← Practical AI in Healthcare
Loading summary
Transcript72 lines
- [00:02]
A
Welcome to Practical AI in Healthcare, the podcast that cuts through the noise to spotlight real world solutions delivering real world value. From patient care to clinical research, from life sciences to patient engagement, we focus on what truly matters in healthcare today. No hype, no theory, just practical insights where AI is making a true impact. Dr. Steven Lapkoff and Dr. Leanne Rosenblitt are your hosts as we explore what's real and moving the needle in this exciting new domain. Welcome aboard and let's get to it.
- [00:35]
B
Hello, welcome to Practical AI and Healthcare. My name is Dr. Steven Lapkoff. This is a special edition this week of our podcast, which was recorded live in front of a studio audience at Baker Tilly during New York Tech Week. We interviewed Fred Bennett, the CEO of Patient Talker, a young AI company that's looking at ambient listening from a patient's perspective. And you'll hear about that in just a few minutes. We do want to do a shout out to my co host, Leon Rosenblut, who's still healing up from a bad nerve in his back. Leon, we hope you get better soon and get back in the recording chair. And with that, we hope you enjoy this live edition. And if you have any ideas for additional podcasts or you have an idea for a live show, please drop us a note. You can reach me at steveractical AI@healthcare and leon@leonpracticalai in healthcare.com and with that, let's get started.
- [01:33]
A
Thank you for joining us this week on Practical AI in Healthcare. If you're ready to go beyond buzzwords and hype and explore how AI is truly transforming healthcare, stay tuned for more conversations that get us and what works. Until next time, stay practical.
- [01:53]
B
Hello, My name is Dr. Stephen Lapkoff and welcome to Practical AI in Healthcare, and we are here this week with Fred Bennett, who is the CEO of a startup called Patienttalker. And Patient Talker is a company that works in the ambient listening space, but from a perspective that most of you probably won't have seen before. So we're going to dig into all this and unpack it in just a second. I did want to do a shout out to my partner, Leon Rosenblut, who can't be here today due to a medical issue. He's still dealing with his pinched nerve in his neck, but he'll hopefully be back on air in a week or two when he's feeling better. So, Leon, I hope you're improving. So, as we open up today, Fred, you know the first question we'd like to ask our guests is, how did you get Your superhero cape. How did you get to where you are today and how did patient talker come into being?
- [02:45]
C
That's a high hurdle to be called a superhero. I earn it with a cape. And I won't leap over any tall buildings here. But the genesis of this, and I mentioned it a minute ago, is improving. Doctor patient conversations is not new, but it became real for me. I guess they would call them the aughts now. So the like 2007, 2008 when my dad had had a series of cardiac events, he had had a couple heart attacks and had had a stroke and he spent a few months in rehab getting over the stroke and he was going to a cardiologist visit for an update and check in. And this was big, it was important, he was older. So my mom went with him and I went with him as well. So we go to the doctor and everything at that point was fine. Did the routine checkup. He talked about the medicine and probably up the statins, talked about lifestyle, said you're in your late 80s, you don't need to live like a monk. You can have a piece of cake every once in a while but really the Mediterranean diet is what you should be focused on. And here's how to live healthy, exercise, et cetera. That was great. We leave, we walk to the car, my mom, my dad and I and said okay, where do we go next? And I, as we alluded to it being in pharma at the time, said oh, he prescribed new medicines. I know how to deal with that. Let's, let's go pick up those new medicines at the pharmacy. My mom said no, no, no, diet, diet and exercise. I'm going to clean out the fridge. You know, we're already pretty healthy. But he gave some really good suggestions. Let's, let's update that. And my dad said no. He said I could have cake, I want to go get my cake right now. You know, and in hindsight it's funny and maybe it's a little glib but, but what I took away from that is we all sat in the same 30 minute meeting, we all were focused on this intently on what the doctor was saying and we all had three very different things that we were, we were coming out with. And you know, I tell a story because it's sort of a light hearted story during a talk like this, but you can imagine with more serious oncology or brain injury or other sorts of of events where it's not cute that you don't remember what happened after that. It's, it's becomes a matter of outcomes and, and the care that you receive. So long story short, that was, that was sort of the kernel way back then that put in my mind,
- [05:27]
D
this
- [05:27]
C
is something we need, this is something that I need. And as I talk to more people I see some heads nodding in the audience right now. This is a common occurrence that people need help remembering. You know, you're in the doctor because you're sick. You're not at 100%, you're not able to pay attention anyway. I could go on, but it's a
- [05:43]
B
real need and it is a real need and it's not just for the elderly. Frankly, if you go to the doctor yourself and you're 20 something or 30 something, your mind isn't always completely there when you're ready to listen to what the doctor's saying and you might forget, you know, up to half of the information or more. So this isn't, I mean Fred framed it from a perspective of his dad's issue when his dad was late in his life. But this is an issue that actually can happen. Let's say you're a young mother and you go to the pediatricians and your kid is wailing because they have an ear infection and you're not paying as much attention to the, to what the doctor's telling you in that setting either. So the point of this I think is that there's a lot of perspectives here that are really important to really unpack. And I think your perspective on it and what you're doing with patient talker is, is addressing that need.
- [06:35]
C
I'll layer on one more, which is this, the emotional aspect, right? There's just distraction, there's understanding the language. I spoke with this woman who got a cancer diagnosis and the diagnosis was given to her in the first few minutes of the meeting. And she described the rest of the meeting. The, I call it meeting, sorry, the rest of the doctor visit, doctor consult. It was like hearing Charlie Brown's teacher. I don't know if people remember the Charlie Brown shows.
- [07:00]
B
Wa wa wa wa wa wa.
- [07:02]
C
She had. Her mind was racing in 50 million directions and she just could not pay attention or, or, or absorb what the doctor was telling her. And so she had to call the next day and say, I'm sorry, could you repeat everything you told me yesterday?
- [07:15]
B
So Fred, the ability to record a conversation has been around a long time. You could walk into a doctor's office and just use your phone, turn on the voice recorder and walk away with that recording. That might break free laws in the state of New York or other places, who knows? And AI, the podcast is called Practical AI in Healthcare. What's the connection with AI to this problem?
- [07:45]
C
Sure. And just so if you feel okay in New York, it's actually you're not breaking any laws because New York is a one consent state, but there are 11 states out there that are two party consent states. So you would be breaking laws if you're not aware of what the laws are in your jurisdiction. But that's a technical aspect. I'm sorry, what was the question again? AI. I remember.
- [08:05]
B
What's the AI connection? Yes.
- [08:07]
C
So the AI connection is, you see this was in rehearse.
- [08:09]
B
This is all that's right. We're live and in person right now. No scripts.
- [08:13]
C
So the AI, so one the thing that you mentioned. Sure. Someone could just bring in their phone, download the audio and then upload the audio to ChatGPT or Gemini or Claude or whatever your favorite is. And people have people of course do that, do that currently. And that's not bad. I think the percentage of the population who's going to do that is pretty small, especially drink things from all the things we just talked about where people are under pressure, they're concerned, they, you know, maybe they're elderly, maybe they have other conditions that are just keeping their mind occupied. So, you know, if I told my 90 year old dad, oh, just record it and then feed in chat dpt, he'd like what? So while you could do it, and while AI is great at taking that transcription, well, doing the transcription, text to speech, speech to text, doing the summarization and then picking things out, it's not as easy as just doing that because there's prompting you need to do to be able to make sure it's not going to try and second guess the diagnosis that the doctor just gave you to make sure it's not going to hallucinate, say, oh, I think you should start taking this other medicine instead. Having an AI play doctor is problematic for a whole host of reasons that many probably people understand. And so one of the things I've done is try and really carve out to let the AI be a helper, not a replacement for the Docker conversation and package everything together so it's one tap for someone to be able to use.
- [09:50]
B
So let's dig into that a second because, you know, one of the things we've done at Harvard is we've looked at how do patients interact with these things, how do they find value in it. And this issue of digesting the information and not diagnosing. I mean, your tool explicitly doesn't do diagnostics or doesn't do clinical decision support per se. Tell us precisely what your tool does and why that separates it from other pieces. Because it does a bunch of different things. But at the core, what's its main purpose?
- [10:26]
C
Sure. So at its core, core it's what we've just been talking about. Recording conversations and summarizing them into health literate, patient friendly language. But there's pieces both before and after that, right? The healthcare is not the only time that you're sitting in front of your healthcare professional. There's as you're getting ready for the visit, you're thinking about what are all the questions that I need to think of? Not to be trite again, but there are hundreds and thousands of videos on how to organize your kitchen, how to organize your closet. Very few people think about how to prepare themselves for their doctor visit. What are the five questions that I want to get answered? I hear this from patients. I hear this from doctors that patients just come in and can't articulate it. So you say in one sentence what you're coming in for. The AI will say, here are a list of questions that you might want to have the doctor talk to you about it, then does the recording and then afterwards the summary. It records it for you, helps you organize it, and very importantly connects you with your caregivers. So if my sister in California wants to know how my dad's visit was to the doctor, you can easily send it to her and sure. Could you look at on Epic Mychart to do that? Yes. But Epic Mychart tends to be confusing. People see the red lines, you know, when your lab test is at a result and it often scares people more than it helps them.
- [11:53]
B
So you just hit on a nerve a bit. Because I'm a clinician, as I've mentioned at the opening, I've used EHRs. I've actually helped write EHRs. I've actually been involved in buying, selling and implementing EHRs. And there's a different focus here. EHRs are focused in on the doctor patient relationship from one direction. It's from the direction of the the doctor wanting to collect information and document information, largely for billing, but also for care. And today that unidirectionality leaves the patients kind of wanting. Patients don't have a similar type of backup or backing for their own personal use. And that's kind of from my mind. When I've learned about what you were doing and why I decided to get involved and to be an advisor. This issue of helping the patient explicitly is a serious differentiator. Why did you choose that angle as opposed to where everybody else is going? Or maybe that's why you chose it.
- [12:57]
C
Well, I think I want to linger on that for a second because I think that is the big takeaway that I have and hopefully can impart on everyone here today in person and on the podcast is the patient is often left out in the design of a lot of the IT and other infrastructure that we have. The patient is so, you know, the patient is the whole reason that the healthcare system exists. But yet when you're thinking about the creation of EHRs or, you know, billing systems or all this other infrastructure that's around it, the patient is often an afterthought. And I think that is wrong. Just to put it bluntly, that's wrong. So you're right. There are just talking about ambient listening solutions. There are a lot of other ambient listening solutions out there. Abridged nuance. There are a lot. They all are addressing real need and doctor's time. And time it takes for them to get the notes and time takes to get accurate billing is important. I'm not going to not agree to that. But what I think equally important is the patient's need. And letting these tools be there to help the patient feel more comfortable, feel more in control, advocate for themselves, and at the end of the day, know what they need to do to get better outcomes is critically important.
- [14:16]
B
Yeah, and I think therein lies a real interesting aspect of healthcare AI right now, and it's one of the things we unpacked at Harvard, is that you'd be really surprised. There are, you know, if you looked at like out of 100 AI apps in the healthcare space these days, those that are specifically focused on patients are well outnumbered by those that are focused in, on. On clinical care. And from the clinician's perspective, I don't want to put a number on it, but I'll take a guess, it's probably far less than 20%, maybe less than 10% are patient focused. And you know, this is the time in our lifetimes where data is becoming democratized, where everything is becoming democratized, where patients are being empowered by technology. And AI is no exception. We've interviewed on our podcast a guy named E Patient, Dave Debrankert. Dave Debrankert runs a group that's the Society for Participatory Medicine, and his focus is on how do you focus not just AI, but tools in general from a patient's Perspective. How do you empower patients to, to take control of their lives and take control of their care in a much more deliberate way? I mean, that led to something called the Open Notes movement, which was something that spun out of Harvard, I don't know, now maybe 15 or 20 years ago at this point, where it used to be that the medical notes in an electronic medical record were literally locked and blocked from a patient to be able to see them. Today you can actually get access to your patient's notes through this movement, Open Notes. And it's become something that is a very serious patient empowerment tool and movement. And it still surprises me we don't have as many folks focusing on the patient side of this. Fred, the aspect of this patient empowerment piece, you know, there's a difference, however, and you know, when you're empowering doctors or you're empowering or if the insurance companies want information, there's money at the play there. There's a lot of money at play there. So helping a doctor to see three more patients a session, that's some serious money. Your app and your, your problem space doesn't touch the world in quite the same way. How are you planning on monetizing this? Because it's going to be a company. You're not going to do this for free. How are you solving for that problem? There's lots of money in. That's why Abridge has just made, you know, another series from a VC perspective. I think they have a round B or round C at this point. They're doing very, very well. So is Suki and Nabla and so forth. You're at the beginning of your journey. How do you see this unfolding from a, from a financial perspective?
- [16:58]
C
Yep, terrific question. I don't think the end user, the patient, should need to pay for this. I mean, I. Again, back to the idea of reducing friction for patients and being able to have them get the health care they need. My model is this is going to be free to patients. Having said that, there are a number of groups, enterprises whose incentives are aligned with the patients in terms of intensive incentives of having better health outcomes, reducing recurrence, readmissions, all those things. So hospital systems, not necessarily the individual doctor, doesn't have the same financial incentives that a hospital or a system would. But those systems, insurance companies, other payers, there's a decent sized list of people who are aligned with those incentives to have patients get better health outcomes. Because health outcomes means less cost to the health care system and less stress on the health system. And maybe it's not the quantitative three visits per day, but it could be X number of fewer hospital visits per year. And that's significant.
- [18:14]
B
That is very significant. So hospital readmission is an area which has gained tremendous amount of focus with the ballooning cost of healthcare. Healthcare had been, at least last time I checked, ballooning at a rate of something north of 3x that of inflation. And anytime anything happens that generates more demand on the system that requires payment, that's a problem. One of the things I like about your, your, your business model is you are aligning incentives. So many times in my career when we've looked at what's wrong with healthcare and what's wrong with healthcare and healthcare, it, it's like who's going to pay for it? Why are they going to pay for it? How is the money going to work? 20 years ago when I was at Pfizer, I was on a project that was looking at health information exchange. Now for those of you who don't know what that is, Health information exchange, or hie is a means in a region and eventually the country where your medical records can be shared with your permission to those around you in a big city or in a region. And the, the problem with that health information exchange model was there was no incentive for anybody to pay for it. And I did a tour. I was really fortunate at Pfizer. I was in a group that really allowed me to do some interesting things. I went around the country and toured eight or nine health information exchanges with an eye towards how's it being paid for. And all but two out of those nine had business models that were actually financially viable. And by the way, here we are 20 years later, both of them have gone out of business. Now health information exchange does exist to this day. There are other ways they figured out finally how to pay for. But the point that you're made here about the incentives being aligned is a really crucial issue because this is great to do for patients, but it's only going to be great if it can actually be a financially viable concern. And I think that tapping into those areas where those concerns are, are aligned is going to probably be the way towards success for, for you and for patient. For patient talker. Fred, let's dig into abridged Suki and Nabla a little bit more. Their, their focus on doctors and your focus on patients. What do you think? You know, is that an exit for you? Do you think they'll be, you'll be bought by one of those, those organizations is what, what's your, how do you look at them in terms of your. Your current and future state.
- [20:47]
C
You know, I don't want to predict M and A or anything at this point, but the economic moat of what I'm doing is not the technology. Right. The. The ambient listening. You talked about democratization of data earlier. AI is really a great democratizing force. You know, anyone could take the technology of what Patient Talker or any of these other bigger ones in some respects, and some of these big companies in probably a couple weeks could duplicate the technology. And I think that's been actually one of my big learnings is that technology is sort of the cost of entry, but it's not a barrier to entry.
- [21:30]
B
Not these days.
- [21:31]
C
Not these days at all. And the knowledge of. Really deep knowledge of how patients think, what patients need, and that obviously is not a one size fits all question either. That's a very diverse set of things. How the industry is aligned and incentives align. I think all of those are what you mentioned. A few names I'll just throw out epic. EPIC is doing a lot to disintermediate lots of people. You know, could EPIC build Patient talker in a week? Sure. Could they pivot from their focus on billing and administration? Administration. Thank you. To be more patient focused, it's hard. There's a lot of inertia there. I talk to people. There we go. I talk to people who work at epicins and say everything that goes out our door gets reviewed by a handful of lawyers, a handful of doctors, a handful of regulatory people, and takes at least, you know, a month or six weeks to get to get anything out the door in terms of new information, updated information, never mind building something. So I just think there are institutional hurdles there to thinking patient friendly and, you know, patient talker being built from the ground up with a patient in mind I think is important.
- [22:54]
B
So I want to go to patient safety for a second. Patient safety in the AI world is something that people have been really focused on. We focused on it a lot at Harvard because AI is anything but perfect. And anyone in the room who uses ChatGPT or Claude or Gemini knows all the trouble it can cause by hallucination. They keep basically making things up when it thinks it's trying to please you. What, if any, concerns do you have about these translation matrices that are being put into practice here and whether or not it's going to tap into those risks and what are you doing about it?
- [23:37]
C
Yeah, so there, there are a couple types of risks that you're going to, that I want to talk about. You Talked about accuracy risks. The other is privacy. Right? Privacy is a big thing. Both of those are foundational to how I'm thinking about it, how we're building things in terms of the accuracy. Not to geek out for a second, but there are controls. There's this thing called temperature. You can set a setting for your LLM to say, don't make stuff up, just be really dry and just tell me exactly what the answer is. So it's things like that, it's prompting, being really clear with prompting. And there's, there's, you know, we all, we all do our own prompts every day when we're tapping, tapping in, typing into the chat window on your LLM. But there can be real science to it as well. When you're, when you're creating those persistent prompts on the back end and doing your own training of the models. So we spend a lot of time on that. Suffice it to say, in terms of security and privacy, again, out of the box, if you will, there are switches that say, don't train the models with the data that you're collecting. So that's, I think that's again, ground entry, but making sure everything is, you know, hipaa, HIPAA standards, HIPAA or hipaa, quality of transcript, of encryption, of rest encryption and transit, sort of. I have a long spreadsheet of all the things that, that need to be done to ensure HIPAA and SaaS requirements. So it's, it's embedding all of those.
- [25:11]
B
And, you know, I guess I'm going to press a little harder on this one because I was warned not to give you all softballs.
- [25:18]
A
Okay.
- [25:21]
B
What happens if it gets it? So basically, the workflow of your app is it records the trend, it records the interaction, and then it translates that into prose, and then it digests it from prose into a shorter digest. What if it hears? How do you check for hypo versus hyperkalemia? How do you check for a missed negation? Because those things, you know, in the past, when I was a fellow, I was working in and around speech to text and nlp, Natural language processing for a good chunk of my career and for my entire career until 2022, the answer was it's five years away. It's five years away. In 1997, when I started at Pfizer, it was five years away. And in 2022, it still looked like it was five years away. It's not five years away. It's here. And it's working relatively well. But help me understand or help the audience understand that problem because I think that's an intrinsic challenge that needs to be bridged in your world. So what's that about?
- [26:27]
C
Well, one, you stole my punchline. As you say, it's not five away years away anywhere. See, if it's here, like a lot of things, it's multimodal. There's improving models, improving speech, text models. There are models out there that are specially trained on medical lexicons and medical just symptomology and other things. But even if you get a model that's 99 or 99.99% accurate, that means there's still in the course of a conversation going to be errors. So the other layers that we put on top of it are actually the AI summarization helps in that. If it's a hyper versus hypo or remember, be sure not to take this medicine. Those are all things that if you miss one word, it can dramatically change the meaning, right?
- [27:25]
B
That's exactly right.
- [27:26]
C
But when you take the conversation holistically, the AI and the tools will be have it in context of all the other things that are being discussed. So it's unlikely that the doctor will say, remember, be sure not to take this, and then not mention anything else about it during the visit. If it's that important, it's be sure not to do this. There'll generally be a question, a clarification. And by not worrying about the accuracy, I mean, we do worry about it, of course, but by not relying only on the accuracy of the transcription itself, but running that through the models to say, okay, what is the takeaway? The surrounding data and the constellation of data that's around. Any individual recommendation doctor gives you helps make sure that the same way a person might miss a word, the overall gestalt of what you're getting is going to be accurate.
- [28:25]
B
So thanks for that. I want to now turn a little bit towards, you know, you've been at this now for a while, like 18 months, 24 months, something like that.
- [28:34]
C
Yep.
- [28:35]
B
And I'm sure you've learned some lessons along the way. Maybe you could tell the audience, you know, in the audience today, I'm sure we have some young entrepreneurs who are probably trying to do similar types of things. What are some of the lessons that you've learned along the way? And let's box it away from the AI. Let's talk about what it means to be a young entrepreneur. Or maybe you're not so young anymore, but an entrepreneur in this space. Don't judge, don't, hey, listen, I'M older than you, so it's no judgment, but what kind of lessons have you learned along the way that have maybe helped change the course of where you're going and where you want to. Want to end up?
- [29:15]
C
Yeah, I mean, I. When I started, I was thinking, oh, AI, there's all this stuff, there's all this exciting abilities that. That 10 years ago, just the technology wasn't allowing us to do these things, and now we're at a place where we can get it done. And to your point, we're going to box it off from this, but I thought that that was going to occupy much more of my time. I think what I've learned is, is that 20% of my headspace is around the technology. And what I feel I've become is a patient empowerment advocate. And I feel like I spend so much time telling people, the patient is not the tail wagging the dog. The patient is the reason why we're doing this. I'll get on another soapbox for just a second. It's tech week and we're all going to lots of events. I was in an event yesterday trying to people about this. They said, oh, are doctors not going to want to be recorded? And like, more than one person said that to me. A number of people said it to me. And it's a question I get a lot. And I said, one, sure, there are always going to be some doctors. Two, I think that number is coming down. But three, when I went to my last checkup, you know, I filled out the forms. I had the HIPAA form that said the doctor can talk to my insurance company. I had the billing form that said, you know, I've agreed to pay to this. And then I had a thing said, you agree to be recorded because the doctors are using all this, all the tools you just mentioned. And I just think it is plain wrong that the doctors feel it's okay for them to record a patient, but that they would have the gall to say, no, patient, you can't record this for your own behalf. I just, I. Anyway, that's like, I can talk about that, but. But in terms of what I think is different from what I originally thought two years ago, I didn't know that I was going to feel so passionate about that and feel so much energy around making sure that people understood that people have a right to understand what their own medical care is.
- [31:16]
B
You tapped on something. So, again, full disclosure. Fred and I worked together at Pfizer more than 20 years ago, and we were both in a department called Business technology. And in order to be in that department, you had to have expertise in two things, technology for one, and anything else. So in my case, I was a physician with technology background. In Fred's case, he was a marketer with technology background. One of the core lessons that it took me a little while to learn was that when you were to be successful in that role, you had to get your head wrapped around the following statement. It's almost never the technology that's going to be the rate limiting step. And Fred just said it himself a moment ago, 20% of his time is thinking about the AI and about the tech. And the rest of it's thinking about patient empowerment, thinking about the business. And I think it's an important consideration for those of you in the audience who are thinking about startups or getting into this space, if you're not already realizing it, the tech is not going to be your problem. The tech is actually going to be the least of your problems. What's going to be the biggest of your problems? Finding staff who can help you. Finding people that aren't going to fight amongst each other. Figuring out ways to get your cash flow. Like all the nonsensical things that are not the fun stuff, the mundane, trivial stuff.
- [32:32]
C
And I'm just going to push back on you. All those things you said are trivial, mundane. But the customer, what does the customer need? What's the customer? That is neither trivial nor mundane. Correct. But it's really the core.
- [32:42]
B
And, but that was, that was one of the core BT lessons from the BT department was if you, if you. And in my, when I started, I was really enamored with technology and I led with technology. And pretty much every time I led with technology, it failed. Whatever the project was, it pretty much failed. For those of you who are out there listening and thinking about it, keep that in mind. The technology is probably not going to be what trips you up. It's probably going to be the rest of the rest of it. You know, we're here at Baker Tilly this morning, which is a risk management company and looking at those things, they're not sexy necessarily, but they're incredibly important. Understanding how to diagnose how your business is doing and what your. Where the risks lie and how to field them. That's, that's not a trivial pursuit by any stretch. Any thoughts, any additional thoughts on that topic? No, it's okay.
- [33:34]
C
No, I say a lot, but. But no, we covered it.
- [33:37]
B
Okay. So, you know, if you wanted to see Patient Talker again, you're a Relatively young startup. Can people see it today? Is it someplace they can grab it? Can they kick the tires?
- [33:54]
C
Yeah. So the point of this was not to be a commercial. I know, I know. But Patient Talker is available today on the App store. Available for iOS and Android. It is available. We have a very active open beta community. You can go download it and use it. I would say use it for two reasons. Not for two reasons. Use it in two ways. One is download, just play with it. A lot of people say, oh, I don't know if I can use it until my next doctor visit. No, it's your app on your phone. You can use it whenever you want. Just practice it. Get a sense of how it works so that when you do take it into doctor's office, you're not filling it for a first time. The other thing is think of it for yourself, but also the caregiver aspect. I had a relative who just last week had to have a series of medical procedures done, and there was an extended family, none of whom could all be in the same place at the same time, but all of which were keenly invested in wanting to know what happened during those visits. So recorded it. You can share on Page Dog, you can share easily and email people password protected and all those safety concerns applied. So my point is, think of it for yourself and think of it for other people who you might be caring for.
- [35:20]
B
So I've run through my list of questions and I think what we would like to do now is open up to the audience for some Q and A. And before I do, I just want to say thank you to Baker Tilly for hosting this live podcast here today. I forgot to do it at the opening, but I don't want to forget. They opened up their offices for us today. They gave us the nice breakfast, and I want to just give them a round of applause. Say thank you for being our host. And I'll also say this. This is the first time we've done practical AI in healthcare as a live event. So if you're interested in other events or you have an interest in anybody that you want to see, come on the podcast. You can find us at www.practicalaiinhealthcare.com. big surprise. There's a QR code on the screens here, and if you link on that, there'll be a form you could fill in and give us some suggestions. And we'd love to hear from you and we'd love you to join the podcast and join the conversation we have typically, like I said at the Beginning high level folks, CEOs, we have healthcare pundits, we have authors. Ing ho over here is an author and a healthcare pundit. She wrote a book recently. We did an interview with our two part interview with her back in the fall. And these are the kinds of folks we're looking for to bring on the podcast. We want to keep it interesting, we want to keep it deep, we want to keep it at a level where we're looking for signal through the noise. So with that, I'm going to say thank you for this and let's open it up to the audience for Q and a.
- [36:49]
E
Hi. Great talk. Thank you so much.
- [36:52]
B
Say who you are and where you're from.
- [36:53]
E
My name is April Smith Herak. I'm actually a former classmate of Leon's. So big shout out to him. I hope he's feeling better soon. And I run 99 directions. It's a strategic advisory. I spent over 20 years on the government side of healthcare and one of the things we always were concerned about was this idea of health literacy. And I think you're really in a position to bridge that with Health Talker. So what I'm curious to hear more about is patients enter the clinician's office with varying levels of knowledge and preparedness to engage with the medical system. And often when they get a really critical diagnosis like you were talking about with oncology or a big cardiology event, they're really not in a space where they can digest and understand the information. So more than just having a record of what was said, how is Patient Talker helping to bridge them and meet them where they're at, whether they're an md, accessing care for themselves or their family members who can really engage on a medical level or perhaps an immigrant where English is their second language and they're really system.
- [38:02]
C
Yep. So few thoughts to that Fantastic point. I'll sort of take it, take it in a few directions. One is helping prep talk to a number of doctors who say the patient comes in and say why are you here today? I hurt. Okay. What hurts? My body. Okay. What prompted you to come in today? My wife told me to. And I'm saying this in a funny way, but the intake history taking process can often feel like you're trying to draw things out. So this what seems like a simple thing of saying why do you want to go to your visit today? And then having a list of questions come out that for those of you who don't know health literacy principles, there's a whole set of guidelines about them Certain grade reading levels, certain presented in bulleted. So it's not a wall of text, graphic, et cetera. It's making all that accessible so patients walk in not having to worry about all the things that they may forget. You know that doorknob thing. Oh, right, I'm exiting the door.
- [39:07]
D
Dr.
- [39:08]
C
I always forgot about this. So that's a huge part of it. The other part of it is I think, Steve, I think you. You gave me this term was the in between time, right? So you're in a visit, then you go a long time. Presumably, you know, whether that's weeks or months or, you know, longer before your next slice of time where you're in a medical visit, things are happening then, and you're both searching for the information. That's where a record that's easily accessible, easily understandable comes into play. It's also where connecting dots can play. So as in our product roadmap is we're thinking about building out, capturing that in between time and documenting it. So when either you want to say or the doctor prompts you to say what's happened since your last visit, it's easily accessible to you and you don't have to feel like you're put on the spot.
- [40:05]
D
Hi, my name is Christina. I am from ERC lab at Max Design Studio, focused on health tech. So patient centricity is music to my ears. Very much appreciate that perspective. I'm curious, with the recession of trust that we have both within healthcare and the tech side of things, how are you addressing that with patient tomorrow? How do you think about that, how
- [40:27]
C
you be those trust? So the first thought I had was rebuilding trust means that you assume there was trust and it was lost somehow and that needs to be rebuilt. So there's a whole bunch of stuff there we could talk about. But I think the main point is trust is earned through transparency and through delivering value. Right? You're setting expectations that this tool or this app or whatever is going to provide you with certain things. And in return, you're giving up your name, your email address, but nothing else. Right? So I think it's whatever those things are, it's being really clear about what you own from the patient's perspective, what's the patient's owning, what they fully control and what's used how. And I think there is not a lot of forgiveness if trust is broken. Actually, I was thinking about this. This is a good chance for me to say something else. I was thinking about everyone always talks about the mvp. This is another lesson I learned. Everyone Talks about the minimum viable product. I think what I learned is that it's not, for me at least, it's not about the mvp, it's about. It was about the minimum trustable product. And because the first time someone uses this and it doesn't work, they're never going to pick up your app again. So you're nodding your head yes. I think before anything goes to market, it has to meet this minimum standard of being safe, secure, accurate, all those things that go into trust.
- [42:15]
D
Hi, I'm Sandy Ballou. Thank you so much for the talk. And I love how you're becoming a patient or are a patient autonomous advocate. I have a question about your product, Bromine app. I'm, you know, a lot of times the visit is focused on sort of the immediate symptoms, immediate condition, and I wonder whether Patient Talker provides an opportunity to bridge the gap between thinking about future implications based on the current condition. Like for, for example, you think about maternal health, they like hypertensive disorders in pregnancy, what that means for cardiovascular risk down road. But very few of these really talk about cardiovascular risk. Hv, eclampsia or whatever. So do you see that in your future?
- [42:58]
C
Yeah. So the way I'm thinking about what you just said is the difference between AI similar, you know, AI is so 2025, 2026 is about Agenic. Right. So it's not just what can the large language models do and gen AI do, but it's the idea that you have this agent running that will be working for you. And you have, whether it's Patient Talker or just, you know, this is, I think, a more broad answer. You have all this health information in various places around you on Page Talker. It's all on your phone, but it could be in other distributed places as well. If there is a tool that is always looking at that and saying, oh, your doctor said have an appointment in two months, it's been six weeks, you want to make sure it's on the calendar. We're looking at integrating tools that will do just that. And scheduling is, of course, one thing, but is it time for a refill, is it time for consultation with another specialist, et cetera?
- [43:59]
B
Yeah, I'm going to chime in on that one because it relates to some of the work we're doing up at Harvard. One of the things that has been traditionally really brutally hard to do is to look at the patient's record in a longitudinal way. Epic and other EMRs, and I don't mean to pick on Epic, but EMRs in general are not geared to do longitudinal views. They're geared to looking at the proximal timeframe of when a patient is being seen the last couple of visits. They're not really geared to look at, you know, what's going on over the course of decades, if you will, or even a couple of years. That's going to change. That's changing now because when you have these AI tools that can actually aggregate data, first of all, you have to aggregate the data, and then you have to bring it all together in a way that is digestible. And then you layer on this next tier of tools that can actually do those views of, like, if you're pregnant, chances are you're probably in your 20s or early 30s. And if you're trying to look and see what happened 10 or 15 years later, and did that have an impact on what happened to you in that next 10 or 15 years or 20 years? Traditionally, that's been really hard to do. It was almost impossible to do on paper. It's slightly easier with digital medical records. But this world of AI is going to make that an entirely different world, because now the data, the tools that are out there to aggregate information, have actually dropping the cost and the pain of that task dramatically. Next week or next couple of weeks, we're interviewing the CEO of a company called X Cures. What is X Cures? X Cures is a company that actually digests information from health information, exchanges with appropriate permissions, and creates these massive data sets that you can actually start to do what you just described. Now, are they doing across decades yet? No, they're not. Not decades, but they are doing it across smaller chunks of time. And they can do things like trend analysis in a much more comprehensive way than has been done before. So what I'm telling you is that not only are we reaching a new point with what AI can do today, but that's going to facilitate entire new business processes, entire new chunks of informatics that just haven't been possible in the past. Leon and I have this little ongoing thing. We look for themes in our. In our podcast. We do reflection episodes every five or six episodes, and we look for the threads that are tying things together or things that are coming out multiple times. One of the things that we've seen, and we're going to. I'll double down on this one, is that whenever a new technology comes on, the very first thing you do is you try to figure out how to do what you used to do with the old world, with the new tech. So, in other words, when locomotives came on the scene, they were, well, how do I get my product? Instead of using a horse drawn carriage, oh, I'll build a carriage that goes on a train. And that's going to change the nature of how I ship my product. Well, nobody may have thought of, there's might be a completely different way of doing that, which might be a different way of doing the planning of how the trains go or how they have hubs or how they have warehousing. They haven't thought of how to solve the problem differently yet. We're in that early stage right now with AI and we've seen this in the work we're doing at Harvard repeatedly. I don't think that we have even begin to hit the beginning of that curve of doing things so dramatically differently. I think we're at the very tip of the beginning of that whole exercise. And as that starts to happen, we're going to see things and no offense to Patient talker, but we're trying to. Patient Talker is trying to do something that we should have been doing 20 years ago, but there's going to be a different way of doing it in general in the not too distant future. I don't know what that's going to look like, but it's going to look probably dramatically different than what we have today. And maybe one of you in the audience is the one who's going to have thought of that. And you've got some AI powered company that's in its formative stages. But my, you know, Steve Jobs had this ad campaign called Think Different. I don't know if you remember that, but if you think different and you look at the new tech from that lens, from a different lens, you're very likely to see things that haven't even happened yet. And I think that's a critically important opportunity for this new space.
- [48:23]
F
Hi, my name is Debtorschuk.
- [48:25]
D
I manage the Marloma Investment Fund for
- [48:27]
F
the Marshmallow Research Foundation. Great interview. A short comment to what you just said. I think what's the possibility that we have for the AI is not just to have incremental treatments or solutions, but really think about, if we were to really start from a blank slate, what solution would we really implement here? Not just thinking incrementally, but really from a much bigger picture. But my question's a little bit different. Your story about your dad asking about the cake immediately was very much resonating to me. And I was wondering, with all of these probational models being so engaging, why is nobody yet using them or gnawing a solution for adherence and compliance. You having Pfizer, this is a huge, probably a 25% issue. So you all know it's for different reasons than patients are non compliant. So have you seen solutions here or is anybody tackling this?
- [49:39]
C
You chuckled, so I'll let you go first and then I'll take a tip.
- [49:42]
B
So when I started at Pfizer in 1997, the very first problem that we tried to tackle that I was involved in was an adherence problem. How do we get patients to take their drugs more regularly? At Pfizer, when Lipitor was in its prime, if patients took their Lipitor just one more month on a prescription, that would have represented something north of $1 billion of untapped revenue. So they looked at that problem quite seriously. I will sit here and tell you that I've probably evaluated my career upwards of 200 to 300 different adherence tools. I have seen exactly one in my entire career that has actually done anything close to adherence tool. And it turns out it didn't rely on AI. It didn't rely on anything fancy. What it relied on was behavioral economics. And from and by the way that many things caused me to stop what I'm doing and go read books and take college level courses or graduate level courses. When I ran into this particular solution, I did that. I dove in deep on behavioral economics. I could not understand having evaluated so many different types of tools in my career that this lady who had no knowledge of behavioral economics, had no knowledge of what she was even tripping into was what she had solved. And it turned in turned out that what she tripped into accidentally was this concept of nudge theory. So you're asking why aren't we trying to solve it with AI? The answer my perspective is I don't think it's an AI problem. I think it's a biopsychosocial. It's a psychosocial problem that can be solved by looking at it differently. And the difference here is behavioral economics that is looking at what are your incentives? What is driving you to make the why are you picking the cheeseburger instead of the salad? Why are you having a sugary treat when you're diabetic? What are the things that are causing you to make these decisions? Turns out that teasing apart human decision making, not so easy. In a couple of weeks we're going to have a guest on the show. The company is called Goodfire. They were a cold call, they came to us, somebody recommended them and we vetted them and we're going to have them on. They made a very interesting claim that they can dissect why a neural network makes the decisions that it makes. Now, for those of you who know anything about AI or know anything about neural networks, that's almost as difficult as figuring out how the human brain decides to have the cheeseburger versus the salad. These guys make a claim that they can do it. We're going to interrogate them very deeply. We want to understand how they can make that claim and make it stick. That's one of the reasons we do the podcast. We want to see the signal through the noise. I think adherence, patient adherence, is one of the most critical challenges that we've seen. And full disclosure, I used to work at the mmrf. Patients who are on cancer therapy who need to take their medicines, even they don't always take their medicines appropriately, which is crazy, because if they don't take it, they're going to get sick again. And it's not just sick, they're going to potentially die. So, again, I don't think AI is necessarily the answer. It can be a catalyst to get there, But I think layering it on and looking at it from other sciences and specifically, and I wish Leegan were here, because Ligon is a social scientist in his training and he could even comment more deeply on this. But that's, I think, where we need to focus. The AI will be a catalytic tool in that experience, but I don't think it, personally, I don't think it's going to be the thing that saves it, solves it.
- [53:07]
C
I agree with everything you said, so I'm not going to repeat everything, but I will say I do think that's the hallmark of what makes a good problem, is a problem that's been around for 20, 30, 100, 200 years. And it's a real problem. It's not AI looking for a solution. It's a problem out there. Having said that, I've seen maybe one or two more adherence programs that have worked than you have longer than six months. We can dive into roi. There are a few that have been good, but suffice it to say, it's a minority. And I do think AI opens up doors and opens up channels to do that in the moment. Highly customized, highly relevant nudges that patients and all of us need to get something done. And so I think, you know, AI is not a solution that does everything for everyone. But I think this is a problem that's been around long enough and that AI can, can actually have some interesting Impact on let's I'm just doing a time check. Yeah.
- [54:10]
B
We'll say one or two more questions. One more. Two more. Yeah, one more that time. Hi, my name is Lu Chi, I'm
- [54:17]
C
a founder in this space as well. I'd love to hear more about your go to market strategy. Are you billing for specific or patient innings whether that's or with this chronic condition and right. The medical specialty and publisher discover your product prevents your positions bestseller. Let's hear more. Sure. I can give you the two minute version now and you know after this we can over a cup of coffee talk more about it but I think you know it's the, the trying to be all things to all people, you know on one hand says oh sure this AI can do everything. That's that's not a good go to market strategy. I think where this solution fits the need really well is in complex high risk situation. So oncology, infectious diseases, rare diseases where people are really struggling for information. Geriatrics falls into that category as well. I think those are the places where it falls most naturally into patient groups. And then I look for the people with aligned interest to that. And so the list of people I mentioned before around insurers. I mean again I we have a whole, we have a whole pipeline of, of B2B. You know for the B2B to C. The B2B people who, who we think have a vested interest in doing this and might be natural partners of ours as we go in both try to. Sorry, you're, you're. You're leaning over and blocking him. Not to call you out on the podcast or anything but I think these people have a real alignment and so looking for them both as a top of funnel being able to acquisition get people in but also reinforcement. As you all know with apps, you know the number of people who use them once and then delete them is a high number. So having it built in, having behavioral economics ways to nudge you to keep using it be that the doctor says oh here's the QR code, remember let's use this app or a variety of other things. I think those are all good ways to both get and keep customers coming back.
- [56:40]
B
So I think we're going to call it a wrap for today. I want to do a couple of things. Number one, I want to thank Fred for being a wonderful guest and really unpacking a lot of work what his company's all about. I want to thank Matt from Baker Tilly for hosting us today. I also want to thank the hosts of New York Tech Week because none of this would have happened without New York Tech Week. And I want to encourage you to all to tune in, thank our guests. Thank our guests. Thank you all for being here today. Yep. And thank you for that. I'd like to encourage you to take a listen to the podcast. I think those of you who are interested in this space, you'll find it to be highly enlightening. I think you'll find interesting guests. We really strive hard to bring on the signal versus the noise. We really try hard to get, like the cream of the crop of folks in this field. And because I've been in the field for such a long time, and so is Leon, we have a nice rich pipeline of contacts that we know and we've been tapping that in. But we're going to run out of those soon enough. So if you have other recommendations, let us know. Thanks so much for coming today and hope to see you on the podcast. Thanks.
- [57:46]
C
So this can be, this can be off your podcast, Steve, now that you wrapped up for everyone in the audience, this is being recorded. Steve's going to go back, edit it. And when will, when will this podcast drop?
- [57:55]
B
Roughly a month. Roughly a month.
- [57:57]
C
So you can go on and listen to it again. Thank you.
- [58:01]
B
Let me turn this off.
- [58:04]
C
Thank you. Let's turn all this off.
- [58:13]
A
Thank you for joining us this week on Practical AI in Healthcare. If you're ready to go beyond buzzwords and hype and explore how AI is truly transforming healthcare. And stay tuned for more conversations that get us to what works. Until next time, stay practical.