Can AI Diagnose Better Than Doctors?

Speaker A: I'm not at all worried that AI is going to put us out of business. I am worried that, like all automated answers, it's going to be valued tremendously. And that means that doctors are going to have even less practice about thinking through a diagnostic problem.

Speaker B: Welcome to Transforming Medical Communications, a podcast by Medcom's experts. We share medical communications insights and advice from the best and brightest in the industry to find out what they're doing to push our industry forward. Here's your host, Wesley Portogees.

Speaker C: Welcome to Transforming Medical Communications, the podcast where we explore bold ideas to shape the future of medical affairs. In an era of data overloads, compressed consultation times and AI driven decision tools, one question matters more than ever. How do great doctors still arrive at the right diagnosis? Today's guest has spent decades exploring exactly that. I'm joined by Dr. Lisa Sanders, medical director of Yale's Long Covid Multidisciplinary Care center and the physician behind the Diagnosis column in the New York Times Magazine, the column that inspired the TV series House M.D. in this episode, we'll explore what medical communications can learn from the real cognitive and human process of diagnosis and how the industry could better support how clinicians actually think. Lisa, welcome to the show.

Speaker A: Well, thank you, Wesley. It's so nice to be here. Such an important topic.

Speaker C: Yeah, absolutely. I know you're very passionate about it, so I can't wait to get into it. So actually then, let's start with the core of the conversation. So in this world of guidelines, algorithms and information overload, what would you say actually determines whether a clinician arrives at the right diagnosis?

Speaker A: I think the guidelines and algorithms can be extremely helpful. I mean, they're often, they should be, they're supposed to be based on data that show what is the most effective way to do something. And so I think most of the time following the guidelines or the algorithms is going to get you to the right answer. The question that is interesting to me is what happens when that doesn't work? And one of the problems with having guideline directed medicine is that people don't spend enough time thinking what they're supposed to do if the guideline doesn't work. When I was trained, when many of my peers were trained, there weren't very many guidelines. And so we really had to go through an old fashioned kind of schema of, uh, what's the differential diagnosis? What are the possibilities, what makes this more likely, what makes that more likely? So I think that current doctors are at a disadvantage because they have these tools. When you have Tools. Sometimes the muscles that you used to use to do what the tools are doing for you get a little weak. But I think that when you don't have an algorithm to go by, then you have to go back to the old fashioned tools. Knowing what something is is an important part of it. So like recognizing the diagnosis is an important part of getting it right. When they've done studies where they present doctors with very smart doctors with complicated cases, it's almost inevitably among the people who get the right answer are people who have seen this before. So recognition is probably the most important factor in making a difficult diagnosis. Knowledge is key, listening is key. The way you ask questions is going to make a difference. I mean, so all of these things combine to help you make a diagnosis when the tools that are easier don't work.

Speaker C: Mhm. Yeah, that makes a lot of sense. And one other thing that I observed as well is that obviously technology has advanced so quickly and it's getting quicker and quicker. Right. So there's so many new therapies. I worked in a lot of rare disease therapies or rare disease areas. Even recently I was speaking with a, uh, client pharmaceutical company. They're coming up with a new therapy for rare disease that hasn't seen any change in a very long time. So the guidelines are actually 11 years old. Right. And this is now going to be the challenge. So guidelines are great tools for sure. I think it makes a lot of sense, it's very logical. But can the guideline development actually keep up with how fast technology and new technology comes to market? Right. Is a good question. So I think this fallback mechanism as you described is so important. So when you watch how diagnosis really happens in practice, what would you say? You think most people, maybe especially people that are not working in medicine at all, misunderstand about that process?

Speaker A: I think the most common misconception is they don't understand that it is a process. I mean, uh, certainly when I went to medical school, I had never had very much interaction with medicine before and I thought the diagnosis was like math, 4 times 6, it's always 24. So I just thought these are recipes that you learn. You know, when you see this and this, it's this. It turns out it's not like that. I mean, it could be a lot of things. And so recognizing that there are steps that you have to go through to approach something in a reasonable way, I think is one of the things that people find hard to believe. I hear frustration in patients when they tell their stories because the doctors seem to go in this sort of stepwise thing that seemed very slow to them, but I think doctors need to do it in that way. It's drilled in us from day one. Common things are common. What do you hear? Hoof beats? Think horses, not zebras, you know. And so that means you really have to make sure you think about the ordinary things before you get to the weird things. And I think that patients who always think that their disease is unique and every disease is unique, get frustrated by being sort of lumped in with everybody else who has fever and a cough. I think sometimes they feel like their individual suffering is not being valued or attended to. That's on us. That's a failure of our communication. I think one of the real problems of medicine, um, and medical schools, and I don't know how it works everywhere, but certainly here, your test scores are the most important thing to getting into medical school and to getting into residencies. I mean, they do do interviews now, which I think is important. But when I did this, when I applied to medical school, the interview was kind of a, uh, hat tip, you know, I'm not sure how much weight it carried. It was there to just try to make sure the crazy people didn't get in. But you weren't being, you know, or people who clearly had something wrong with them, even though they were super smart. But a communication skill which is at the heart of medicine, no matter what kind of medicine you practice, wasn't evaluated, it wasn't part of the process. It wasn't something that was appreciated in me. And so when you start off with a bunch of people that aren't good communicators, your work is really cut out for you. Because medicine is all about communicating.

Speaker C: I'm speaking my language here, that's for sure. So what I'm hearing from you is that like people think it's kind of maybe almost like a logical binary exercise, right? You check a data point and then you know what's happening. But in reality it's more like messy and uncertain, which must be kind of also sometimes a nerve wracking experience. As a physician, I would say now to the point of where you feel that physicians could maybe communicate better for patients to not feel their lumped in with these other groups. What could they do to have their patients more of a feeling that they are unique with their unique problems and feel more engaged in the process as a result?

Speaker A: I mean, you should listen to them. You should be working in a situation where you have the opportunity to listen to them, where you're not allowed 10 minutes to have an, uh, interaction with a patient where you have the time that you need to listen to them. And then I don't know who designed medical exam rooms, but somehow in medical exam rooms, the doctor's back is always to the patient or to the side. They don't seem to make it so that you can take notes and look at the patient. Good communicators master this. Yes. And what else happened? I see. So that you can look at the patient while you type and not see. But I think being quiet and letting a patient talk and actually listening. There's this old technique I learned when I worked on a hotline when I was 18. Hotline where people called in and talked about their trouble. And they taught us this thing called active listening, where you let the person who is talking understand that you heard them, and you're checking to make sure that what you heard was correct by kind of repeating back what they say. So what I hear you saying is, uh, and doctors need to do that more often. It, uh, first of all makes sure that their information is correct, and it lets the patient know that they've been heard. I cannot tell you how many patients say to me, and I'm sure say to other doctors. I just don't feel like doctors are hearing what I have to say. I think doctors might dismiss this as cosmetic, but how you appear, um, and whether you appear to be listening is really part of that whole communication bargain. I have, uh, a talk about this, and I have what patients want. You know that great Luke Fields painting where it's the doctor looking at this sick child in the bed bending over, and the parents are in the background worrying. That's what patients want. Then there's another picture of a bunch of doctors surrounding the bed of a patient, all of them on their phone. Now, uh, probably. Probably they're looking something up. They're listening, but they don't look like they're listening. Certainly, if the patient is talking, they're not feeling heard. And so it might seem silly and superficial, but looking like you're listening is important. And also, if you're texting on your phone, I don't believe that people can actually multitask. If you're looking up something on your phone, you're probably not listening.

Speaker C: I share that opinion with you. I think we're just telling ourselves that we can multitask, but we actually cannot.

Speaker A: I think that's true.

Speaker C: Yeah. And then you're right. Like, even if it's the impression you're not listening, even if you are. It's not nice. Right. Because you would feel dismissed and not important. So I think this is really admirable. Uh, and hopefully that's an area where medical training can include in their trainings as well. So these communication skills get improved.

Speaker A: But uh, let me just say even when medical schools do focus on that, when you get into practice, what you see around you is different. And so you have to bulk people up so that they're ready to be different. Not just in training, in real life, you don't see it all the time.

Speaker C: Yeah, yeah. So it's like an existing practice that has been going on for so long that it's kind of hard to change. Right. So you probably need to be very insistent. Yeah, I can see that. I was just wondering, since we're in the business of medical communications here on my end, is there any role for like materials in your mind? Like materials you can show when you talk with a patient, when something is maybe complex to explain or materials you can give to them so they can read it at home, which could be in any format as well.

Speaker B: Right.

Speaker A: Uh, you know, I give handouts to my patients all the time. I have a wonderful super smart nurse who put together little two page explainers in long Covid. You see things that I'd never seen before that people have to live with. So she gave explainers about that. I feel very comfortable picking up a pen and drawing pictures of what I'm trying to get across if that's going to do it. So I think those are important, but they can't replace what the physician has to say or the practitioner.

Speaker C: Yeah, it's more like an augmentation, Right?

Speaker A: Absolutely.

Speaker B: Is medcoms evolving faster than your team can keep up? You're not alone and you don't have to navigate it alone. Wesley Portiges, host of um, Transforming Medical Communications, is offering a limited number of free, free 30 minute consultancy sessions for medical affairs professionals. Whether it's medical communications, field medical or medical affairs training, these sessions provide an opportunity to address your real world challenges, pressure test ideas and walk away with practical strategic insight for your team. No fluff, just strategy. Follow the on screen link or in the show notes to book your first free session now.

Speaker C: So before when we were talking about how we think diagnosis is being done and how it's actually done, I heard you say that phrase about common things are common. Right. And the zebra and so on. You are active in the long Covid therapy area and I could just imagine that this has come up quite a lot. This concept Are there any anecdotal experiences here around diagnosing and that concept that you could share with us?

Speaker A: You, uh, know there was a paper early on that said 200 symptoms associated with long Covid. So you think, oh, my God, this is going to be so complicated. Which actually, to be honest, is why I started the clinic. I thought it would be fun and complicated as fun and complicated. But I see that there are patterns in all different kinds of diseases, and I see the patterns in long Covid. And so that makes it much easier now than it was when I started. I mean, I think an important part of diagnosis is to have a background of experience so that when something new comes in, you can sort of see where it fits into other things you've seen before or whether it doesn't. I would say maybe 10% of the patients that I see, maybe more are there because their doctor said, I don't know what this is, maybe it's long Covid and send them to me. And, uh, it's useful to have this experience of having seen a lot of long COVID patients to say probably your diagnosis needs more investigation, because long Covid is a diagnosis of exclusion for the most part. There are a few things that we can actually diagnose with testing, but even diagnosing them with testing doesn't mean that long Covid caused them or that Covid caused them. Um, so it's been exciting. When I see zebras, then I know that the chances are it's probably not long Covid. I mean, I think currently, researchers who put together the underpinning of what we do think that most of long Covid is sort of an immune system gone wild. And that's crazy. That's crazy and interesting, but it also does it in predictable ways mostly. So when somebody doesn't fit into that pattern, there's a very good chance they don't have what I'm looking for. And then the job is to figure out, well, who's the right person for them to see next, which is also fun to do.

Speaker C: Yeah, I can absolutely see that. So, and I guess, like, there must be some level of cognitive bias as well, right, when it comes to diagnosing?

Speaker A: Oh, absolutely. Very early on I wrote about this. A patient came in and she was tachycardic. So we have this diagnosis in post infectious illnesses like long Covid of postural orthostatic tachycardia syndrome pots, where people's heart rate goes up just from standing upright. It's a dysautonomia. So this woman Came in, she had a positive test. I'm like, okay. And she had had Covid. Of course, everybody's had Covid. And so I said, well, I think this is what you have. And I started her own therapy, and she didn't get better. And she came back, and I realized that I had just assumed she had long Covid. And I was thinking of the long Covid things. What long Covid thing does she have? She didn't have long Covid. She had hyperthyroidism. She had Graves disease. But you come in with a certain, uh, sort of expectations. You come in with blinders on about what you're going to see, and it's not surprising that that's what you see. And we have to fight against this all the time.

Speaker C: Yeah. Yeah, I can see that. Do you think there is a way where the pharmaceutical industry could somehow help with some of this? Right. Like, they basically have the data on their different products. Right. And do a lot of research. And if the answer is no, that's fine. I'm just wondering what your view is on that and how you consider their role potentially in helping clinicians diagnosing better.

Speaker A: I'm not sure they have a role. Their role is in treatment, not in diagnosis. Treatment is the step that comes after diagnosis. Uh, if you'd ask me about laboratories or companies that make imaging systems, I might have, uh, more to say about that. But for. Pharmaceutical companies put out all this advertising information about whatever rare diseases they have an orphan drug for to try to raise awareness in physicians, and I guess that's okay. Pharmaceutical companies come with a lot of baggage and history of not being honest or not being reliably presenting everything you really need to know. And that has, I think, interfered with. With trusting them. Certainly for me, for example, pharmaceutical companies are not allowed in our medical school at all. They're not allowed in the hospital. So, I mean, we've really been. We really keep them at arm's length, or Yale does. And I'm not sorry about that, because once you hear a fact, it's very hard to unhear it. Like, I'm very careful about where I get my information from, because once you learn something that you think is true, it goes into this big oatmeal box with all the other information on that topic you have. And you can't go, oh, right. That came from this unreliable source. Sherlock Holmes described his brain as an attic. And you have to be very careful about what you store in your attic because you have limited room. And I think that's true. And that's why I'm less enthusiastic about pharmaceutical companies information on diagnosis.

Speaker C: Yeah, I do respect that. Like, I do hope that the industry overall has cleaned up their act over the last decades. Right. And I know there have been several examples, of course, of what you just stated. But at the same time, there's this concept of the medical affairs side of pharmaceutical companies that has been working really hard to be completely separate from their commercial function with scientists sharing information and discussing science, which I think is pretty interesting. But I'm not trying to rebuke your point here.

Speaker A: No, no. But when you're getting paid by somebody, it's really hard to go, uh, oh, they're totally wrong. It happens. But it doesn't happen all the time. I mean, the fact that people are paying you subconsciously exerts pressure.

Speaker C: Yeah, I do agree with that. So I'm happy there's a lot of compliance rules in place nowadays.

Speaker A: Yes.

Speaker C: And hopefully there will be other ways where they can be helpful with that. So, switching to a completely different topic. I hear a lot about AI, Right. On a daily basis when I speak with people, either from the medical practice or from the industry. And not so long ago someone told me, like, listen, Wesley, I think the role of doctors is going to be so much less important. Now we have AI because basically you can just like ask AI what test to do, do the lab test, feed the results into whatever ChatGPT, and it would probably be able to make a pretty good diagnosis, which I think is a very bold statement. I'm not saying I agree, but I would just like to hear how you think about it, and especially knowing that that would mean there's no physical exam, right?

Speaker A: Well, the physical exam provides some information. It used to be the only way we had to get information. And I think if you include pocket ultrasound as part of the physical exam, I think it can still provide information. So I don't think that's a limit, making a diagnosis. I think AI should be very helpful. There are studies that show that when fed a case with data into AI that they outperform doctors in terms of making a diagnosis. But who got that information and how, and who put it in and who decided what was important and what should be left out? Did AI decide that? Did you just dump everything in? And maybe you did, but all the things that you dump in there are things that were written by doctors who are making decisions about what should be included and what is it now when patients interact directly with AI. Uh, I haven't seen any studies about that about whether AI continues to be better at making a diagnosis when it's the patient who's inputting the information compared to the doctor. So I have no doubt that that's going to be true one day. You know, I look forward to the day. Did you ever see that show Star Trek? There was a doctor there named Bones McCoy. He had this little device that he just waved over patients that told him what was going on. He still had a job. I was still a doctor. A, uh, doctor was still needed. So I'm not at all worried that AI is going to put us out of business. I am worried that, like all automated answers, it's going to be valued tremendously. And that means that doctors are going to have even less practice about thinking through a diagnostic problem. I mean, if algorithms and guidelines have impacted our ability to think, you can imagine that AI will. So, I mean, I guess, uh, we'll see what happens.

Speaker C: It feeds into maybe in the guidelines topic we spoke about earlier. Right. So maybe the new way of doing this is not a guideline as we know it like that is pretty static, but maybe it's more fluid so that they are always updated. Right. You know, my takeaway from what you're saying, and I think personally I see this as the most important part, is like, AI is a power tool, but you need to be able to operate the tool properly. Whoever is operating the tool is also what, ah, defines the outcome. So sometimes, as they say, garbage in is garbage out. Right. And that's kind of what I hear you say. And I think that's very exciting because I think, I feel this is an opportunity where hopefully we can reduce the burden a little bit on our physicians, because there is a lot of data out there that many physicians are completely overworked and overburdened for many different reasons. And hopefully this new onset of technology will make our lives a little easier.

Speaker A: Well, I have A.I. uh, that helps me write my notes. And I would say it's very a. And not very I. But I'm eager for it to get better as long as it does the best it can. It's just that it's not there yet. I take my own notes while I'm in the room with the patient. But when it comes to the assessment and plan, I use what AI has developed. So it's changed the way I talk to my patients. At the end of the appointment, I always say, okay, here's what we decided to do on the topic of your fatigue, we're going to do. And so I lay it all out and AI still gets it wrong, but I revise it. I know that there are doctors who are going to feel pressured to just take the note and put it in the chart. And I see that happening all the time where the information is just not right. And I don't know what we can do about that. Because what's really going to happen is our employers are going to say, well, now that you have AI that writes your notes for you, you can see more patients. And I think when people embraced AI, they thought, oh great, I'll have more time to talk to my patients. But in fact it's going to work to give you less time to see your patients. Or that's my fear.

Speaker C: Yeah. I think honestly that is the same in the corporate world. Right. It's just like the expectations will increase as a result. And I think when we discuss this topic today, it's very clear that the diagnosis is not just a scientific process. It's really like a human one. Right. And I would say preserving some time to establish and keep and improve that human connection between clinicians and patients is so important. So hopefully we'll be able to leverage that way. I just want to say thank you, Lisa, for this inspiring conversation. This was a lot of fun. Thank you for opening a window into your daily life and sharing your expertise from both practicing in medicine, but also all the different articles you have written in the Times. So thank you so much for being here today.

Speaker A: Well, thank you.

Speaker C: Thank you for listening to Transforming Medical Communications. If you found value in today's conversation, please share this episode with a colleague who needs to hear it. And don't forget to rate and follow the podcast so we can help moving Medical Communications forward.

Speaker B: Transforming Medical Communications is brought to you by Medcom's Experts. To find out more about Medcom's Experts and how we create some of the most cutting edge medical communications programs anywhere in the world, visit www.medcoms-experts.com and then make sure to search for Transforming Medical Communications in Apple Podcasts, Spotify and Google Podcasts or anywhere else podcasts are found. Make sure to click subscribe so you don't miss any future episodes. On behalf of the team here at medcoms Experts, thanks for listening.