Eric Topol and Abraham Verghese on What AI Means for the Physical Exam

Medicine and the Machine

; Abraham Verghese, MD


July 03, 2019

This transcript has been edited for clarity.

Eric J. Topol, MD: Hello. This is Eric Topol with Abraham Verghese, and we're thrilled to start our second podcast of Medicine and the Machine. Today we're going to talk about the physical exam: how it evolved over time, where it's going, and how artificial intelligence (AI) may someday play a role. Abraham, what are your views of the physical exam?

The Ritual of the Physical Exam

Abraham Verghese, MD: I love to talk about the physical exam. One thing that surprises students is to learn how recent our ability is to examine the body and make the most basic observations. I date it to Auenbrugger in the late 1700s, who began tapping on people's bodies, much as he'd seen his father tap on the casks of wine in the bottom of the inn where his father worked. In a sense, Eric, that was the ultrasound of its day. That was the way to detect fluid in the lungs, consolidation, enlargement of the heart and liver, and so on. It was a profound moment in medicine and it was followed very quickly by the development of the stethoscope, the reflex hammer, the blood pressure cuff, and so on. I think the carrying of the stethoscope represented a signal that we were no longer barber surgeons but were committed to trying to make a diagnosis on the body, and maybe even characterize what kind of illness this was. So, [the physical exam] is very recent compared with the history of medicine in general.

You and I trained in an era when the bulk of our dataset came from that exam of the patient and an unlimited number of laboratory results carried in our heads. One of the reasons I went into medicine was seeing physicians who had this magical capacity to come to the bedside and feel the pulse and PMI (point of maximal impulse), and look at the neck veins and tell you, "You're going to hear the murmur of mitral stenosis, and you won't hear it unless you use the bell of the stethoscope lightly applied to the patient in the left lateral after some exercise." It seemed like a kind of magic—that they were divining things.

Topol: Some of my most memorable experiences in training were with Kanu Chatterjee, the master of the bedside exam of the cardiovascular system. Just as you say, he could predict, to within a millimeter, the pressures in every cavity of the heart. It was extraordinary. Let's get to it from the patient perspective because you have so elegantly described this ritual and the expectations from the patient.

Verghese: It's become terribly important to try to understand that because, clearly, we have technology that surpasses the accuracy of the physician examination for most things. A few things, like neurology, still need the exam to determine functional deficits. What is the argument and the meaning of all this? Why do we need this from the patient's point of view and from the physician's point of view? It's a very important ritual. If you look at rituals, in general, they are all about crossing a threshold. We marry, we have baptisms, we have funerals—all with ceremony to indicate the crossing of a threshold. If we step back and look at the physical exam, it has all the trappings of ritual. You are coming into a room with furniture that does not look like the furniture in your house or mine. One person in this dyad is wearing a white ceremonial outfit with strange instruments in the pocket, and the other is in a paper gown that no one knows how to tie or untie. That is part of the mystery. Incredibly, they divulge things that they would not in any other setting. And then at a moment in time, they disrobe and allow touch, which in any other context in our society is assault. So, it already has all these trappings of ritual. I'm struck by the fact that patients from many different cultures and ethnic groups, who may have very different beliefs about illness, quickly understand that you're about to enact a ritual. They can be disappointed when they notice that you're not doing it well because, to them, it's a kind of inattentiveness. It's very similar to what we judge in our baristas, our hairdressers, our short-order cooks. We don't necessarily know how to do what they do, but we know when someone is doing it smoothly and easily.

If you go beyond that, it serves an important function of localizing the disease on their person—not on an image somewhere, not in numbers, not in histology slides. There is something about the exam that I think really validates their personhood. We live in an age when we're almost becoming disconnected from our bodies. We can spend hours on screens and on our phones, and be almost unaware of our bodies to the point where we hurt ourselves with neck injuries, carpal tunnel syndrome, and so on. It becomes a very important moment, especially when they are ill, to reconnect to the body and to affirm the bodily nature of this illness, not some abstract thing happening out there.

The Basic Need of Touch

Topol: The term "laying of hands" has a lot of meaning. Of course, what you're getting at is that there is less laying of hands and there is an expectation of patients—they want to be examined. They want the touch and they know when they are not having the real exam. After having a knee replacement and having such a tough time, I remember the orthopedist not even examining the knee or having me walk, or whatever. A patient of mine from Cleveland who flew out to San Diego was having problems with angina. I had my colleague here see him because they wanted to whisk him away to the cath lab. He did an examination and I was in the room. But after the procedure, when he had a stent that my colleague did, I went to see him in the hospital and he was very mad at me. This is somebody who I had a relationship with for more than 10 years. I said, "What's the matter?" He said, "You didn't examine me." I've never forgotten that. That was an expectation.

I think there is a profound human need, especially in the context of illness, to get the sense that you have the attention of that person.

Verghese: To echo what you just said: I think there is a profound human need, especially in the context of illness, to get the sense that you have the attention of that person. You mentioned therapeutic touch. There is a whole body of somewhat fuzzy literature on that, much from the nursing community. I don't doubt the importance or validity of it. But when I talk about this, I think it's helpful to accept that there is perhaps such a thing, but to focus more on the ritual and what that accomplishes. In my mind, I think that your desire to have yourself examined by the knee specialist, your patient's desire to have you examine him, is almost like a fundamental human need in times of illness.

We're learning now that the placebo effect, when it works, is profound. It does not just trick the patient; it's producing a real neurohumoral surge of endorphins and things in the brain. We're recognizing that you can have a placebo without a placebo— meaning touch, tone of voice, setting. All of these things have a tremendous influence and also cause the same neurohumoral surge. This profound need of ours has a therapeutic implication. I have an anecdote from a distinguished colleague who just passed away, a very eminent cardiothoracic surgeon who late in life also got his PhD in English. He had a shoulder problem, and apparently ultrasound is state-of-the-art now for shoulder. He saw the ultrasound tech and then saw the best surgeon around for this problem. That surgeon could not have been more attentive and more solicitous of the senior colleague. It turned out that they could not do surgery. Later, when he was telling me about it, he said, "But you know, Abe, I wanted him to touch me." Hearing that from a cardiothoracic surgeon just threw into relief how basic this need is to be touched. Beyond touch, don't you think that in this complex age, we're still in great danger of overlooking the simplest sort of things if we don't make that contact with the patient and don't do a quick survey to make sure all of the numbers and stuff are gelling with what we're seeing on the body, with what we're seeing on the patient's face?

Importance of Teaching the Next Generation

Topol: I agree. This part of medicine will never be digitized. It's not a machine learning experience. There is not going to be any algorithm that is going to touch the patient, that is going to provide that check, that oversight, that human experience. Bilaterally from both the patient and doctor, aside, it's a way to confirm what is going on. It's of fundamental importance. One of the reasons that may have eroded the willingness to do a more thorough exam is lack of time. Doctors feel rushed. And they also rely too heavily on getting scans. They wonder what they are going to do with the exam when they could just send the patient for a total body scan or something short of that, of course. As you wrote about some years ago, basically the default is to revert to the iPatient rather than the patient.[1]

Verghese: I think you are right. Part of it is our failure to communicate these skills and teach them to the next generation as well as we should, and to make them relevant and valid. It's clear that in any ritual, whether it's saying high mass or hitting a baseball, you need a long apprenticeship to do it very well. And you need a lot of coaching. Clearly, we're giving lip service to this aspect of the training. Our residents can talk your ear off about sodium and what it means if it's low or high, but that same facility they don't necessarily have at the bedside. It's not their fault; it's our fault. We have not imparted that to them and they need to see us doing it in a way that is useful.

We did a study some years back with my colleagues, John Ioannidis and Jerry Kassirer, where we asked physicians to tell us stories and anecdotes of oversights in the physical exam that led to consequences. The study is still ongoing, but we published in the American Journal of Medicine when we had 200.[2] The basic problem was that most of the oversights were because the exam had simply not been done. And yet for the recorded exam in the computer, every checkbox had been ticked, suggesting that it had been done but the person submitting the anecdote knew that it hadn't. We have 200 of those kinds of oversights, such as diagnostic delay, contrast exposure, surgical misadventure. I told my friend Atul Gawande that I want to write another book called Checklist, but mine will be very simple. It might be three or four pages. And the first one will be: "Patients have a frontside and a backside. Look at the backside; you can see things happen between the backside and the bedsheet." I'm being facetious, but I know you agree with me that the moment of seeing the patient is not just a ritual. You might see things that do not gel with the information you have otherwise (especially in the intensive care unit), and rather than act on the numbers, it's always helpful to take a good look at the patient too.

The Modern Physical Exam

Topol: I like that short checklist. Another part of this story is the modern physical exam. We don't want to lose any of the things we've been talking about, but we want to build on it. The thorough physical exam is one of the memories I have had for many years because of the influence of my mentors like Kanu Chatterjee and Arthur Moss. I used to love to teach at the bedside the splitting of the second heart sound and other things you could hear, see, and the touch. Now it's hard to belabor that and spend all of that time trying to listen and convince the students, residents, and fellows on rounds that they should hear things that you're hearing when you can see it. Now you can bring out a smartphone ultrasound as part of the exam and see much more information. Some people are revolting about this, which is not so dissimilar to the stethoscope in the 1800s, which took 20 years to get accepted. We're in the early years of that perhaps 20-year revolt. But the point here is that people think that it's going to lead to all sorts of incidental findings. In my view, it reduces the need to even think of sending a patient for an echocardiogram, or for other parts of the body. For example, you may think there is an abdominal aneurysm, and then you get out the ultrasound and see that the aorta looks pretty good. Should an ultrasound with or without AI support, whether in obtaining or interpreting the images, be considered part of the modern exam? Do you think it will evolve that way?

Verghese: It's an unequivocal yes on my part. But before I say more, I want to pick up on something else that you said, which is that there is this trope out there that the physical exam is useless and that there are studies to back that up. I think that that is referring mostly to the "executive physical," where you do a well-person exam. It's helpful to do—a bonding takes place and you stumble onto things—but I would agree that, by and large, the routine physical isn't much use. However, people have conflated that with a patient coming to the emergency room or your clinic with a specific problem. I think the physical exam done with that problem in mind has a very high yield. For someone coming to your clinic with shortness of breath, you're looking for a whole different set of things than if you're just doing a routine physical on a well patient. Making that distinction, the physical exam is extremely helpful, very efficient, and it allows you to ask better questions of your tests when you're dealing with a focused problem. In some cases, like neurology, there is no other way to get at the functional deficits and to correlate all of these images with what the patient actually has.

But coming back to the tools we carry, I'm all for it. The more things that we can carry that allow us to not have to send people to suites and wait for reports to come back, and the more things that we can do at the bedside and interpret for the patient, the better. The ultrasound is a wonderful tool, provided we are all in agreement on what it is we can use it for reliably. We need to learn, as well as we learned from masters like Kanu, how to get a sense of contractility and fluid status, and how well the inferior vena cava collapses. Using the ultrasound to teach students at the bedside that that is, in fact, a neck vein and showing it to them on your little screen can be extraordinarily helpful. Anything that we can bring to the bedside that allows us to look in is wonderful. The key element is that we should still be there; it would be a mistake to do those things and disappear ourselves.

Topol: You are bringing up a really important point that I hadn't thought of until now. As Kanu Chatterjee and I became friends beyond him being a mentor, I asked, "How did you get to be so good at this? You are one of the best in the world at being able to detect and differentiate in the heart exam." He said that for all the time he was at Cedars-Sinai, he would examine patients at the same time as getting the pressures in the heart with the Swan-Ganz catheter. This sharpened his exam. He did this correlative type of long-term training, almost like training an algorithm. Nobody had that experience because we didn't have hundreds and hundreds of Swan-Ganz catheter insertions, and he deliberately went after that. [Similarly,] you're bringing up the point that the use of this adjunctive tool to see things can actually sharpen your exam.

We've disowned the body and we've taken ownership of the data. We don't trust the body.

Verghese: It should sharpen your exam. William Osler was by all accounts a wonderful clinician—good at the bedside, and probably no different from the top people of that era. But he had to rely on autopsy. One year he did something like 400 autopsies to get at what was really going on. I've always thought that this was peculiar because we can, in life, see all the things that Osler had to wait for on autopsy: vegetation on the valves, size of the spleen, size of the liver, and all that. You would think that it would make us so much better at the bedside, because we'd have these Kanu Chatterjee kind of affirmations with our data. But I think exactly the opposite has happened. We've disowned the body and we've taken ownership of the data. We don't trust the body. And I think pitting one against the other is just a false dichotomy. We don't need to be doing that—they are both important. See the patient, examine them, find out what is tender, look at the data, and then bring them all together and make sense of it.

Embracing New Technology

Topol: This is another facet that I found recently to be illuminating. We were taught to take the pulse because the pulse is a window into the cardiovascular system. And it's a way to start the touch. We still want to take the pulse. However, the modern pulse, which I only experienced in recent weeks, was with a six-lead smartphone ECG with their exquisite tracings. You have the patient put their fingertips on the center of something that looks smaller than a credit card and they put it anywhere on their left knee or their left leg. Then they get a six-lead cardiogram, which you look at together with the patient. Again, here you have some doctors saying, "This is crazy. Why would you do that?" But I'm saying, "Well, wait a minute. This is more informative and it's actually kind of fun." It preempts the need for sending somebody out for a cardiogram when you can see things now.

An experience kind of shook me the other day when I started using it. I saw this one patient who came in for a second opinion. He had been worked up twice for a heart attack with high troponins. He had had a cardiac cath which showed normal coronaries, and he also had atrial fibrillation. So he came to see me to find out what was going on. I started off with the six-lead ECG after I did his pulse and it showed very low volt. I wondered if this could be amyloid of the heart. I got the probe, put it on the phone, and there it was: a speckled heart, the classic finding. How could you diagnose cardiac amyloid in 2019 with the smartphone? I mean, why do we reject this type of potential advance?

Verghese: You are absolutely right. It should not be either/or. Why not pulse and ECG? I'm thinking about the recent spate of airplane crashes, one of which was very close to my heart. A plane crashed in Ethiopia, the land where I was born, affecting an airline that I knew well. It was the particular model of airplane where the software was telling the pilots something and yet they were looking at the horizon knowing that it was wrong. I think we need those constant checks and balances.

Topol: I've already experienced that, by the way. If you don't put that sensor in the right upright position, you get a funny result that is all screwed up. So you are absolutely right. I do like the idea that these are complementary, that they add to the informativeness of the exam. They should not be summarily dismissed as yielding incidental findings, but rather enriching. That is the key. People will say, "You need to do prospective studies of the value." But was there ever a prospective study of the use of the stethoscope? Or palpation? Should there be a study about these things? These are like basic tenets of what an exam would be, and it's going to continue to evolve.

Verghese: There has been a whole area of scholarship on taking individual aspects of the exam, just one sign, and then talking about specificity and sensitivity and operator characteristics. There is something to that but we never do that in isolation. We're taking all these pieces of information together and coming up with something.

There was an anecdote in the New England Journal of Medicine that struck me.[3] Again, these are not common, but they are cautionary tales. A patient came in with liver dysfunction but a high lactate that didn't gel with his appearance. He looked pretty well. He had moderately elevated enzymes but his lactate kept being quite high. To make a long story short, the puzzle was only solved by the patient's visiting sister who happened to be a phlebotomist. She made the observation that the phlebotomist kept drawing from his veins right below where Ringer's lactate was hanging because she couldn't get it from some other site. It might have just been the proximity to the Ringer's lactate. What machine is ever going to tell you that?

Topol: Up until now, the physical exam was essentially analog. It's from doctors' and clinicians' observations, whether that is through palpation or visualization or auscultation. Then there is this digital component (eg, ultrasound or an ECG recording) as part of the exam, not to lengthen it but to add this other dimension. It is archived and given to the patient and can be put in the chart. This is a different dimension. There is a lot of discomfort with this, that it somehow crosses a line when you go from analog to digital. I don't really understand it.

Verghese: That is strange. I don't understand it either. I'm surprised.

Topol: For example, a patient comes in with abdominal pain and instead of reflexively sending them for an abdominal ultrasound, you say, "Okay, I'm going to look at their gallbladder now." And you see that there are stones in seconds. Part of the other story here is that there is no reimbursement for this and additional training may be required. We're seeing companies, like Ultromics, that are trying to basically use AI to provide the interpretation. The company Butterfly is telling you that you need to move the probe up to the left 2 cm and to the right to acquire better images. I don't know if we are ever going to override these obstacles about the need for accepting some added dimension to the physical exam without detracting from what ought to be part of it.

Verghese: If the ultrasound sonographer, namely someone like myself, isn't a trained ultrasonographer, I think what the data does is pretty much the same thing that data does for me in the rest of the physical exam. I might pick up that this patient has aortic stenosis, and rather than sending them for an echo to say, "What the hell is going on? I hear a murmur," I'm asking a better question of my cardiology colleagues. I know this person has aortic stenosis. I'm sensing that the valve gradient is borderline; can you tell me what it is? So this is asking better questions of our tests. Similarly, if we were to do ultrasound and pick up something like a gallbladder, then that next step would be to send them for a more formal study to ask a better question, which is: "We know this is a gallstone; can you tell us about the status of the bile ducts?" We can ask about things that are beyond what our training and our little handheld machines can tell us.

It's all about human connection. It's all about that sense of identification and connection.

Topol: People think that technology is depersonalizing and dehumanizing. Yes, that can happen. But when you have captured this video loop and are showing it to the patient in real time, this is something they would never have seen if they went for the formal study because they are not allowed to have the sonographer review that with them. You have this shared experience of teaching and going over the findings essentially in real time. And to me, that actually adds to the bond or intimacy.

Verghese: It's profound. We're talking about the physical exam and ritual, but we're also talking about connection. There have been some profound studies now to show that patients do better or are more compliant, or might even have better survival, when taken care of by a physician that looks like them. Why is it that they are more willing to get their information in the barbershop than in a sterile clinic? It's all about human connection. It's all about that sense of identification and connection. There is nothing like a skilled exam to break some barriers and instill a certain confidence. I'm not confident if someone does not touch me with finesse. It just tells me, "Well, it looks like you need to have this done." It's just hard to take that without a little bit of awareness. I want to I feel that I'm connected to this person, that what they said is backed up by how they were with me and how it felt.

Topol: It is interesting to look back. When René Laennec started the stethoscope, people revolted, in part because they said that this tool would disconnect them from the patient. And now as we evolve, that part of the exam is what people feel does, indeed, connect them. The points you are making are great.

This has been a fun discussion. The physical exam will continue to evolve. The interesting thing is that there are such rich aspects of it that hopefully will not devolve as they have, but will be supported and enriched as technology becomes part of the exam, which seems like it's inevitable in some respects. In some ways, it's exemplary for the medicine and the machine, this tension that we have.

Verghese: I like the way you summarized that. I think you are absolutely right—it's going to evolve. But it's important that it evolves. Neither of us are Luddites trying to keep things as they are. You, especially, are not a Luddite.

Topol: I'm accused of the opposite, of course. But hopefully, our discussion is enlightening—that there are some reasons to be thinking about this stuff.

Verghese: The most surprising thing about first meeting you was to recognize how much common ground we actually had, which is how this has come about. Thank you, Eric, for that.

Topol: I appreciate it. We're going to keep going with our podcasts. In the future, we're going to also be inviting some guests who will add to the perspective. This discussion on the physical exam has been fun. We could not cover everything, just like it's very hard to do a total-body physical exam thoroughly, but at least we hit some of the key points. Thanks very much, Abraham.

Verghese: Thank you. It's always a pleasure.

Follow Medscape on Facebook, Twitter, Instagram, and YouTube


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: