Don't Upload Your Medical Record to ChatGPT — Here's Why | Dr. Terry Adirim

Only 12% of people are fully health literate, a statistic Dr. Terry Adirim calls "astounding." As patients increasingly turn to ChatGPT and Claude before their appointments, she has one clear rule: never upload your full medical record into a general LLM.

In this episode, Chris Hutchins sits down with Dr. Adirim, pediatric emergency physician, former Acting Assistant Secretary of Defense for Health Affairs, and author of the upcoming book "The Prepared Patient" to unpack what happens when AI enters the exam room. They cover the accountability question when AI shapes a diagnosis or flags a coverage denial, why she believes AI should be designed with clinicians and not around them, the informed consent gap around ambient AI scribes, and why patients researching symptoms on AI should treat it as the start of a conversation with their doctor not the final word.

In this episode:

• Why only 12% of patients are fully health literate, and what that costs them
• The one thing you should never upload into a general AI chatbot
• Who's accountable when AI shapes a clinical decision
• What informed consent should actually look like with ambient AI scribes
• Why the most successful health startups have physicians as founders

Chapters

 00:00 – Cold Open: Three Seats at the Table
 01:14 – Dr. Adirim's Path: From Pediatric ER to the Pentagon
 02:34 – The Gap Between AI Policy and What's Happening on the Floor
 04:55 – What Individuals, States, and Industry Each Need to Do
 08:45 – Introducing "The Prepared Patient"
 11:20 – The Interoperability Problem Nobody's Fixed
 17:20 – Layering AI Above the Electronic Health Record
 21:21 – Patients Arriving With Their Own AI Research
 24:46 – Who's Accountable When AI Shapes a Diagnosis
 26:16 – Designing AI With Physicians, Not Around Them
 34:56 – Why Board Exams Still Ban AI Use
 38:34 – Should Medical Students Use AI Early in Training?
 39:34 – Informed Consent and Ambient AI Scribes
 43:13 – Getting Test Results Faster — At What Cost?
 47:07 – The One Thing to Never Upload Into AI
 49:40 – Where to Find Dr. Adirim and Pre-Order the Book

Transcript

Chris Hutchins 0:00

My guest today has seen healthcare from three places. Most people never get to stand in. The exam room, the Pentagon, and the Federal Policy Table.

SPEAKER_01 0:07

What you see or what you get from ChatGPT, Claude, or any of these other large language models, LLMs, is not the definitive answer. It may not be accurate. They don't have the context. They don't know you. You may not give all the right information. When you sit with your doctor and you come with that information, it's the start of the conversation. It's not the I found on Chat GPT. Only 12% of people are fully health literate. We don't know where that information is going. It has your identifiers. A physician has the relationship with the patient. If you're using AI and it tells you something or it does something autonomously for the patient, you are still that practice, that physician is still responsible. Do not upload your medical record in total up into a general LLM. Just don't do it.

Chris Hutchins 1:07

Welcome back to the Signal Room. I'm Chris Hutchins. My guest today has seen healthcare from three places. Most people never get to stand in at once: the exam room, the Pentagon, and the Federal Policy Table. Dr. Terry Angle, welcome to the Sigma Room. I'm very excited to have you on this morning. It's just been an extraordinary few days, and I really enjoyed our conversation last week and am very happy to let some folks in on some of the amazing work that you're doing. You spent more than 25 years as a pediatric emergency physician. You've led health IT overhauls at the VA. You served as acting assistant secretary of defense for health affairs. This is just amazing. Before we get into too much, I'd like to hear a little bit about your personal story and a bit about your why. I know that people don't just stumble into a kind of career that you've had not only as a physician, but involved in so many different aspects of policy making and education and writing. Definitely want to make sure we take a chance to take a minute to show to really get into your new book. I'm very excited to hear about this. But if you could just start us off with a little bit of background and what makes you tick.

SPEAKER_01 2:15

Yeah. Well, thank you very much. And thank you very much, Chris, for inviting me on uh for uh this discussion. So, what makes me tick? I've always wanted to be a physician. I've always wanted to be a doctor, I've always been curious about health and healthcare. And most especially, like all my colleagues, we very much wanted to help people. And as I was going through the process of becoming a physician through education and my years of training, I realized that I could make a difference by treating patients one-on-one, or I could treat larger populations by moving into academia and then into the federal policy sphere. So that has been my career trajectory. I've been very fortunate to work with a good number of great colleagues across the spectrum from clinical care, academia, and federal, I would say, uh health agencies that I've worked in.

Chris Hutchins 3:24

So amazing. So I want to kind of get into some of the interesting stuff for for folks, because right now, when we're almost every conversation that comes up about AI, there's that there's some policy and uh regulatory kind of aspects that seem to be coming up, governance a bit too. Um you've sat where some of these big decisions actually get made, like in the VA, the Department of Defense, federal health policy. Where's the widest gap right now from your perspective between what leadership believes is happening with AI in the critical encounter and what's actually true out on the floor?

SPEAKER_01 4:02

Sure. Uh, thanks for that question. I think right now there's a tension between how much regulation, how much uh policy making to ensure that AI used in healthcare is safe with innovation. There is that tension. And prior to this current administration, there was an appetite for coming up with frameworks on ethics and regulation and governance. But when this new administration uh came in, they wanted to unleash AI and to ensure that innovators can innovate. I think it has to be somewhere in between. I think in a current environment, we're not quite getting it right because I think the healthcare field, uh, healthcare community, as well as patients, need that guidance. Um, and it's just not happening really right now. And hopefully, because we're in this transitional period, we'll get to a place where we strike that right balance. Because you don't want to hinder innovation. But in healthcare, it's very different than other sectors in that we need to make sure that we protect patients. So that's where we are right now on a national scale.

Chris Hutchins 5:31

Can we I want to double-click on that just for a minute, if I could. Could you uh speak about uh some of the specific things that you know people should be aware of uh, you know, from your your perspectives? Like one of the reasons I love having this platform is for when for someone like you who comes on, you you can actually send some really uh clear messages out there so that people that may not be paying attention have some idea what they should be doing and how to lean in and maybe hope push push the influence uh in the in the right direction.

SPEAKER_01 6:02

Well, I think there's uh kind of could look at it in two ways. There is what individuals need to do, and then there is actually three. There's what we need to be doing nationally with regard to the federal environment. Um, there is stuff being done on the state level, and then there's what the industry should be looking out for. I think there with technology in of itself is not ethical or unethical, but how we use it brings up the issues of ethics. I think privacy and healthcare is huge. And I think there are a number of things that that, for example, an individual can do to protect their privacy when AI and AI tools are being used, like one of them being don't upload your entire medical record into Chat GPT. Uh, with regard to you know the federal sphere, I really think they need to update a lot of the regulations and policies around patient privacy that brings in AI. Some of them, that these policies and regulations didn't even think about AI. It came before, you know, AI became a thing and healthcare. So I think we that that could be attended to. And then lastly, I think the industry too needs to respect um healthcare and needs to understand it differently than some of the other sectors. So let me give you an example. When you people who are, you know, for lack of better term, tech pros, they like to move fast and break things. In healthcare, the culture is very much about do no harm, don't hurt patients. So there could be conflicts between those two ethos, which um I think it's important to reconcile. And I think um health tech companies have a responsibility to understand the do no harm and to respect that and understand that those of us who want to use these tools very much are looking forward to all these innovations want to make sure that there's that respect. And I think that tension will be there for a while longer, but I think that's something that could be paid attention to.

Chris Hutchins 8:27

That's amazing. And I really appreciate you bringing some clarity to this stuff. It's just such an unbelievably fast-moving trend transformation that's going on. To your point, the innovation's definitely outpacing uh the responsible and ethical use and a lot of the policy and guardrail things that really need to be put in place to protect patients. Well, like you said, I mean first do no harp. We we didn't all go to medical school and take that oath, but we need to support that oath and make sure that we're backing it up by what we deliver. No question about that. We have a lot of opportunities to do better. This kind of nicely steps over into the topic of your book, which I'm I'm really excited to hear about, but the prepared patient. Talk to me about that. So, what does that phrase mean to you? And and you know, how did you how did you come up with this particular title based on all your experiences in what's happening now? And I suspect it's not about an AI but an unveiling type of thing and trying to prepare patients for that. But I mean, I'll leave it there and just I just want to hear all about it.

SPEAKER_01 9:31

Yeah, no, thank you. Um, it's a broader uh topic. And and the book is called The Prepared Patient, Your Guide to Surviving the Healthcare System. And I was inspired by two things. One was the statistic that only 12% of people are fully health literate. I'll let that sink in for a second. That is a really astounding statistic. And what does that mean? That means that if you don't know fully about your health coverage, what to expect, and how to navigate the system, that could have impacts on your health outcomes and on your finances. So that's number one. But what really inspired me was early in my career, I took care of a lot of families with children who have special health care needs, medically complex children. And what impressed me about them is that they came to the emergency department, which is where I spent my clinical part of my career. They always came with these white binders that had all of their information. And as a physician who didn't necessarily know this patient fully appreciated having all their medications with their dosages, their medical diagnoses, who their specialists were, and just all the information that I needed to care for that child. Now, we, of course, scroll ahead 25 years later. We don't necessarily have a white biter, though you can put it in paper, you could do it electronically. But um the point I'm trying to make is that they wanted the best outcomes for their children so they were fully prepared. So that's the inspiration for this book. Our health system, no matter how smart you are, no matter how educated you are, could be challenging to navigate and difficult to understand.

Chris Hutchins 11:22

That challenge is this is strangely enough. I actually had had someone someone on probably six or eight weeks ago. I'm not really sure timing-wise, but she she she had a child who had really, really rare disease. And she talked about this exact thing that you just mentioned where she's constantly having to keep everything updated and you know how difficult it is that every time you've got to you want to go see a new specialist, it's Groundhog Day. And if I mean from your perspective, um, isn't are there some things that we should be challenging our our technology companies to do? Because I mean that's a pretty big part of what the care.

SPEAKER_01 12:07

Right. And I 25 years ago, I would have kind of hoped that we would have been further along, you know, in 2026. Um, but there are several factors that have been challenges to make it more of a smooth, super well connected healthcare system. I mean, even just a few years back, I was working in the emergency department and I got a patient, a young patient in family, was transferred from another hospital. So as I'm asking the father questions, he's like, Yeah, we just we just answered all these questions and we had all these tests done. Isn't it in your and he pointed to the electronic health record. And I said, Yeah, I'm really sorry, but you came from a different healthcare system, right? So within the same healthcare system, if you're taking care of at Kaiser or Mayo or Cleveland Clinic, like, sure, your records would be in there. You go see a subspecialist who's affiliated with that system, that information would be there. So that would be step one. But that's not how things operate. Just like your guest on, you know, several weeks ago, she may need to see different specialists in different healthcare systems. And one of the fixes is not really been a full fix, has been these health information exchanges. But even that, those are kind of clunky. You have to pull the information, doesn't automatically populate, and you know, people have to agree to be a part of that. So we still have a very disjointed, fragmented health care system, which makes it so challenging, especially for the over 50% of people in this country who have chronic medical conditions. And for that child who's got multiple conditions, they're seeing multiple specialists, they're getting multiple tests. Unfortunately, and it's not fair and it's not necessarily right, but the responsibility then becomes on the patient or that patient's family. So that's what I was trying to do with this book is to help people get themselves organized, get themselves prepared, figure out what types of information that you need, for example, to bring to your appointment, what types of information you need with regard to your healthcare coverage. Now, that it's all not all doom and gloom, right? AI is helping to smooth some of these friction points for patients. So, for example, you can collect all this information electronically, right? You can do that. Um, and there are companies that are working on doing this for people who've got complex medical needs. There's also within the exam room things that could help the physician, right? So, for example, if a patient's medical record is within that system, the in a lot of places the physician could get a summary that's, you know, that is created by AI, right? So now there's a number of institutions. I would, I don't think it's most, but there's a lot of institutions and doctors' offices that are using something called AI ambient scribes, meaning that the physician isn't sitting there typing away while they're talking to you. A voice-activated system turns on while while the patient and physician are talking and takes notes, organizes the notes, and the physician can ask for a summary of that. You could even do a patient summary and give that to the patient about their visit. So there are these things that are not solving the problem, but they're really reducing those friction points that you brought up for people who have to see multiple specialists.

Chris Hutchins 15:59

Yeah, I I think um to people out there who can actually develop and design stuff, we're still talking about the same lack of interoperability here. And I don't know how many times I've heard this over a decade now. And we're still we're still having solved for it. And then there's just no good reasons for it.

SPEAKER_01 16:18

Yeah, it's been longer than a decade.

Chris Hutchins 16:20

Yeah.

SPEAKER_01 16:20

You're being nice by saying a decade. It's, you know, it's been fraught and difficult because, you know, there's a number of laws around privacy. The way that EHRs were established, companies establish their electronic health records and they're proprietary. So they don't want to share and connect. And so that took a long time to improve that problem and then also develop health information exchanges. They're not perfect solutions, but you know, it's it's a lot of regulatory issues and proprietary issues, and how electronic health records were established that have made it not easy for physicians or patients.

Chris Hutchins 17:07

Right. Well, I mean, that the unfortunately it was it was designed to do more accurate billing instead of actually supporting clinical workflow. And that's and if quite honestly, it hasn't done a great job with either. That's the part that's unfortunate.

SPEAKER_01 17:21

Yeah, we have such a complex health system. But what the the future's gonna hold, and where I think um we're gonna see improvements is that the electronic health record is not gonna go away. It's gonna be that system of record. But hopefully, what we're gonna see are innovators, and we're starting to see this, build and design products that are layered above the EHR, can extract stuff from the EHR, but then operate kind of above it. So that from the very start, let's say it's a primary care practice, you have AI tools that help with appointments. There'll be chat box, patients don't have to wait, and they could do all that kind of upfront work automatically, all the way to ambient AI scribes in the exam room, um, to patient discharge instructions. You already have electronic prescribing, but you know, really amplify that or I'd say augment that. There's other tools on the back end too that also help patients with referrals and so on. And there's the companies, there are companies that are doing uh what's called RCM, but that billing kind of work. So one day I predict we're not even going to need to necessarily touch that electronic health record. Maybe just the people within the companies and big health systems might need to touch it. But hopefully we'll get to a place where we have systems in place that make it with less friction for patients and physicians.

Chris Hutchins 18:57

That should be the objective right there. I mean, I get where the litmus tests that I've I like to talk about is first and foremost, before we talk about any other aspect, I just want to know is it going to give time back to the people who need that? Whether it's you know the physician, the nurse, the MA, uh, more most importantly, you know, the the time for that encounter between a physician and their patient. Uh that's gotta be the bar that we're shooting for. We can I just think that we do a lot of stuff because we think it's cool technology. Unfortunately, that's happened a lot in healthcare. And that's just not been helpful. It's been disruptive from everyone I've talked to, including yourself. There's there's definitely some some areas that have been a little you know just glaring and then they haven't been touched. I mean, interoperability is not a small thing. Honestly, I don't think it's the most complicated thing in the world, but there's got to be some tinks to the regulations to force that to finally happen.

SPEAKER_01 19:53

Yeah, it's it's more complicated than you'd imagine. But to your point, I think AI is going to be a help. I think on balance, it's a positive thing. And I think the future things will be better. It will not solve the problems of our healthcare system because our healthcare system is built on policies, regulations, laws, and so on that make it this multi-layer. And now I'm gonna use a big SAT word, accreted. You know, everything is just all kind of glommed together, hard to extricate all the different pieces. So AI is not really going to fix that. We're gonna have to have the political will to want to make the changes that we need in order to unfragment, unopaqueify, I don't know, whatever terms you want to use, uh, our healthcare system. But I really believe that technology can improve the experience for both the you know clinicians and for um patients.

Chris Hutchins 20:54

I I I agree with you there. You let's talk a little bit about what you know the dynamics there that are actually changing now, where uh a patient may show up already having worked through their own differential before they even sat down. What are you seeing in in practice? I mean, I mean in this case, I'm assuming it's you know the the the parents who would be coming in having done their own AI we'll call it research, but yeah.

SPEAKER_01 21:19

Well, you know, uh yeah, pediatrics, uh, pediatric emergency medicine, children who are medically complex and with special health care needs, their parents know their condition. And if it's a rare condition, they know more than you do. Because in the emergency department, you're a bit of a generalist. So they've done their research. And I've you know started practice when people were beginning to go on the internet and check, you know, Google their conditions and stuff. So it's been around for a while. What I would say is different. There's several differences. One of the differences is that back then people would Google a lot of it was wrong, people didn't know how to sort through. That kind of stuff. And there was a lot of resistance by the medical community, right? And but I think scroll ahead to now. I think AI could actually be like a patient co-pilot or you know, really help patients. And I think physicians, you know, uh many of them, not all of them, are less resistant, see it as an opportunity, right? Because I I really like when my patients know know things. I, you know, I want them to be engaged in their um health and their health care. But I think there's a responsibility on the patients too, right? What you see or what you get from Chat GPT, Claude or any of these other um large language models, LLMs, um, may is not the definitive answer. It may not be accurate. They don't have the context, they don't know you. You may not give it all the right information. When you sit with your doctor and you come with that information, it's the start of the conversation. It's not the I found on Chat GPT that I have, you know, I must have lung cancer, right? No, that's not that's not how that should work. It should be the beginning of that conversation. And it takes it to a different level, right? As a physician, you're not starting from scratch. You're like, oh, well, so tell me a little bit about what you found. Well, here, let me tell you why that's not quite right. I did this test and it showed this, it showed that. So I think it could be a good thing as long as both physicians kind of know, you know, how to work with it, and patients also know how to work with it. So I think overall equals the playing field a bit, levels the playing field, which I think is good because in medicine we've traditionally been a very paternalistic um field. But I I I like the fact that it can help empower patients and prepare them for their care. So I see it on balance as a good thing.

Chris Hutchins 24:12

You've written on AI and accountability and care. When AI shapes a diagnostic recommendation or flags a coverage denial, who actually owns that outcome, in your opinion?

SPEAKER_01 24:22

Yeah, okay, that's a complex issue. If you're using it in your practice, you as the clinician, the practice is responsible for the output of that AI because you, as the clinician or provider, are the individuals or organization that is delivering the care, right? The LLM, the GPTs or whatever they are, or the AI are tools in your practice. So you have to look at them as tools in your practice. They are not delivering that care. So it's who is responsible for the delivery of the care. A physician has the relationship with the patient. If you're using AI and it tells you something or it does something autonomously for the patient, you are still that practice, that physician is still responsible for the care. So always the physician or the practice.

Chris Hutchins 25:21

You also talk uh about the the approach to designing AI. You know, essentially, you you're I think that you you're agreeing with probably almost any any physician I've had a chance to talk to in the last several months. Almost everyone's telling me the same thing. You said we need to design AI with care providers, not around them. Where have you watched that step get skipped and what broke as a result of it?

SPEAKER_01 25:46

Yeah, wow, that's a great question. I've talked to a lot of well-meaning founders of startups. And, you know, not having somebody that's either a co-founder with you or within that leadership structure who is a clinician, and I believe it should be a physician, means that you may identify a problem, but there's a good chance you have not found that solution or you're not creating the solution. Physicians know how care is delivered. We know our patients, we know what works, right? You could be smart, but a business degree or a computer, or I should say, software engineering degree does not make you an expert in healthcare. So that's foundational, right, from the start. And I have talked to founders who identify the problem and I'm like, that's great. I know you want to solve this, but then they come up with a solution that doesn't, which probably wouldn't fit, right? The other thing too is physicians have to use these tools, right? So not having them in the loop is probably not that, probably not the wisest uh approach. So and the reason why I think this happens, that physicians are not uh brought in early in in subcases. I think we may be seen because we have the ethos of do no harm. And people, you know, startups have the ethos of work fast, break things. There could be some seen as some conflict between those two. And what I see is that people who are inventors or innovators, I think some of that speed in getting things done could rub off on physicians. And physicians, they're do no harm ethos should rub off on the innovators. So I see that as really important that it's they work together. Because at the end of the day, we do not want software engineeres doing health care. I don't think they want to do healthcare, but that's at the end of the day, that's brilliant.

Chris Hutchins 28:03

It's inter it's interesting. I uh I've seen it go the other direction. I've seen some brilliant uh clinicians who have developed really great solutions for things. And that that's always been like beyond my ability to comprehend how someone can have so many different things going at the same time. And that creative things, it just seems to manifest itself in really interesting technologies. I mean, I had a chance uh a number of years ago to to work with a it was a pediatric neurosurgeon, and he actually developed some software in his garage. Uh I think he probably used like four or five different programming languages, and every once in a while he'd see something that inspired him. He'd he'd go figure out how to use that that language and he put it into his tool. But uh, you know, mind-born.

SPEAKER_01 28:50

I think it's easier to teach physicians how to be in innovators and business than it is to teach, you know, business people health care, right? Because, you know, I spent many years in school, I spent years in training, um, extra training. I had at one time three board certifications. So learning that does not happen quickly. I know a lot of people think, you know, they know healthcare. I think we're seeing we're in a transitional period where we're seeing a lot of physicians who not only have business degrees, I have a business degree, but also are in the innovation field, developing technologies. And it's especially important that it's people who've practiced, right? The neurosurgeon that you cite probably been practicing for a good number of years. And over and over again we see the same problems and we say we need to fix this, right? But there wasn't a day that went by. I was in the ER, like, why can't we fix this particular thing? And I would say, even early in my career, I was like, why are we handwriting notes? Why can't we have electronic records? And and you know, I did try and create something, but it didn't work out. Um But we but that's you know, we want to we want to uh fix the problems that are in healthcare.

Chris Hutchins 30:12

So it it is amazing. I think the first couple of months when I launched my my company, there were, I would say it was almost a third of the people I would meet were that were founders, uh about a third of them were were physicians. And you know, it was that's one of the things that motivated me to to start this platform to give people a voice that that really have something to say about this stuff because you people are inventing things because everything's working for them in their electronic health care. There's there's gas.

SPEAKER_01 30:45

Yeah. Across the board, I think everybody knows that our health system is broken and and has problems. But the one thing is, and I don't have the statistic for it, but I know it's true that the most successful health startups have physicians as founders or co-founders.

Chris Hutchins 31:00

Yeah, that's not that's not surprising to me at all. It's it's a strange approach that's I've seen it in even working in revenue cycle systems years ago, and I'm sure I'm sure you've had been around this too. One of the a vendor that you've you've been working with for years, they want to do something um unique to really take a nice leap forward and give you better capability. So they come in, they conduct a few interviews, uh and sometimes they'll do that as much as a week. Uh, then they go away for a while and spring a new product on you that doesn't even resemble the things that you talked about when they did the interview and they don't understand why you don't want to use it.

SPEAKER_01 31:36

That's why you have to have the people who are using it there. And it's critically important. And I found this out when I was at the VA where we're doing electronic health record deployments. We did have the end users there. But changing how they did their work really can be upsetting. It's disruptive. Um, and so having the clinicians there could help smooth that out. They can help with product design, they can help with how it's being used. And that's critically important because if the end user won't use it or they use workarounds, then your efforts have gone to waste.

Chris Hutchins 32:14

Let's just talk about a couple of different things. No, you uh one of the things that you've mentioned to me before that previously was um some of the policies that are in place now, the rules for maintaining board certification, uh forbidding use of AI to answer exam questions being one of those. Yeah, in real practice, you know, commissions could, you know, they consult outside evidence constantly. What does that tell us when the way we test doctors is out of step with the way they actually practice?

SPEAKER_01 32:44

Yeah, well, first I have to say I am only aware of my board certification, the American Board of Pediatrics, they say you can't use AI to answer the questions. But I think, you know, just like other parts of uh health and healthcare, the whole environment, they're gonna have to catch up. And, you know, the old way of doing things uh was to memorize, right? We you learn a boatload of stuff during medical school, and you're supposed to um memorize all of these things and then you know, memorize the crib cycle, and then you take board exams, and we take several exams. We take, you know, uh boards, you know, part one, two, three, and then we take our specialty boards and our subspecialty boards. And it was the old-fashioned way was memorizing a lot of facts and material. But that doesn't align with how we practice medicine, right? We've had these clinical decision support tools at our fingertips for a while now. Uh, things like up-to-date algorithms from our professional societies and now AI with uh tools like open evidence, and there's others as well that I've used. You know, when you're practicing, you don't have to rely on a memorization necessarily to extract that information. So the testing for keeping up your board certification needs to kind of come up, come up and be a little more aligned with how we really practice medicine. Not that I don't think memorization is important. I still think in in medical school it's important to memorize a lot of these facts so that you have them and have a very good understanding. But when it comes to later in your career, after you've practiced, you know, asking to regurgitate facts is probably not the best way uh to test physicians. So we need to think about ways that we can we can use AI to update how we assess uh physician uh capabilities.

Chris Hutchins 34:49

I'm curious about the the way medical school the medical schools run now. Do you have concerns um that they start to get people using this AI or early on in their training? Uh what I'm kind of getting at there is there's a lot of stuff that you go through. I mean, I know I don't know the half of them, but you go through all the all the the class time, all the you know, all the uh rotations with when you're doing a residency, all those types of things. I I wonder if there's a a need to make sure that we're not I don't want to call say dumbing it down, but I mean I think there's a risk maybe for for inhibiting the development, the reasoning and the judgment that has to be be ingrained in someone when they're being taught how to how to be a physician.

SPEAKER_01 35:40

Yeah, no, and I think um medical educators are aware of that. Um and it is you know a concern that those critical thinking skills and you know I'm hopeful that it'll it may change how we operate and work, but that we would still need those critical thinking skills. Um, and we just need to figure out how we uh promote that and culture cultivate that, but it'll be in different ways because we can't say no to AI when it's being used already pretty extensively. So I think it's gonna be up to medical educators to kind of think about how we educate for the future, right? And I think they're starting to do that.

Chris Hutchins 36:25

We're gonna get into an another area. And then this one's interesting for me because I because you're you're you're um you're also heavily involved in in public policy and you obviously you've worked in in government as well. The concept of informed consent, uh, there's a lot of conversations uh that are happening, especially now because there's been some some lawsuits because of the uh how how things have been implemented with ambient listening. Um how do you think about that? And and what is it that we need to be be doing in order to help a patient to understand what it is they're consenting to? Because I don't think historically we've done a good job of even explaining it that the informed consent, even in the way that it was was prior to you know this this whole ambient listening thing.

SPEAKER_01 37:10

Yeah, um I think there's been people who've criticized um informed consent, but informed consent is critically important, right? Because if you know you're of sound mind, you're an adult, you should be making decisions about your care, right? Right. So there's different levels of consent. There's consent, somebody just to walk in the emergency department that you consent to somebody delivering your care, right? So um, I don't need extra consent for drawing blood, getting x-rays, doing a physical exam. But if you're doing something more invasive, then you need an additional consent. And that always involves laying out what it is that you're going to do, emphasizing that this is a choice, you don't have to do it. But if you don't, here's what's going to happen. And the pros and cons of that particular, whether it's surgery, whether it's a procedure. When it comes to ambient AI scribing, the central issue is privacy, right? And being informed means that the patient understands where that recording is going, where it's stored, and what are the privacy safeguards, uh, how long it is stored, uh, whether or not they have a choice to use AI scribe or not. So those are some of the things that patients need to look out for, but physicians need to really be um intentional. When a patient walks into their room, they need to say, I use something called an ambient AI scribe. It is a recording. We keep it only for a month. It is, you follow all the laws with HIPAA and so on. Do you are you okay with me using this ambient AI scribe? Right? Right. Um, that's informed, right? That took a couple, that took a minute, right? But I really believe that if you're going to use something like that where privacy can become a concern, then I think you know, consent should be had. Now, not everybody believes that way. I've been in rooms where people have said, oh, I wouldn't ask for consent. There was a conference I went to where they were talking about a chatbot that does appointments and all that pre-appointment work. And I asked, Well, do you get consent? Um, do you think you have to get consent? And half the room said no. And I'm like, Well, yes. I mean, yes, you need to tell, you know, you need to at least say at the start, this is a chat bot, and I'm going to be doing, you know, making your appointment. And that could be enough, right? Because it's not such a big deal, right? But with healthcare, there's special regulations and policies around privacy, and we need to respect that.

Chris Hutchins 39:59

Yeah, you're absolutely right. I think one of the challenges that I've been concerned about is how do we make it really easy and consumable so that, you know, as a physician, you you don't have to go get another degree to be able to explain AI. I mean, that I don't I don't mean to be it'll be flip about. I'm just I just think it's something that we have to think about as we're as we're scaling. And it's gonna get there.

SPEAKER_01 40:24

It's gonna get there because we use tools all the time that I couldn't tell you how it works, but we use it. You just need to get to that level of trust. U right now we're in this transitional period, right, where more asking more permission and consent is probably the better practice right now.

Chris Hutchins 40:44

We've talked about uh uh a few of the challenges in the in in and some of the interesting benefits. But one thing that um I started to see this a few years ago when I was working was in New York. Um there was this big push to make sure that we're we're giving people their test results as quickly as we possibly can. The pace now that we can do these things, AI is kind of speeding that up even further. What are your thoughts around that? Because I think one of the things that was was concerning was getting a diagnosis on a Friday that's coming out of a lab result going into the weekend and that and the results are not something that someone's going to be comfortable with. Uh talk a little bit about that from your perspective and what we should be thinking about in trying to make sure that we're syncing up the uh the speed with uh the actual preparation that's required to be responsible with it.

SPEAKER_01 41:36

Yeah, yeah. Sometimes when laws are passed, and this law was passed in the Cures Act, best intentions, but then the unintended consequences are not necessarily thought about, right? So you identified a scenario where it would not be ideal that somebody gets their test results on a Friday night, right? I think overall it's a good thing because people should have access to their test results. Uh, the responsibility is on the physician who ordered the test to make sure that the patient gets the test, but there may be a different way we're gonna have to practice, right? So let's say um a patient gets a test result. You know they're going to Chat GPT or one of those things to find out, you know, what it means, right? Well, how about a physician? If I'm doing a test and I said, Look, I'm gonna be doing a complete blood count. I'm anticipating you might have a low hemoglobin. That's a red, red cell. If it's low, it may mean you have anemia. It may not, it, you know, I'm not thinking that you have cancer or this or that. So you may need to do a little bit of prep preparation when you're doing the tests, so that when the result comes back, and that was too simple, that's a routine test, but let's say you're doing a um a biopsy, right? And you're worried that it's cancer. What you should, I think what best practice would be is to say, I'm doing this test because I'm concerned that you it will have an abnormality, like a, you know, a cancer or mass, and discuss it pre preemptively so that when the patient does access that information, it's not so much of a surprise. And in that preamble, say, but we have we have treatments for that. So I'm happy to talk to you about it if you get that test result. So we may need to change a little bit that, you know, how we practice, you know, adjust a little bit that workflow. But um, I think in general, it was a good thing to do. I think it would have been good to think about it, about what the repercussions and what the consequences uh were going to be beforehand. I think that law was passed like in 2018. So hopefully practices have worked on that. But yeah.

Chris Hutchins 43:56

Interesting that it it comes full circle. Um, back to Where we started your book, uh, the prepared patient. It it's I I love how you just described that that whole scenario because that's really uh what what's needed when you when to reach when you've got these kind of requirements. Uh the you got to think through this. How do you prepare the patient? Uh so I want to make sure we leave people with something that they can they can use here. So if you're talking to a patient right now who wants to bring AI into their own care, what's the one thing that they should always do and one thing they should never do?

SPEAKER_01 44:29

All right. Let's start with never do. I advise everybody do not upload your medical record in total up into a general LLM. Just don't do it, right? Uh it's we don't know where that information is going. It's has your identifiers. Like, don't do that. That would be one of the things I could think of others. I would also just because this is going to get to what you should do. I would not take everything that is said, if you put your information, you'd say, Oh, I have a mole on my arm and it comes back with what it could be and it gives you this long list. Use that as a start of your conversation. Don't automatically assume it's melanoma, right? It's not cancer. Bring that to your physician and say, you know what, I put that into Chat GPT and it gave me this differential. Let your physician talk to you, have a conversation about uh what it's more likely to be based on what, you know, uh her experience and what her she's looking at and all the data in order to come up with that diagnosis. I think people should understand what they get from these LLMs. There's probably a high amount of accuracy. I think we have studies that show that there's a highly accurate, but I think understand that uh general LLMs are just spewing out information, right? Physicians are not information spewers. We do more than that, right? So the LLM does not have your context, your family history, your physical exam, all the other pieces that we put together as that puzzle to come up with a diagnosis or a plan for you. So as I said, it needs to be the start of the conversation, something you bring to your appointment, um, so that you could start at a place that is ahead of having to get, you know, all the basic information. You can actually have a good conversation. I think it's also a good tool to ask it what kind of question should I be asking at my appointment? Knowing that appointments are, you know, for routine appointments could be 15, 20 minutes long, you don't have a lot of time. So you should prioritize what questions you're asking. And that's what I talk about in my book, The Prepared Patient, about being prepared for your appointments, um, would be very helpful uh to you in getting the most out of those appointments.

Chris Hutchins 46:50

And as we if we wrap wrap up our conversation, uh tell tell folks where they can get a hold of you and where they can pre-order your book.

SPEAKER_01 47:00

Yes, no, thank you. Well, my book is available for pre-order on Amazon and the Johns Hopkins University Press, which is the publisher. It's being released on August 11th, so weeks away from it actually being released. But right now you can purchase it on Amazon. Um, and if this is recorded, if this is released after August 11th, then then it's available um on a lot of booksellers. Um and I could be reached. Um, I'm on LinkedIn under my name, um, on at Terry Dramd, um, on X um uh Instagram, and I have a website called www.thepreparedpatient.com, uh, which has all the information about where you could get the book and so on.

Chris Hutchins 47:53

Fantastic. And for listeners, uh, you you'll find all this information that Dr. Terry just shared, and it'll be in the show notes. Uh, we'll make sure you know how to find her. And please make sure that you you take a take a snapshot of the of the screen when you see the the pre-order link, and you you definitely want to get this book. I've I've learned a lot from you this morning, and I really do appreciate you coming on the show. Uh thank you.

SPEAKER_01 48:19

Thank you so much for having this conversation with me. I truly appreciate it.

unknown 48:23

Dr.

Chris Hutchins 48:23

Ms. Terry Adram, physician, former federal health leader, and author of the Prepared Patient Out This August from John Topkins University Press. The thread I am taking with me today is the patient already has a copilot. The real question is whether the system is prepared to go meet them. Go pre-order the book. And Terry, thank you so much for joining me today. And for my audience, I'll see you next time on the Signal Room. That's it for this episode of the Signal Room. If today's conversation sparks something in you, an idea, a challenge, or perspective worth amplifying, I'd love to hear from you. Message me on LinkedIn or visit Sigma RoomPodcast.com to explore being a guest on an upcoming episode.