Welcome to Real Talk on AI and I care with your host, Dr. Scott Morris. And I'm co host, Dr. Rehan Ahmed.
In each episode, we discuss current news, adding a little AI education, and debate innovative AI tools and topics that will change the industry. Let's get started with news of the week and it's got there are really two stories that caught my attention this week and both sort of signaled AIs where it's headed. Sort of not in theory, which is, you know, I get we get letters and emails saying like tell us more about where it's going in practice like how people actually use it. So there's there's two.
I think one is big and effectively whole AI field and one's a little bit smaller. But the first one is GPT five. So chat GPT, you know, it's on a version of 4.0. Now it's going to GPT five.
And this is coming out basically any day now they said it's reportedly being released in early to mid August. So I'm, you know, I Scott, I know you're a Gemini guy. Right? I know, but you know, I mean, anytime Sam Altman says something I kind of pay attention go, is this going to totally shock me or is he going to like amaze me with brilliance?
I never never really know where he's coming from. Yeah, you know, say, same with me. Sometimes I wonder how much fluff there is because you know he has to keep the enterprise value ever increasingly going up. But you know, obviously super smart dude.
And you know, I'm a, I'm a chat GPT person so so much so that you know, I had this thesis that I think we're going to. I want to switch the friction to me switching because today I was actually going through it. I'm going up and tangent sorry, but like I was going, I was talking chat GPT about. Schools from my daughter and public school private school all these things and I talked to like a therapist frankly, and I showed my wife like, hey, you know, CC what you know, like I talk what school should go to I'm going through all these like various iterations.
And I showed her the answer and she was blown away. She's like what and she's like what it knows my name like, yeah, I guess like a told her told you, you know, I told chat GPT, you know, you're one a wants and I know that kids say all these things and so. I really do believe that people are going to be segmented if people are going to to choose. They're going to be like fire socially or so far and they're going to choose one.
Because there's too much or it's such a cost. I'm so far down. It's kind of hard to switch. It's kind of hard.
So, but that's a chat GPT 5, so this is going to be different because it's being described as a multi modal foundation model. about this on the last, I think, news, not news bites, but the educational things we do. And this multi-modal means it's combining not just language, not just a large language model, but vision and what they call reasoning or planning into unified framework. So it's becoming less so a chatbot, it's much more than that, but we're there a model that can interpret quote unquote interpret text images and respond.
So what does it mean for I care? So it means that a patient's a customer can upload a picture of their red eye and it could tell them the symptoms. And it could look at the red eye and then correlate it with the symptoms and give sort of a differential diagnosis, probable diagnosis, explanations, et cetera. A company could potentially take that and sort of build upon it, but that's what multi-modal does really in terms of healthcare.
So this is really, I think, again, knocking on the doors of the next generation of AI triage tools like bots or assisted diagnostics, support systems, stuff that we've talked about a lot. I just think it makes it bigger and better when you talk about multi-modal, including images and potentially video and then of course. Well, I think we're going to talk in our knowledge bite and yet another news item bit when we get to our knowledge bite. I'm going to talk about how you use that multi-modal in terms of what evidence-based medicine really is going to look like in the next few years and is it really possible?
So let's let me finish up the news of the day and then we'll kind of get into the knowledge bite. Yeah. And so the second one is there's a paper that came out, they're using from Cedar Sinai. They published their data on an AI-powered virtual care platform called Cedar Sinai Connect.
And that was developed with another AI healthcare startup called KD Health. They did a ton of patient hours, 42,000 patient encounters and it generated diagnoses and treatment plans. And here's sort of, so a lot of people are doing studies like this, but this was a very large study of a very large and well-known academic center, of course. But here's what's most interesting about this.
They found that the AI's recommendations were actually rated optimal more frequently than the physicians working without it. So by 10%, so 77% respondents, patients prefer the AI responses compared to the physician-only responses. And it included a broad range of conditions, including eye conditions like red eye, eye pain blurry vision, visual disturbances. So this is probably the, in terms of one, I've seen a really high quality peer reviewed, a large publication using real-time AI assistance with actual patients and actual medical complaints, including eye complaints.
And so it challenges, I think, the assumption that AI and medicine really has to be narrowly scoped. This is really a generalist AI system that's looking yet acute conditions among thousands of people. It's probably tuned to some degree on health care, but it wasn't probably tuned specifically to eye care, right? So definitely, patients are already using it.
Doctors are already starting to trust this. And so I think, again, the knowledge is sort of building that, and again, it goes to what we talk about. It's not going to be whether AI or doctor, it's the doctors who use AI and use it intelligently and these types of systems are going to far, and academic systems and healthcare systems in general are going to far outpace those who sort of don't or slow walk it. It's that comment we made a couple weeks ago, AI plus HI human intelligence, that's going to be a pretty unbeatable combination.
For all of you listeners, you cannot get behind on this. You just can't, you've got to stay focused on this of where it's all going. So, brings up, and that we're going to get into knowledge bite because you kind of set me up perfectly, or Ann, I'm not sure if you knew you were going to do that or not, but today I wanted to talk about our knowledge bite of this week is evidence-based medicine. And we've talked about this, we've kind of worked around this subject a couple different times about what evidence-based medicine is, and it made me think about, one of my friends the other day, I was at a continued education meeting out in Oklahoma, and he said, well, tell me what evidence-based medicine is.
You keep referring to it on the podcast, and he goes, I'm not sure I really know what it is. It made me think, okay, then I really copy need to explain that. And this is my opinion, and I think evidence-based medicine might be much bigger than what I'm going to go over, but I got two minutes, so we'll try to knock this out. So, you know, I want to think about like a little bit of what is evidence-based medicine, and then my opinion kind of, how do I think it's going to change the future?
Well, evidence-based medicine, we often hear it referred to as EBM, is really kind of a different, let's say systematic approach to healthcare that takes the best available scientific evidence, and then may help us guide clinical decisions. Now, maybe evidence-based medicine is going to, A.I. is going to directly say, this is what you do, and we take humans out of it. I personally don't think that's going to happen, but it's not going to be just about what the doctor thinks is best, which is kind of what we do now.
If I say, if we're out, having a meeting somewhere, and I say some doctor says, well, this is what I think is best. I'm like, was that what the evidence says? Well, of course, the evidence. That's the worst evidence.
Well, that's what I think. Well, that's not evidence, that's your thought pattern. So, it's what we've always done, right? It's like what we've, the opposite is sort of, you know, what we've always done, that was the proof, and that's why you did certain things, is, you know, that's what they've been best in the work.
And what they've been doing. What they've been doing. Exactly right. Which is what we've been doing.
But I think the future is not going to be about what the doctor thinks is best. It's about taking what all the research says. And, you know, I'm not sure I completely, if all research is completely valid anyway, but you take that, and then you add that to the clinical clinicians kind of expertise or you augment knowledge as we often talk about. And then you take the next step, and maybe now with the help of AI, you can coordinate that with the patients, kind of what do they value?
I mean, what's their preferences of how they're treated? I had a patient today, he said, you know, like, well, here's your five options for treatment for this diagnosis, and they said, I don't do antibiotics. And I was like, okay, this gets a lot harder when we're not going to treat an infection with antibiotics, but here are some other alternatives. But, you know, I mean, now a patient will be able to put in their values and preferences.
And I'll help the provider and the patient go, well, based on what research says, these are some of the evidence-based medicine. I kind of think it's like this, I'm literally sitting on a three-legged stool right now, so it's appropriate when I said that. I was like, you know, you're going to take evidence, and you're going to take the clinician's skillset, their IE experience, and then you're going to take the patients' preferences or circumstances or choices, and we're going to roll this together, and we're going to get some, the most effective, kind of safe, and really personalized care possible. So, we're moving away from, let's call it traditional medicine, where the doctor knows best.
Well, do we really, and are we really taking the patient's perceptions into play? You know, I think we look at it and I go, I think that both augmented and artificial intelligence are really poised to help and revolutionize evidence-based medicine in some pretty profound ways. I mean, if you think about it this way, our biggest single biggest knowledge promise provider, A-Hon, is that sifting through this mountain of medical literature every day is basically impossible. And I think we might have talked to this once before, but this was a study done many years ago when I graduated school in the 90s, it said medical knowledge was doubling every 20 years.
You know, then you move forward to the 2010s in medical knowledge is doubling every seven years. Now you move to the 2020s in medical knowledge is doubling every, I think it's 71 or 73 days. So once every three months, medical knowledge, that we need to know as a provider in our industry is doubling. And as evidence-based medicine and all the stuff that we talk about, A-Hon, I think medical knowledge is gonna double every day within a decade, every day.
So how can we do that? Well, when you think about these large multimodal language models that are out there, I mean, they're gonna be able to search, summarize, synthesize that data in seconds. I mean, seconds, way beyond our capability to do that. But that's just the processing side.
You still have to take into, there's a human being sitting in front of us. And you know, I was just reading about it yesterday as there's a company in IKare now called Amoros, AMAROS, not involved with them in any way, say, perform. But they're already starting to do this where they're actually able, their AI systems able to synthesize all of the data about a particular subject and feed it to you or feed it to me as a provider. And that's giving me way faster access to data.
I mean, boom, instantly done. I can look at any part of relevant up-to-date research. You know, not, and I was thinking with this way, it's not something I learned six months ago at a conference and maybe I remember it and maybe I remember it correctly or don't, or maybe it's not something I learned 20 years ago in school, but it's something that changed today. You know, that's a big change of what AI is going to help us to do in terms of bringing evidence-based medicine to the peoples, which is all of our patients and our families and our friends and everybody else.
And I think that evidence-based medicine, when it fully kind of starts getting rolling, I mean, this could change, I mean, this could change the financial picture by hundreds of millions, if not billions of dollars. Because no longer will we prescribing some treatment that's old or doesn't work, just because we learned it school 20 years ago or 10 years ago or somebody told us about it at a conference and it wasn't really true. Now all of a sudden, doing those unnecessary procedures or prescribing drugs that just simply don't work or there's better drugs that are, you know, better effect for same price. I mean, all of those are going to kind of change the standard of care.
I think that's really the big part of this is that evidence-based medicine is going to slowly or maybe not so slowly change the standard of care. Because you had mentioned it earlier, you said, well, when thousands of patients, I look at and go, I think in a few years with some of these AFIS systems that are coming out, these advanced or adaptive systems, you're going to be going through patients, you're going to go through hundreds of thousands, if not millions of patient data points and see what really does work. Right now, we don't know, we count on our scientific studies and those are all somewhat biased. But, you know, I mean, soon we'll have hundreds of thousands or millions of exams that we can sift the data through and say, what really does work?
What's in an unbiased way that still maintains, you know, data privacy? What really does work? And I think when we can all start building upon that evidence, that may really change what we do. I absolutely love it.
And I think as you rightly pointed out, AI is built upon data and evidence is nothing more than data. And so it's just a perfect, a perfect marriage and the three-legged stool of evidence, expertise, and patient values. I think it's a great framework to sort of think about this in a sort of humanistic way and how actually physicians and patients are going to think about this. So, I hope so.
You know, I mean, evidence-based medicine is just a piece of that knowledge tool that's going to augment to all of us as providers. And then use our human to human element to actually cause change. So, wait, you know, so I have to say this, man. So you asked me before, you know, the podcast, you said, what do we do as a real talking?
You know, I've been really focused on insurance and I'm not sure almost to the point of driving me crazy lately. And I started thinking about, well, how can AI really change the whole insurance third party primary care, you know, Medicare, Medicaid coverage? And over the last couple of weeks as I've been developing algorithms for my own project, I've really spent, I mean, I'm probably $100 into just digging deep into how insurance really works. Or sometimes how it really is built to not work.
And you're not talking about... We talk about preauthorizations and voice bots. Well, I don't know if you saw this kind of news that last week, I guess, you see RFK, he said last week that he's trying to make preauthorizations go away. Then they're going to start these pilot projects.
And, you know, I mean, if your insurance company, you sign with the insurance company, they say they're going to cover it. Why the hell do I need a preauthorization for that? That's part of the plan, right? And so...
And there's where I think evidence-based medicine is going to help drive insurance coverage and go, hey, there's better ways. There's way A, there's way B, and there's way C. Way A costs X, way B, cost 2X, and way C cost 3X. But which one's going to give us the best results?
And I think the whole preauthorization, you know, and then let's talk about, I guess, let's talk about the way it is, right? And so I think all of us are pretty familiar with that, but then we can poke holes in it. It is, right now, if you want to, as a provider, right, you make a diagnosis and then you're a treatment, and you hope that your treatment, especially if it's a procedure, surgical procedure, or that you hope it's going to be covered, and you hope the diagnostic test that you did to do that is going to be covered. And in some case, you have to go get a preauthorization, I don't know about you, but I can't tell you how many times we've got a preauthorization, we have a preauthorization number.
We send it in. When we do the thing, they come back and go, yeah, it's not covered. I'm like, we preauthorize it, yeah, we changed our mind. And I'm like, then what the hell's the point of a preauthorization?
I mean, I, not making any political judgment, but it calls, I'm kind of in favor of making preauthorizations go away. If you're covered, you're covered. 100%. And you know, the preauthorizations, I think the critique is that they've been used as a way to obviously create friction between the provider and doing a procedure.
I mean, it's not, it doesn't take a, it doesn't take a neurosurgeon to figure that out. That's true. To figure that out. It's just another step in a very complicated, convoluted cycle to prevent what may need to be done.
Of course, that's not what the insurance companies would say, but as a physician, providers, and we all feel that these are here to make it harder so that we just get so tired that we just stop. Yeah, and I think I love your, I love the wording you said, a complicated, convoluted cycle. As I sat there and unfortunately for listeners, I'm a freaking geek. But you know, these big huge algorithms that have arrows going everywhere and these are workflows and processes and trying to figure all this out, it is the most complicated, convoluted process.
I have ever seen when you draw it all out. I mean, and I think it's made, as you said, I think it might be made to do that because you have friction in. So what do we do? We do the whole process.
We're trying to help our patient. We do what we think is best. We send it in, it gets denied. And so at that point, as a provider, maybe I don't even know it gets denied because maybe my billing person says all it was denied and maybe they refile and maybe they don't.
And if they don't, we just lost all that money. But maybe they refile and then it's denied again. And then you're like, guys, well, it's denied twice. Do I keep trying?
And so ask the lady who does all my billing. So how many times sometimes you have to try? And she goes, four is the magic number. By the time I submit it, the fourth time it almost always gets approving.
I'm like, four times. How much time does it take you to do everyone's authorization? She goes, I don't know, five, seven minutes. I'm like, oh my gosh, that's a lot of freaking time.
And I think about with where AI is. And I'm not so sure health insurance companies are gonna be all hot to try about implementing an AI to help our process. They're gonna be about driving their business model totally. But are they really gonna want the ability to have hyper personalized coverage?
Like where we go, this person has this plan. It covers X because I go back and forth, and I'm curious about your opinion. Sometimes I think that, you know, like United Healthcare. And obviously they've been in the news a lot the last couple of years, but it's actually the last year.
But you know, they have thousands of different plans under their umbrella. So the reality, should we really have three or four plans or should we really have millions of plans? Based on each person gets to design their own plan. They, I enough, are the companies, for the insurance companies could literally design perfect plans for that person with incentives to try to, hey, if you submit your data through a wearable or, you know, you, you're compliant with your medications or, you know, you, you, you, hit your draw, you know, you don't always choose the most expensive drug or you're eating the right foods or whatever.
Could we literally truly have a hyper personalized insurance coverage? I think that's possible. What are your thoughts? Well, that's a very fascinating something.
I have not truly given a lot of thoughts. Because to me, you know, obviously this is a, insurance is all about risk. And the way you increase your decreased risk is with, and we started talking about before, it's with data. And when you have information asymmetry, that's what one person has an advantage of for the other.
Right? So insurance companies, they sort of like look at things that sort of risk adjust basis and they sort of try to figure out, you know, how many people are going to cover and who they're going to cover and what their ages are, et cetera, et cetera. And they get pools together to make sure their risk tolerance is satisfied to, to their liking type of thing. And so it makes sense.
It's a business at the end of the day. Those are like big, huge actual aerial tables, right? But right. And so, oh my god, you're sick.
Yeah. What you're saying is if we have, if we're in a system where, you know, there's just so much individualized data and the patient has maybe a ownership of that, I mean, it sort of turns the whole system, I think upside down a little bit. I'm not sure. Well, you know, my philosophy on that, Rayhan, is I believe that, you know, I'm not a big pro EHR person.
I believe that patients should have access to their own data and it's all be held by them. But that's the only way you're going to get to this kind of hyper personalized coverage. I know that's kind of a way out there thought, but I've been thinking a lot about that lately. You know, I mean, we kind of start out going, you know, what if we can automate and kind of expedite prioritizations?
If not, just get rid of them all together. That might be best of all. But I'm not so sure the insurance company is going to want to do that because they want to have that. That's one more loss of control that they have.
I kind of say, well, maybe instead of getting rid of it, we make it real time. You know, and I always look at that and go, I think there's two parts of this, right? So there's, there's, ours is somebody eligible for any form of coverage, right? So the first thing we always have to look at is eligibility.
And that's a pretty straightforward. Either you're signed up for a plan or you're not. I think that one's real time eligibility should be something that, that's not a question mark. Like right now, I think that's not a question mark.
That's not a question mark, but I think you know, I think that's a very important thing to be aware of. That's a very important thing to be aware of. I think that's a very important thing to be aware of. It's something that, that's not a question mark.
Like right now, many times we don't find somebody's eligible. We go through tri-zettel or some other program out there and find out do they have eligibility, but that doesn't, and it tells us what their benefits are, but it doesn't always guarantee benefits, right? So then we do the prior authorization in hopes of potentially guaranteeing benefit payment, though, you know, the fine print always says not approved until, you know, payment went out. And so, you know, how we fix that where we have now, real-time eligibility.
And I think we're getting really close to that. And real-time prior authorization that's expedited instantaneous. Now, on the other hand, I do think that the insurance companies should be able to say, hey, this is experimental. It might be covered, but patient you may have to pay more than something that's not, right?
I mean, it is a business, as you said earlier, right? I mean, it is a business. And I want insurance companies to be around just to help protect all of us. But we got to have, we can't take all power away from them.
So we have to give them the things they say, really, if we have evidence-based medicine, like we talked in the knowledge, but couldn't they then say, hey, if you want to have something that doesn't have evidence, but you firmly believe as a provider and a patient is a combo, that this is a better procedure, that you can do it, we'll cover you up to the cost of the standard. And anything above and beyond that is now you guys have to work it out, is what your auto-pocket costs are going to be. It's a great way to collect more data, to have more evidence, and it's not necessarily increasing risk for the insurance company, and it's decreasing overall costs for the patient, because instead of them getting nothing for an experimental, they're at least getting what the evidence-based medicine treatment might be. I don't know.
I don't know. I mean, I think that there's that, and then I also wonder, why does it take so long? Why do we submit something and it takes a week, two weeks, 30 days, 60 days, to get through a processing period? I can conduct financial transactions in my bank and have it in there in 10 seconds, right?
I mean, I think about, my son was in Europe this week, and dad, I need more money, Shokr. You know, in Yvendmo, he got it 30 seconds later, right? Why does this have to take so long in insurance? I mean, you can't tell me there's a human being processing on the insurance company, every one of these claims and looking going, yeah, we accept it or no, we don't.
That's gotta be automated. And if it's automated, why can't it be instantaneous? I just- Yeah, I mean, I think we're, we've mentioned companies on this podcast before that have the pre-authorization bots that go and call insurance companies, and there's a clinic I work with that's, that's starting to use that and spin fantastic. But you just wonder, you sort of do some second order thinking here and I go, well, soon the insurance companies are gonna have their, obviously, their voice bots, and you're gonna have two voice bots basically talking to each other and they're gonna go through their iterations.
But then why do you even need voice bots? Like, it seems so ridiculous to have two AI agents pick up a phone literally on the VIP and call each other. Like, why, wait, well, they can't- That's the pro-processing. They just share, and then you just do, they should just figure it out in advance.
Right. I agree. I mean, why would we waste time on a phone call? I mean, just, it's digital.
There's going to be rules. You set the rules. And, you know, everybody starts playing by the rules. But then, you know, if the insurance company changes the rules, our agents say, wait, rules changed.
We got to fix our rules. I think we're worried about- I'm- I'm worried about insurance companies as more of the people who are in the industry are in the industry. I think we're worried about insurance companies as more providers, especially in our field, and optometrometer optimality start using these. I wouldn't be surprised if they came out and said, nope, you are not allowed to contact us with an AI agent.
Like, that is often- So they're just going to put another friction in the rules that you have to be a human. And so I'm- I'm- I don't know, but, Rayhan, you know, I think about that too. I really do, but then, you know, I always believe I'm a little bit of a capitalist, I guess, but an evolutionary capitalist is if they say that, that's going to provide an opportunity for a third party, a different company to start in the insurance market and go, we're okay with that. Yeah.
And if there- there'll be so much faster and so much more efficient in their process because they're going to use AI plus HI to change the healthcare insurance market, right? I kind of wonder if- You know, we look at the big- You know, we look at CMS, and we look at, you know, the big three or four of United Healthcare Group and Blue Cross and Etna and sign in all those guys and go, yeah, but they're kind of the dinosaurs, right? I mean, they've been around a long time, are they really keeping up with everything that's going on? Or is there going to be some new player that's going to come in and completely disrupt the market?
I think AI might just give opportunity for a completely different insurance landscape. In the next couple of years, for those companies who are willing to invest in that hunt that down. I think it's huge. I wish I had a couple of billion dollars to just start one because I think we could do a better.
I really do. And then what we're doing right now. But I do believe, Rayhan, that AI is going to change that entire- what was the word to use? Complex, convoluted process.
Anytime there's a whole bunch of curly cubes and something, there's somebody that's going to make a straight line and make it faster, better, quicker. I just, I don't think, I don't know if it's happened yet, but I sure think it's going to. Hopefully somebody listened to this podcast to decide to take a shot at and go, hey, let's go do this. Well, hey, everybody, then you got your next week.
You got to stay tuned to this week was Evans Space Medicine. Next week, we're going to get into the really meaty stuff of value-based medicine. For this and all of our content, whether it be our innovators podcast, our real talk, and all the, I mean, geez, I don't know, Rand, if you've looked at the website lately, we have so many articles by so many different people coming in and different opinions. I really love reading it now and go.
There's so much good stuff on there. When we started this, geez, what, seven, eight months ago, we just hoped we'd get here. And I think we're here. And for any of you who want to write, please don't hesitate to contact Rayhan or myself.
And we are super happy to have writers give their opinions on how AI is going to has changed their life and how it might change the life of all of us in the future. You've been listening to Real Talk, an AI and Icare, your weekly podcast to keep you informed about AI technologies revolutionizing Icare.