial you up tomorrow and say, David, how'd that walk go? And I could problem solve. If it didn't go well, I could say, what can we do to overcome those obstacles? If it went well, I could say, how are you feeling about it and give you an opportunity to get all excited and brag on yourself. And that's that between session of supplemental care that you've been alluding to. It's really good at that. And particularly, you know, if it's an advanced digital system and you set a goal to walk tomorrow, Wednesday, the 22nd, our app will know that and we'll contact you late in the day on the 22nd to see how it goes. Very few therapists are able to do that. There's some concierge services out there, but the average counselor just doesn't have the bandwidth to make that extra contact. Whereas the digital environment, it's really a very low-cost additional service.

David 13:16

So broadly speaking, it would it be safe to say that outcomes really determine more in everyday life than in the exam room? Yes.

SPEAKER_02 13:25

And we have a term for it now, and we call it just-in-time intervention. And that's meeting the person in their real life. So a very good example is if someone is trying to reduce the sodium in their diet, we can determine through our digital technology that they're in a supermarket. And we can even know the products that are available in that supermarket, and we could even know their sodium content. And so we could say, David, you're in Kroger right now and you're trying to reduce your sodium, as we talked about two days ago. Here's some really low sodium soups. We could really get in there just in time as the behavior is happening. And imagine having a therapist shop with you. That that doesn't happen very often, does it?

David 14:07

No, it's kind of like shopping with your spouse. And, you know, they remind you, hey, you know, that that's got a lot of sodium. Buy this one instead. So I, you know, I I get, I think anybody, anybody who's ever shopped with a friend or a spouse understands

What Employers Gain From Digital Counseling

David 14:20

that. As you know, most of our audience are benefit advisors. So let's bring this back to the employer realm. Why should benefits advisors care?

SPEAKER_02 14:30

Well, there's sort of two very important benefits of these agents. One is cost, that accounts are ranges, uh, depending on if you're using bachelor's level all the way up to PhD, anywhere from a 30 an hour to 150 an hour, and in even more in some situations. The cost of the digital support is a fraction of that and generally will be delivered at 110th, 120th the price, the typical things, you know, 50 cents a month per person, things like that. The other thing is efficacy. It's it's going to enhance the impact on people's lives. So if we're lowering people's cholesterol blood pressure, weight, blood sugar, that saves money because of healthcare costs. It reduces hospitalizations, acute care. In some cases, it reduces use of expensive medications. So there's sort of that two-pronged benefit of saving money and improving health.

David 15:26

Does it also affect absenteeism, presenteism, and maybe even in some cases, workers' comp?

SPEAKER_02 15:32

That's still emerging, that literature. The worksite health promotion literature from the 90s and early 2000s established that of on-site Weight Watchers, on-site cooking classes, on-site exercise in general, improved retention of employees. It was seen as a generous benefit, which is a wonderful uh cultural advantage you could achieve in your work site if people feel you care about them. Now, is it true when people add apps? That literature is just emerging. I think it's a little young to know, but it's going to happen soon, that type of research.

David 16:07

That's

Redesigning Care And Fixing EHR Integration

David 16:08

fascinating. Let's let's build on that a little bit and go broader. If you could redesign behavioral healthcare from scratch today, if you had that magic wand or whatever universe you're using, what would it look like?

SPEAKER_02 16:24

Well, I would want more trained counselors. I think the training we give folks is inadequate. So, for example, I think everyone should learn motivational. It doesn't mean they should do it with every patient and every encounter, but it's such an essential skill. I can't see certifying people to be coaches without it. So I think I would increase the rigor of what skills they're required to exhibit and uh make that part of their credentialing. But what I really would like to see, and it's happening, but it is a lot of retrofitting, is the integration of systems. So our apps should talk to electronic health record with uh great ease. So it's wonderful if our app can look inside your health record. The biggest one we work with is called Epic, that is used by thousands and thousands of providers. But not all apps integrate very well with Epic. So we'd love our app to be something that a doc could prescribe in Epic. And then when the person sets a goal in our app andor walks, as the app verified, you know, I talked about I'll check in with you tomorrow night because you set a goal to walk tomorrow afternoon. I want my app to push that back into the EHR, into the electronic health record, so the doc and the other team members can see it. That has not generally been achieved. We're all working towards that. There's a mad dash to integrate the digital with the human. And I think that will be a huge benefit to both the provider and the patient to have an integrated digital health system that links to our current electronic health record, which, as you know, is largely a billing mechanism, less so an intervention mechanism, although it does both.

David 18:11

But the technology is there today. Actually, one might argue the technology's been there for a while. And yet the systems are fractioned. Why hasn't that come together yet? It's a good question.

SPEAKER_02 18:24

I would have thought that the electronic health record companies would be more aggressive. That's why I think it's largely there for billing purposes and record keeping and protecting your, you know what, legally. I think that's they're not primarily meant to enhance care, even though, of course, that's part of it. So I think if we had that orientation of how do we enhance care, I think there'll be much more creative ways to integrate the two worlds.

Digital-First Care And Clinician Partnership

David 18:52

So let's go back to that crystal ball I asked you about. In in in the perfect world, would AI become the first touch point or the last? Right.

SPEAKER_02 19:00

We generally operate from the digital first perspective. If we can do it by moving molecules and the person responds, even if it's not quite as effective as human, we're going to save society a lot of money. And that's generally a good thing because healthcare is so darn expensive right now. So digital first, so we would try the app. Generally, it'd be it's a good idea if the trusted provider recommends the app. That's you know, that warm handoff that we talk about. So a warm digital handoff, I think, is a good start. Try a couple of sessions with the uh agent. And then if that doesn't work, then we up them to the next level of care, which would probably involve a human.

David 19:44

Do clinicians today see AI just as another digital health tool, or is it becoming a new member of the healthcare team?

SPEAKER_02 19:52

That's a good question. I think we're in the middle of that. I I don't I think providers have a skepticism about whether or not it's going to do any harm and whether or not it's gonna replace them. I don't think you know neurosurgeons are worried about that, but there's a lot of you know people below that level who are worried that the tasks of these agents are getting more and more sophisticated and therefore their jobs are in jeopardy. So I think there's an ambivalence about this whole technology. It's a threat as well as a exciting new opportunity, but that's the ambivalence.

David 20:28

Is there a role that AI can fill that humans just can't?

SPEAKER_02 20:34

Yeah, there are a couple things. So one is the timing, as we talked about. They can be available 24-7. Another one is this, I think, complete sense of you can't freak me out, that there's it this no judgment. Apps have there is actually a little judgmentalism in in the LLMs, but we have teased that out with our programming, with our sort of retraining that we give the agents. But uh that aside, that the sort of moral free zone that it can establish. There's a couple other things that it does really well, retrieving information so that if you ask a good question, it's gonna serve as a very sophisticated uh web searcher. Um so it does that very well. So it will get you a summary of the web-based knowledge very quickly, faster than you could do it, because you'll have to search, you'll have to read WebMD, you'll have to read the May Clinic, it's gonna consolidate that for you.

David 21:29

It almost sounds like as AI gets better, unlike the ambivalence they may have today, clinicians become even more valuable. Do you see that?

SPEAKER_02 21:39

Yes, but it there's gonna have to be a shift that it's a partnership that healthcare providers encourage the effective use of these agents and give them tips how to get more out of it. But I don't think that's happening as much as it needs to yet, because I still think there's this tension between them. But with your Magic wand going forward five or ten years, I think we'll make peace and that providers will be considering these an asset and just part of their toolkit, not a threat.

David 22:12

Ken,

Hard Limits Safety And Prescribing

David 22:13

are there some things that just should never be delegated to AI?

SPEAKER_02 22:17

Yeah. There's some high-risk situations. So, of course, the one everyone talks about is suicidality, that it probably shouldn't be doing that counseling. It should be getting them to suicide hotline or 911. There's sorry. And not take over that role of the emergency call center. At some point, you know, right now, again, we don't have it ever prescribe a medication. It will describe a medication, but it won't prescribe one. I'm very fascinated who's going to break through that wall and how that's going to play out. Because at some point, it's going to do it. It's not going to be forever this way. So I'm very curious how long it will take for someone to break through that wall.

SPEAKER_04 23:02

Would that have to be combined with some kind of pharmacogenomic testing?

SPEAKER_02 23:10

It might have to be combined with a pharmacist who has to approve it rather than the MD, which is the where the cost is, is their time is more expensive. So sure, it would be nice. I'm just thinking if it was linked to testing. So if I if I prescribe a GLP1 for your diabetes, it would be linked to your every three-month A1C values. That would be wonderful. That's the this idea of, again, linking it to the EHR, because those values are in the EHR.

David 23:41

So broadly speaking, do you think that AI could become the single biggest expansion of access to behavioral health that we've ever seen? Perfectly said, yes.

SPEAKER_02 23:51

And both improving reach and efficacy. There's a wonderful equation in public health that says impact is the product of reach times efficacy. And AI agents could be improving impact on both of those dimensions, both on the reach, but also because some people respond better to it, both increasing the reach as well as the impact and therefore improving its overall effectiveness.

Hallucinations Guardrails And Second Opinions

David 24:16

So again, broadly speaking, what's the biggest misconception or concern that you think we need to talk about more?

SPEAKER_02 24:23

Yeah, very there's a lot of fear about hallucinations and the agents going out of bounds. And it does happen, but we're getting better and better. It's going in the right direction. I don't know if it'll ever get to zero, but in our agents so far, and in other ones I've been involved in with colleagues, the harm and safety have not been a problem yet. And, you know, there's some very public examples that are two or three years old already where it hallucinated. And this goes back to uh Apple Maps. I think there's a, there was this case. I don't know if it's even true, but you know, it led someone off a cliff, uh, literally. And people are worried about our agents doing it figuratively of leading them off a behavioral cliff. But we just don't see a lot of it. We listen in and watch, you know, uh people use our products and try to monitor them. And we haven't seen uh much harm. We do have a very cool feature in our product. I'm not here to, you know, shill, but we did add a second opinion. So anything that involves intense emotion or the potential for prescription or diagnosis, we have a second LLM look at that and ask the first LLM to redo it. So we have a second set of digital eyes on all high-risk output. And that should greatly diminish this already rare event.

What Changes Most In Five Years

David 25:48

So again, looking into the future, five years from now, what do you think people are going to say that AI changed the most about healthcare?

SPEAKER_04 25:55

Great, great question.

SPEAKER_02 26:00

I think it will become acceptable and normative. And that, you know, people say, my agent, you know, I use so-and-so, and I really like it. And people will just like they talk about their therapists, oh, I, you know, I like Dr. Sauzman, he's really good. They'll say, I like, you know, uh, whatever, X, Y, Z app, and they'll say why. And I think it'll become part of the normative discussion. Right now it's still a little bit of a boutique y item or, you know, an extravagance or something, but I don't think that will be the case in five years. I think it'll be part of your normal discharge process from the hospital. It'll be something that docs prescribe and are covered by benefits. We haven't talked about that yet. Not a lot of these apps are currently covered by benefits. You're paying, as you know, because you do this for a living. People are paying cash for it, in effect.

David 26:53

Or self-funded plans are paying for it. Yeah. You know, if that happens to be the funding mechanism that a particular client is using. But let's follow up on that. Yeah. What's coming that most people just don't see yet? What do you think is out there that's heading our way? And most people haven't even dealt with it yet.

SPEAKER_02 27:11

Yeah. What people find most interesting about what we're developing is how well these new agents handle emotion. And that you could get very emotional about your kids and your diabetes. And it can handle it really well. The level of emotional intelligence. And I have to put it in parentheses because it's a different type of intelligence, because it's really just anticipating what's the next logical set of words. It's an amazing thing. That's really what it was designed to do. Um, but I think people will be very surprised that you it feels empathetic and it feels like it's understanding your emotions and your worldview and your foibles and your aspirations. And I think that's what when people hear some of the better apps do it, they're like, whoa. Because I think people think about falsely but understandably, that it's just a very sophisticated search engine, that it's just gathering information for them. It's really much more. And I think people will start to understand that.

The One Takeaway For Benefits Leaders

David 28:17

So as we wrap up here, if if every employer, broker, healthcare leader remembered just one thing from today's conversation, what would you want it to be?

SPEAKER_02 28:28

It's here, it's not going away, and it's probably better and safer than you think.

David 28:35

That's certainly a great place to wrap up our conversation, Dr. Ken Resnikau from Chronologics. Thanks so much for helping us understand this new and emerging technology and how it can help. Thanks for your insightful questions, David.

SPEAKER_00 28:49

The Shift Shapers Podcast is a production of Shift Shapers LLC. The content and images of this podcast may not be used without our express written permission. Copyright 2023, all rights reserved.

Recording Gremlins And Rapid-Fire Questions

David 29:12

So here's a fun little piece of information that in, I don't know, 600 interviews, I've never had this happen. Your headset came on in the middle of the interview. The microphone on your headset?

SPEAKER_02 29:23

That's so weird.

David 29:24

Yeah. Does that mean you we lost audio or no? No, we never lose audio. What what happens? A couple of Israeli kids had a really great idea. Years ago when I was doing radio in the 70s, if we did a remote interview like this, I would only hear you over a telephone line. You would record where you're sitting, I'd record where I'm sitting, we'd give the two tapes to an engineer, and they'd slice and cut and marry them together. What happens with with Riverside is that what we record is not what I'm hearing or seeing from you. It's what is on your computer and similarly what's on my computer. So that even if we had an internet glitch between the two of us today, that wouldn't be recorded. So as long as we don't have as long as we don't gain enough delay so that I can't ask and answer questions, there's no problem. And even if for some reason yours didn't upload, I can send you a link. You can click the link sometime in the next week. That's it only lasts for about a week, and that file's still floating around your computer someplace, and you can upload it manually.

SPEAKER_02 30:26

Okay, cool.

David 30:27

So they do a great job.

SPEAKER_02 30:29

I didn't click on anything. You know, I did it before we started, but I didn't in the middle.

David 30:34

No, I saw you. It was weird because all of a sudden your sound changed in my ear. And you know, I listen like through one of these guys, and I went, okay, what's happened? And I looked over at your list of equipment and it says microphone, Logitech, USB headset, which it did not say before. So gremlins. I chalk it off to Gremlins. That's I chalk everything off to Gremlins. So if I may, let me ask you those three survey questions real quick.

SPEAKER_02 30:57

Absolutely.

David 30:58

What's the one thing that people consistently misunderstand about what you do?

SPEAKER_02 31:03

I think people think we just give advice and information. And we're we're really doing a lot more sophisticated intervention than that.

David 31:13

If someone had only seven seconds to understand what you do, what would you want them to remember?

SPEAKER_02 31:19

We try to help people motivate themselves by exploring the meaning of the change. And motivation is how much energy are you willing to devote to this? So when we ask someone, are you motivated? what we're really asking them is, are you willing to devote the cherished research of your personal energy to attack this problem?

David 31:40

Great answer. And finally, what's the one thing you wish your customers understood much sooner?

SPEAKER_02 31:46

It's safe, it's effective, it's here to stay. And from our perspective, it's highly cost effective as well. That's awesome. Hang on, I'm going to stop the recording and we'll make sure we're both uploaded.