[Automated Transcript]

Earl: I mean, to practice medicine is to, assume some risk because, invariably you're gonna be wrong at some point in your career. And you can also be sued when you're perfectly right, and, things either by chance didn't turn out or something was misperceived and, it's kind of a, a messy world. But, back in the day before all of the internet and the AI a- and everything, people used to ask around, "Who's a good doctor?" And they would make their judgments partially on personality and partially on outcomes and, partially by what was heard in church.

Earl: And the whole system wasn't perfect, but kinda worked. 

Deep: Yeah. 

Earl: And, and now we have this, information on steroids, and we're trying to find a new equilibrium. 

Deep: Hello, I'm Deep Dhillon, your host, and today on Your AI Injection we'll be exploring how AI is driving real and measurable outcomes in healthcare with Earl Campazzi, author of Better Health with AI.

Deep: Dr. Campazzi is a concierge physician who holds an MD from the University of Pittsburgh School of Medicine, as well as graduate degrees from Johns Hopkins University and Duke University. Dr. Campazzi focuses on turning AI into practical real world impact in healthcare. Dr. Campazzi, thanks so much for coming on the show.

Earl: Well, thanks for having me. It's my pleasure. 

Deep: Fantastic. So, maybe start us off and tell us, what is a concierge physician, and then maybe walk us through how do people typically manage their healthcare today, and what does that process look like, where does it fall short, and what changes when you get involved and when AI becomes a part of that?

Earl: Sure. Well, concierge is, uh, a little of the old time and, and also a new concept that it's a membership base, much like joining a club. So people pay an, an annual fee, and because there's a fee, we have a smaller group of patients and are more available that we don't have a voice tree. We... or a phone tree.

Earl: Uh, we, we actually answer the phone. We're available on evenings and weekends and, uh, the way, doctors used to be. So it's, it's kind of a nice, uh, to practice in that environment because the finances are pretty much taken care of with the membership fee, and we just focus on taking good care of the patients.

Earl: And so, uh, that- that's concierge medicine, and you were asking how AI is starting to impact it? 

Deep: Yeah, but maybe before that, tell us- Yeah ... like what's the problem that you're seeing out there? Like what are people doing who aren't using your approach and, where is it falling short?

Deep: And then maybe tell us a little bit about the solution. 

Earl: One of the big areas where things are falling short is in supplements, uh, because people are going by, you know, all kinds of recommendations from their cousin to, you know, uh, podcasts to commercials to marketing of all kinds. And, uh, uh, I say if, uh, if you had a new Ferrari and your brother-in-law came with a, a gas additive, you would never put it in your tank.

Earl: But yet people are putting all kinds of supplements, a lot of which, you know, aren't doing,, any harm, but, a few of which are really pretty potent and can get people in trouble. So that's one area. And then I think, they're just not taking as big of advantage of the wearables and the smartwatches that are really recording a lot of data, and you can get some insights.

Earl: But you know, uh, in your email to me, you mentioned, a overwhelming amount of data from this and, and that's a problem, is AI can help you get your arms around it, whereas no doctor really could. You know, we're getting pulse readings, uh, you know, every minute, et cetera. 

Deep: So you mentioned the... mean, and maybe I'll kinda project a little bit, but, like, typically- Yeah ... you know, physicians have 15 minutes to see the patient. It's very much- Yeah ... problem and immediate sort of solution- Yeah ... oriented. Typically, physicians have a few tools in their toolkit.

Deep: It's either an RX. They'll write a prescription for something. They'll hand you off to a PT, or they hand you off to a specialist. Or they send you off for some labs and basically do those things later. What's, what sorta happens in, in, in your kinda clinical setting that's, that's different 

Earl: i... We have a lot less time pressure. The, the 15 minutes is a little bit generous because it's 15 minutes to see the patient and also document. So, uh, the actual face-to-face time of listening and, uh, examining the patient is even less than 15 minutes. And I've practiced in that environment, and you can't blame the doctor.

Earl: It's, they have to see 36 patients, in a day, and you have to do it that fast, uh, in order to keep up, otherwise you're... By the end of the day, you have a two, three-hour wait for you. And that's a luxury that we have. I mean, essentially our patients are, you know, w- wealthier and, you know, are getting better than average healthcare because of that.

Earl: I have to be honest about it. But we can listen and, ask more probing questions and just, uh, you know, do it slower and more in depth. And, uh, that's a big asset. 

Deep: Yeah. And then I imagine you also build relationships with some of these folks over time. 

Earl: Yeah. I, in fact, I, you know, I've even, my wife and I have traveled with our patients.

Earl: We've gone to dinner with, oh, gosh, a couple dozen different couples o- over the years. Uh, sometimes we invite them, they invite us. But they've, uh, really, grown into friendships, which uh, is very nice. And then, you know, we have close to 500 patients with, uh, three MDs and two nurse practitioners, but w- we know everyone as they walk in the door.

Earl: It's not like, "What, what's your name?" And it's, it, in that, it's, it's the old school kind of practice. 

Deep: And maybe tell us a little bit about what do you think happens when you have a, a long multi-year relationship with a patient that doesn't happen in a typical physician- Yeah ... bouncing setting or, you know?

Earl: I'll, I'll tell you, we had a, a, an elderly patient, he's, uh, was in his mid-80s and he came in a bit pale, said that he wasn't feeling good, and we did an EKG. The EKG wasn't that bad. Uh, his vital signs weren't that bad, but we didn't like the look of him because we'd seen him many times, and he, uh, something was off.

Earl: So we sent him to the emergency room, and I actually went there with my nurse, and the, uh, ER doc started to give us a little hassle and said, you know, "He really doesn't need to be here, and we're gonna send him home as soon as we get the CT." But and lo and behold, he was having a, a dissecting aneurysm, had to go right to surgery.

Earl: Fortunately, had a great outcome, uh, uh, but they literally had to stop his heart to repair the aorta and then restarted his heart, which is, uh, amazing physicians, amazing surgeons that did that. But just by that kind of gestalt thing, uh, something doesn't seem right, is what led us to send him, which led to the operation, and that, uh, that's probably the best of, of all examples of, uh, where that subtlety really made a difference.

Deep: And you think that's because you guys had a baseline understanding- 

Earl: Yeah 

Deep: ... of what this patient looks like, as opposed to- 

Deep: Yeah ... a, a physician maybe seeing him for the first time, thinking, "Maybe this- Yeah ... patient always looks like this," or whatever. 

Earl: Yeah, because then you go by the the exam and the EKG and the vital signs, none of which were that bad to really, uh, declare an emergency.

Deep: I wanna kind of bring up another scenario. I, I will sort of fully admit that I'm privileged in the sense that we have a, a lot of physicians in the family. Ah. And so whenever I'm, whenever I have to deal with some complex healthcare scenario, like my, you know, my father was in the ER and, and I'm talking to them and trying to, like, navigate what's going on and I'm being told that, "We're just gonna send him home.

Deep: We're gonna send him home." And I'm like, "Yeah, this is, this is not right." So I call up my sister. My sister's a physician. She gets on the horn with the doc. They, "Do y- you guys have this-" Back and forth. It's like machines talking to each other really fast and- 

Earl: Yeah 

Deep: ... and all of a sudden, you know, within two minutes he looks at me and he's like, "We're gonna keep him for a few days.

Deep: We're gonna do this, that, and the other. We're gonna run these tests. We're gonna run that." I don't know what exactly happens in that handshake, but I have some theories. One of the things I think that happens is the physician who's attending the patient knows they're being watched pretty closely

Deep: Yeah. And I think that, I mean, I think that just changes the dynamic. I think the second thing- Mm-hmm ... is they got somebody to bounce a longer history with, because- Mm-hmm ... you know, anyone who's representing somebody in our family knows their full history, as opposed to- and so I'm curious, like, in your practice, are you guys playing that role?

Deep: Are you playing a role- Yeah ... where your patient winds up, in the ER or with a specialist and you get to actually jump on the call and be there or help out? Well, y- 

Earl: we're most often with them. Uh, that's one of the things our patients like the best is if, if they need to go to the ER, first we try to keep them out of the ER.

Earl: If we're, they have a small aspiration, we'll put a suture in it or, you know, or whatever we can do to avoid the ER. But then if they do need to go, say, for chest pain, uh, we go with them, and w- we do exactly that kind of conversation, uh, often not in front of the patient, but we'll walk to the to the doctor's area in the ER.

Earl: And we also know a lot of the ER docs because we tend to use only a couple of hospitals. So all of that, and, you know, the answer is a little bit of all of the above. It's, you know, some deference to a colleague that's saying this might be serious. It's some, you know, liability concern that, you know I'm being watched, um, having, you know, somebody with equal credentials saying something different.

Earl: Uh, you know, and then also, and w- where we really make friends with especially the nurses is we bring in, uh, a paper copy of their medical history, not, not hundreds of pages, but like a three-page summary, and, uh, that, that's a good way to start off because it makes the nurse's job easier. You make friends with the nurses and, and life is much easier.

Deep: Yeah, and I think correct me if I'm wrong, but the literature really supports, um, the idea of kind of team medicine, where if more than one person is involved in a conversation, a dialogue about a patient's situation, whether it's a surgery context or, or- Yeah ... you know, a diagnosis context, that the care level and the error is reduced and the efficacy goes up.

Deep: Yeah. Okay, so now- I think it's kinda, you, you mentioned by saying that, you know, you got wealthy patients and, um, I get that. But I think getting this level of care to everyone seems like a good goal, which I think you probably share 'cause you're publishing books and trying to get a lot of this information out there.

Deep: Um- 

Earl: That's exactly right. Yeah. 

Deep: But part of that is maybe not in the patients themselves side of it directly, but in, like, maybe the clinic operational efficiency side or in the physician's assistant side. So for example, like in my company, you know, we do a lot of consulting work and helping products companies out, but we've, we've done a lot of medical applications.

Deep: We, we work with a company that builds a s- um, that builds a smart toilet. So everything, uh, or, you know, a lot of what a gastroenterologist can figure out by looking in your toilet. But, you know, obviously they don't actually do that 24/7, but the system does. We did a like a, uh, let's see. We built a system with a cardiologist to kind of do a lot of kind of more nuanced heartbeat anomaly detection from an iPhone case, for example.

Deep: So there's a whole, a- a- and we've done a number of these kind of physician's assistants kinds of projects. And I'm curious are you seeing can you, are you seeing the ability to maybe take your business and maybe take it down an income strata by, like, incorporating a lot of these sorts of efficiencies into your business?

Earl: Um, we have a, a really healthy demand. Uh, we've sort of gone, go on and off a wait list. So we're ... But we're not contemplating going down on the price point. But as far as efficiency within, uh, with our patients' permission, uh, we have AI listening now. It's, uh, it's only been for last few months, and it generates the draft note, which is, you know, w- writing a note is kind of the laborious part of seeing patients.

Earl: They, you have to be very detailed and and the drafts that our AI are producing through the electronic medical record system that we have are really good. Now, they need editing, but they're really good. Uh, and that's playing out, uh, throughout healthcare that so much at a time... I, I went to a conference and they said, "I'm gonna teach you how to see a patient in seven minutes, uh, write the note in seven minutes, and that'll save one minute.

Earl: That'll add up to four minutes an hour so you can go to the bathroom." And I'm like, "This is the wrong conference for me." Um, but, but, um, but this, it's that kind of, , time commitment to writing the note which was equal to seeing the patient is the point there. And, so, the efficiency AI i- in a number of ways, and, you know, I can go into- Yeah

Earl: other examples. Yep. 

Deep: Yeah, I mean, my physician uses that system as well. I think it's pretty common. Yeah. And I actually have a friend that started a company that does that, and I think it, it's still running pretty successfully. Yeah. One of the things that you kind of pointed out as one of your guys' key roles is, is, like, having patients have more time to ask you questions and really kind of- Yeah

Deep: understand, um, their- Mm-hmm ... healthcare and in sort of a more nuanced, maybe detailed way than they can get in the 15-minute, you know, once-a-month kind of scenario that usually happens in a clinic. So I guess my question there is, um- For those not in your context I think a lot of people are just leaning on the LLMs now.

Deep: They're talking to Gemini, they're talking to GPT or Claude. And I'm curious from your vantage as a physician, how did we go from an arena where the FDA has everything to say about what a phys- what medical information a patient can get, to a world where the FDA has virtually nothing to say about any of the LLMs?

Deep: And they're absolutely 100% giving all kinds of medical advice, 'cause I know 'cause I use it all the time. Yeah. And you know, I vet things and like check it out, and generally they're really quite reasonable, and it's much better than what people were doing before with Google and suddenly, you know, diagnosing themselves into oblivion.

Deep: But I'm curious, like, what's your take on the FDA's lack of a role there, or- 

Earl: Well, I mean, I, I think it's a bigger societal thing that, uh, the government is grappling with how much to regulate AI because, you know, it's having explosive growth that it, it wouldn't if it were regulated, uh, et cetera. But, it, you know, in medicine probably 90, 95% of the time it's right.

Earl: Uh, it, you know, because it's pulling references, et cetera. But the trouble is it gives the same degree of confidence when it's absolutely wrong. And y- you read the answer and it is very convincing, and this is why it's kind of repetitive in my book, do this in, you know, as a team with your physician, you know?

Earl: Use AI to come up with ideas or to organize your information so that when you go see your physician, it's, more efficient. But don't, uh, you know, I said it as many times in as many different ways as I can, don't just, think AI is your new doctor because it can really lead you astray with sometimes it, it's called, um- hallucinating where it just makes stuff up. I've seen it make references up and sometimes it just, takes something out of context or you don't give it all the information, whatever the reason, but it can be flat wrong in, in a dangerous way.

Earl: And you won't know it as a consumer because it'll look just like all of the right answers. 

Deep: Yeah. I mean, I think one of the things that highly trained humans are good at, you know, let's say, uh, physicians in this context- 

Earl: Yeah ... 

Deep: is knowing when to ask follow-up questions, and it's one of the things- Mm-hmm

Deep: that the AI systems are frankly terrible at because they're super overly biased towards giving as thorough a response as they can given whatever information they have- Mm-hmm ... which almost never ends in them asking follow-up questions. So where the- Yeah ... appropriate question might be, "I need to send you for labs, uh, and I'm not gonna say anything else," the AI- Mm-hmm

Deep: will just prattle on about and have all kinds of speculations, right? 

Earl: There, there is a, a better way, uh, rather than asking AI questions, uh, to give AI a goal. Say, uh, say you, it's a sad circumstance where your wife was just diagnosed with breast cancer. Instead of saying, you know, "Where could she get the best treatment?"

Earl: or, "What's the best treatment for breast cancer?" Say, you know, "My goal is to get the best treatment for my wife. She has this type of breast cancer." What else do you need to know and generate a, query for yourself, and it'll generate, ... It'll probably ask you some follow-up questions, which is exactly what you're talking about, and then it'll generate, uh, you know, a question, and AI is formally named a query.

Earl: So it'll generate its own question, uh, which might be a couple of paragraphs long, and then the answer to that is a far better, far richer, and far more personalized answer, and i- it's called meta-prompting is the, is the technical term. 

Deep: Yeah. I mean, I think a lot of, a lot of a lot of the kind of state of the art right now in terms of how to interact- Yeah

Deep: with these systems is to focus less on question answering and more on getting into the right feedback loop. So- 

Earl: Yep ... 

Deep: you know, like, okay, ask me a line of questions until you feel like you have sufficient information, and then tell me that, but don't answer my question. And so a lot of, like, figuring out what the feedback loops are that are going to lead to answers, and that's something that's very unintuitive for end users.

Earl: Sure. 

Deep: But, you know, I mean, the systems are gonna get better at figuring out when to facilitate the entry of that kind of thing, I think. 

Earl: Yeah. 

Deep: So, okay. Well, maybe, So going back to the, the FDA question, though. Yes. Like, do you ... You know, like when we think about devices, the FDA has this, you know, multi-year very exhaustive process for navigating that.

Deep: When we think about, like getting a pacemaker on the market, I mean, that's gonna take you at least 10 years by the time you get- Mm-hmm ... something on the market. So, but when it comes to advice, the FDA used to be really strict about regulating it, and as a physician, do you think the FDA needs to step in and, like, intercede at the foundation model regulatory level?

Deep: Because when ... I mean, like I'll, I'll use a very specific example. The foundation models were sort of notorious. This came out maybe like four, five, six months ago for psychiatric patients with s- serious mental health disorders. Well, then one of the things that the systems loved to do was play along with their delusion, which is probably the worst thing you could possibly do with somebody, you know, suffering from a severe case of bipolar or, you know, or any number of, of h- mental health conditions.

Deep: That that is a problem, like a really big problem, and these folks are already maybe, um, challenged for social interactions. And so now they're like ... And as society is increasingly leaning on bots for social fulfillment, if you will, it just seems like- you know, there's, there's just this vast arena for problems, and I'm curious what your take is.

Earl: I, I, I totally agree, and, and that's like the one black box warning that I, I put in my book was on, uh, mental health care because there uh, a couple of examples of, somebody in their late teens, may- may be a little so- socially awkward and that the AI programming was kind of to agree with people, and they would say, "You know, I was, uh, depressed," and, "Oh, I'm sorry to hear that," and, "I can see why you're depressed," and, "Oh, gosh, I'm thinking of ending it all," and, you know, "Oh, yeah, I'm sorry to hear that," and, you know.

Earl: And then it got into maybe I can help you with that," and it led to some r- suicidal planning that then the, the, uh, teen took his own life. And so now AI does... I- is very sensitive to anything about suicide, that it will put the brakes on and say that this, um, uh, I shouldn't say AI in general, but at least some of the chatbots that, that I've been using are very sensitive and say, "W- we won't answer anything with regard to suicide," "Please call a mental health line."

Earl: And, uh, that, that's one, one big warning. I mean, there's some warnings about privacy, so, uh there is the opportunity that the FDA could step in other areas besides mental health to put some curbs on, on the advice. But, you know, the FDA is kind of funny because, again, we're going back to supplements.

Earl: Supplements are regulated like food, uh, where if they don't cause immediate harm, uh, then, uh, pretty much has to be proven that they are harmful in the long run to take them off the market rather than like, prescription drugs where have to be proven prospectively that they are helpful, that they are safe.

Earl: So, the FDA, uh, does a great job in some areas and kind of has some blind spots, I think, in others. 

Deep: Yeah, I mean, it always struck me as odd that the foundation models don't at least put in a disclaimer whenever they're giving medical advice at a minimum, 'cause if you listen to, like, any financial advice, like every- Yeah

Deep: podcast and radio under the sun will do a whole blurb. Yeah ... they don't even really bother, which is-- I've always found to be really curious. And, and- Yeah ... it's so easy to, to break them out of whatever security context they are locked in. You know, usually just saying that you're working on a movie script will get you whatever you need from the bot- Mm-hmm

Deep: uh, or something along those lines. So going back to this idea of I, I still have this mental model of trying to understand your practice in the context- Mm-hmm ... of maybe efficiency and tooling that could be used society-wide to bring a lot more of the capabilities that you're bringing to these wealthier patients, but that could be brought to maybe society at large.

Deep: And so part of what I'm thinking is maybe we could go through an exercise of breaking down some of the functionality that you guys are offering in this concierge service, and think about how those could be, like, facilitated and automated or pseudo-automated so they could be expanded into other kind of maybe more traditional practices.

Deep: So for example, a lot of the informational facilitation that you're doing, it seems to me like you could have a bot maybe more carefully trained in the way that you communicate with your patients that is under kind of complete, call it physician supervision, where, maybe it helps arm patients with questions that they can, you know, ask to their physician.

Deep: Or basically it tries to like-- or maybe it's working through curated materials and heavily curated materials so that there's no, you know, mixing of whatever garbage made it onto WebMD or something. Like, it's, you know, not pulling things off of discussion forums or whatever, but it's pulling things out of, you know, standard of care guidance.

Deep: But I'm curious do you think, there, there are systems coming out like that, but are-- what's your sort of take on that? Do you think, A, that that could be beneficial for those who can't afford a service like yours, and maybe what do you think are some of the pitfalls there? 

Earl: One of the simplest things that people can do is just, uh, list their medications and, you know, with the number of milligrams and how many times a day that they take it.

Earl: And again, add in the supplements because frequently, uh, a couple of specialists, uh, might e- each prescribe in their area of expertise and, and, uh, the notes may not get to the-- back to the primary care doctor and, um, some medicine that conflicts with each other or adds to each other or if you're taking, uh, an SSRI, which, uh, an antidepressant medicine and one supplement in particular, St.

Earl: John's wort they both work in the same way, and they can increase serotonin, uh, in your brain, which a little bit of increased serotonin is good and can lift depression. But too much, it just, uh... It's called serotonin storm, and it, it can be-- give you a tremendous headache and really, um, at, at the highest level, could even cause a seizure.

Earl: You know, there's, uh... That would be a very simple thing. A little more complicated that we're doing, you know, and our cur-- our patients differ a lot, are, elderly and aren't very tech-savvy. But the ones that are more middle-aged or 60-ish, whatever they're-- with their iWatches and their wearables, we're helping them download the data and then, uh, analyze, looking for any trends.

Earl: I mean, a classic trend is if you don't get a good night's sleep, you'll tend to ingest more calories just to get through the day. Um, you'll tend not to exercise as much. So that's an easy correlation to pull out. But, um, you know, there's a lot, Uh, exploring the data can just, uh, uh, give you a lot of insights.

Earl: And, you know, it's amazing today that it's almost like, uh, what we would get for an ICU patient, that we would probably get 12 different parameters of, of health just from, uh, iWatch and, and plus you add in wearable 

Deep: So maybe tell me a little bit more about that. Like, what are the... Do, are you at the point where you're actually recommending that patients wear some of these devices?

Deep: And what's your guys' methodology for, like, collecting the data and analyzing it? And- Yeah ... what kinds of insights do you think you get there that you can't just get via, asking questions? 

Earl: Yeah. W- we certainly d- we're not heavily pushing the wearables, but, but anyone that asks us the question, we, we encourage them to, to go ahead because uh, a lot of the technology's becoming very good.

Earl: It can be a bit clunky, to download the data because you gotta download it into kind of a spreadsheet format and then analyze it from there. But, there's companies that are what they call data aggregators that are coming along, and they don't necessarily take all of the data from multiple sources.

Earl: But, they're getting better and better very fast and really automating that, data downloading and analysis overall 

Deep: But what's your take at the end of the day? Like, do you really need a sleep analysis device and its data, or can you just ask, "How's your sleep?" And then let them talk? I mean, what, where, where do you see the advantage of the devices?

Earl: I, I, I see the advantage, uh, because people can kind of get used to being tired, and they can convince themselves that they're short sleepers because, you know, they're getting five, six hours of sleep and they're, maybe hard-driving, career-oriented people or just, Type A in general. And they convince themselves, uh, that they're, they don't need much sleep, where actually they've just kind of gotten used to being perennially tired.

Earl: And you can take a look, uh, especially at the sleep stages, which I think are more important than the length of sleep. And people need deep sleep, uh, especially called a glymphatic system, where it clears, uh, some of the proteins. Essentially, what happens is the, cerebral spinal fluid, which is the fluid that bathes the brain starts to, pump at a higher rate and kind of flushes the brain when you're in deep sleep.

Earl: And there's some good studies to indicate that decreases the lifetime risk of dementia. So, yeah, I th- I think it is handy, and I don't think that, self-perception is perfect. And I don't wanna go on too long, but, uh, let me give you just another quick one, is that, you know, people naturally assume if they can see well, nothing's blurry, uh, that their eyes are okay.

Earl: Where, when in fact, they could have glaucoma that... which is a pressure increase in the eyeball, or they can have some early retinal changes, which is kind of the movie screen in the back. And so the body doesn't necessarily monitor itself perfectly. I, I have a saying that living is like flying a plane with, uh, imperfect gauges or faulty gauges that, you know, sometimes it's accurate, you feel great.

Earl: That could be true, but there are some quiet things like a small cancer you wouldn't notice 

Deep: So you've mentioned kind of devices. Yeah. We've talked a little bit about maybe taking your priors and, using those in context with maybe LLM conversations. Like, what are the other kinds of things that you find that you recommend to your patients that are maybe some of us who are listening aren't doing that you think are advantageous from a health standpoint?

Earl: I think, you know, there, there's was a study long ago called the H- Hawthorne Factory, where they went in and, uh, they increased the, the lighting, and they found that the factory was more productive. And so then, uh, they went back, and they're like, "I just wanna prove this." So they, they decreased the, uh, lighting level, and it was a little hard to see inside.

Earl: But once again, it was more productive, so that didn't make any sense. And then they came to realize that just doing the study watching people made them more productive, at least for the short run. And I think just collecting the data sometimes psychologically makes you more attuned to your body.

Earl: Uh, you know, we tend to be a little competitive. If you weigh yourself every day you'll be a little bit more, uh, diet conscious or, you know, or portion cons- conscious than you would be, be otherwise 

Deep: Yeah If 

Earl: that's a, if that's a fair answer. 

Deep: I can see that. Another thing I was kind of thinking along this idea of maybe epistemic trust.

Deep: So patients have this sense of like over weighing the outputs of AI. Sure. I'm wondering, A, if you see that, and then how do you sort of actively maybe counter that false authority? I think you were sort of alluding into it earlier, but better to open it up again. 

Earl: But, what I don't like is when I, uh, patients send me uh, output directly.

Earl: Like, uh, I put this into AI and, uh, here it is. And, you know, you can overwhelm a physician with that. But, uh, but if they come in, you know, and they've spent some time and sort of can summarize their concerns, as a physician first, you gotta be willing to say, "I don't know." And, and that, that's important.

Earl: You also have to be willing to accept that a lot of the time AI is correct and that, but you need to double-check it. And it's, it's fairly easy, especially for a doctor, uh, to, to double-check AI recommendations just, just by, uh, pulling the studies and, and seeing, uh, if the studies indeed support what AI is showing.

Earl: But I, I think it's now, it's gone from like a, a two-person dance, the doctor and the patient, to three, where it's the doctor, the patient, and AI. And the thing that we're trying to avoid against is, uh, it just being AI and the patient because, there's a, a certain level, you know, where, airplanes can fly themselves and they do most of the trip, but it's certainly good to have the experienced pilot be there so, you know, a, a computer glitch doesn't crash the plane.

Earl: And that's, it's kind of the same role physicians are, are starting to assume that we're maybe a little bit less hands-on with every, uh, data point, but we're, uh, watching 

Deep: I mean, I think that's part of why I'm, you know, I think your model is so good, is that you actually build a long-term relationship with the patient.

Earl: Yeah. 

Deep: Where if you look for example, back during the COVID years, there's so much misinformation and- Yeah ... you know, our information ecosystem is highly corrupted at this point, so people are suspect listening to garbage they see on TikTok to, you know- Yeah ... bad talk shows, cable news, all kinds of garbage.

Deep: And but what's frightening is the level of trust in institutional authorities has dropped a lot. So, you know, I was chatting with my sister a lot, and she had a lot of anti-vaxxer, you know, folks, and she would finally have to come all the way down to like, "I'm your doctor. I've been your doctor for 15 years.

Deep: Listen to me or don't, but that's your call. But either way, this is what you need to do." And then even the most hesitant ones that anyone that had a long relationship with her, they would usually, like, come on board. Um- Mm-hmm ... but if you don't have that relationship, you know, and you're just walking in blind and talking to somebody, there's just so much suspicion of institutional and intellectual authority at this point that I think a lot of people are just suffering from a health standpoint because of that.

Earl: I totally agree. it's, when you're coming into... What came back to my mind, early in my career, I used to, uh, moonlighting would cover the hotels. Uh, and if there'd be an emergency or urgency, nights and weekends and, the first thing, like, who's a hotel doctor? You know, like yeah, and they've never seen me before.

Earl: And to immediately establish that rapport and that authority without going in like, you know, you're a military officer and you're barking orders it's kind of, you know, medicine is kind of an art. Uh, and in order to maybe get a scientific point across you have to have the personality of listening to somebody, of connecting with them of, you know, uh, having eye-to-eye contact.

Earl: Uh, you know, all of those things are what AI can't do. 

Deep: Yeah, and I think that's kind of like a... it's an interesting theme of this conversation, at least what I'm kind of yanking out- Yeah ... is that we've got a lot of this kind of increasingly compartmentalized healthcare activity going on, right?

Earl: Mm-hmm. Mm-hmm. 

Deep: Um, where outside of your healthcare context, but in a normal scenario, like you got your 15 minutes with your general, physician, your family practitioner. They send you to a PT, then you might go to a specialist. The specialist might have a relationship with you.

Deep: they may or may not tell the, prescribing physician. Everything kind of... And then you might move, you might get different health insurance. You start bopping around. And so it does seem like the bulk of these decisions just end up falling on the patient, and you combine that with a general lack of...

Deep: there's been a, I think a difference between 30 years ago, my doctor just said what I needed to do. Now my doctor, she does, but, like, I think it's very common for other doctors not to just say what you need to do. 'Cause the question I always use is if, "If this was your kid, what would you do?"

Deep: And they won't answer the question the first nine times you ask it, but eventually they'll answer the question. But I feel like there's something really broken in in the medical educational establishment or in the lawsuit fear climate where physicians just feel like they can't tell you what needs to be done.

Deep: And they just wanna present patients who oftentimes I don't even know are maybe even capable of understanding all of the nuance and complexity to even make the right decision. 

Earl: Yeah. And, you know, this is, again, uh, we're knowing your patient because there are patients that, I just have to tell them what to do.

Earl: And there are other patients that if I did that, they wouldn't accept it. That can only come after the end of a long conversation o- after they've asked good questions. You know, there isn't such a thing as a generic patient. It, it's a pretty wide-ranging spectrum. And then, certainly there's different levels of physician personality, physicians that are comfortable with risk.

Earl: I mean, to practice medicine is to, you know, assume some risk because, you know, invariably you're gonna be wrong at some point in your career. And you can also be sued when you're perfectly right, and, things either by chance didn't turn out or it was, something was misperceived and, you know.

Earl: It's, i- it's kind of a, a messy world. But, you know, back in the day before all of the internet and the AI a- and everything, people used to ask around, "Who's a good doctor?" And they would make their judgments partially on personality and partially on outcomes and, partially by what was heard in church.

Earl: And the whole system wa- wasn't perfect, but kinda worked. 

Deep: Yeah. 

Earl: And, and now we have this, this, information on steroids, and we're trying to find a new equilibrium. 

Deep: Yeah, I mean, I think that's a really good way of putting it. Kinda changing topics slightly. Mm-hmm. I imagine with, um, your patient population, you have a number of folks maybe trying to just not only just s- stay healthy, but, like, maybe optimize their health.

Deep: So when we sort of walk into the wellness arena or- Yeah ... some of these more expensive tests where you maybe take some blood or cut some hair or whatever and send it in. Um- Yeah ... are you are there any lessons for the rest of us that you're seeing from all of that pay a lot, get some fancy information, like, that actually matters?

Deep: Or what are some lessons that maybe from the wellness arena that you think matter? 

Earl: I think some of the basics are important. O- one is, uh, DNA. I and this is coming back to what everyone can do is, maybe not everyone, but a lot of people can a- afford the less than $1,000 to get their, their DNA tested because, uh, that's, uh, one time for a lifetime and if you have control or, or you keep that data there, there is some valuable information.

Earl: You may be, at higher risk for colon cancer and then you should just simply get colonoscopies more often or, you can have lungs that are very sensitive to, to smoke and you, you shouldn't be a firefighter or you shouldn't, be around tobacco smoke in any way. So that can add some real value.

Earl: The whole body MRIs, uh, is one thing that it's less than 10%, probably less than 5%, but w- we have found some, uh, kind of eureka diagnoses where, uh, they have a, a tumor they couldn't feel that was much more treatable now than it would've been by the time they felt it. One in particular, uh, the back of the tongue, uh, and he was successfully cured with treatment that he wouldn't have been had we waited.

Earl: Another one, uh, had a benign growth pinching on a nerve which is, you know, very rare or hard to diagnose and actually needed neurosurgery to remove it but with a good outcome. Uh, again, if it, it took forever to diagnose, it would, uh, have been much more difficult to manage. So I'm a big fan of that.

Earl: What I'm not a big fan of is the ones where the testing is linked to the supplement sales. And so you know, you basically test for enough things, you're gonna find something out of the normal range, and then you pair that with, "Well, we can fix th- those numbers with this pill," or, uh, especially the ones that contain, you know, 14 ingredients so you have no idea or proprietary ingredients where like no one has any idea of what's in it.

Earl: You know, some of the high tech, like biological age I think is still a bit, Bit speculative, which, which speculative, I'll share this with you and your audiences. It took me a little, uh, bit to get up my nerve to do it, but there was a, a, like a death clock where you put in, your age and just a f- you know, like a dozen questions on your health, and then it would come back with the day you're gonna die, which, you know, is like tarot card reading.

Earl: Uh, but anyway, so I did it, and it came back with, I die in August, at age 115. And it said that I was healthy enough that I would reach kind of the AI boost that, medicine and diagnosis, is gonna be so positively affected by AI that in 10 years, if you make it that far, you'll, you'll get a real turbocharge or afterburner to your health for that.

Earl: And, uh, you know, I vowed never to do that again because I'm very happy with, uh, 115. But so some of it's smoke and mirrors, and some of it's good. And, uh, you know, DNA whole body MRI, I like. There's no radiation. Um, but, you know, it comes back to the pillars of health. Uh, it's sleep, number one diet, exercise.

Earl: But just as important- That's probably- ... social interaction and purpose. Uh, so those are the five that I speak to my patients about 

Deep: Yeah. Okay. That makes, all makes sense. Um, so maybe one question for you. When you sort of look around your tool set, you know, of what you have as a physician, what you see maybe out there, w- if you had to pick, like, maybe one thing that's lacking, that doesn't exist, that maybe you think some, uh, machine learning AI systems could kind of bring to the table, like, what kind of tooling do you think could really maybe alter your ability to improve patient lives?

Earl: I think if you could get better reading on, uh, brain health, uh, because, you know, too many people, you know, with a joke, uh, I'll kill some brain cells and, you know, like a wild party where for, you know, a bachelor party or whatever. And, you know, I'm not a moralist. Go have fun. But i- if we can really have

Earl: You know, the brain is a whole ballgame, and if we had some way to really assess brain health and to kind of attribute what's causing the harm and, uh, you know, whether it's sleep or smoking or alcohol or, or what it is for you personally that is putting you at a higher risk of dementia later on and then, then jump on that, uh, I think that would be miraculous 

Deep: Okay.

Deep: With that maybe one final question. Okay. If we fast-forward out 10 years out into the future what do you think the healthcare setting looks like for maybe for you and just, you know, in general? Like, with particularly with an angle on all of the new technology advances. 

Earl: I think that what we talked about earlier the iWatch and the various wearables you will routinely download them, uh, going to your ph- physician and that data will then be analyzed and, and presented to your doctor just like, uh, the nurse taking your blood pressure pulse, et cetera, i- in the office temperature.

Earl: Uh, I think that's one of the things. Imaging's gonna become radically better. Treatments, the new drugs that are basically, a, you know, AI-assisted in drug discovery are now entering, some of the early clinical trials, and they're way outperforming, say, in pancreatic cancer where we haven't had a good answer.

Earl: Some of the new drugs are looking very good, not a cure, but, but at least a life extension with pancreatic cancer. So there's a lot, I think a lot of positives, in healthcare. 

Deep: All right. Well, thank you so much for coming on the show. This has been really valuable.