AI May Save Some Time, but It Can't Listen to a Patient

Medicine and the Machine

; Abraham Verghese, MD; Danielle Ofri, MD, PhD


August 22, 2019

This transcript has been edited for clarity.

Eric J. Topol, MD: Hello. This is Eric Topol, and I want to welcome you to this "Medicine and the Machine" podcast. With me today is Abraham Verghese and our special guest, Danielle Ofri, who is our first guest for this podcast. Welcome, Danielle.

Danielle Ofri, MD, PhD: It's great to be here, Eric.

Topol: Danielle and Abraham go way back, so I'll turn it over to Abraham to get things rolling.

Abraham Verghese, MD: Thank you so much, Eric. Danielle, it's great to have you on this podcast. Your whole career, unlike many of us, has really been spent at one storied institution, Bellevue, which we associate with every wave of immigration and the last bastion of care for so many of the needy in New York City. You have written so much about this. You have watched technology move in from the era when most of us were using paper to where we are now—and I don't just mean the electronic medical record (EMR). What has been your grand look back at how it has been?

Ofri: It depends on how far you want to widen the lens. Bellevue is the oldest public hospital in the United States. It started back in 1736 and every generation has been importing technology, from the syringe to stethoscope to all of the new technologies. With every wave there is always a period of hesitation, concern, and a little bit of umbrage, but then we make our peace, find what works, and either take issue with or try to improve what does not work. Current technology is no different.

We're always a little bit behind in the public hospital sector compared with our private counterparts, who usually hop on technology first. But sometimes that is an advantage because when the negative parts come through, as they inevitably do, we let the folks in the private sector take the brunt first and then we can be a little more thoughtful in how we approach it.

Bellevue Literary Review

Verghese: I always thought that the wonderful humanism and humanity of an institution like Bellevue was so well captured by voices like yours. I always marvel at our mutual friend and your former chair's idea of putting together a literary publication, the Bellevue Literary Review, which went on to become a publisher with a Pulitzer Prize–winning book among its titles. Talk a little bit about that contrast of technology, cutting-edge medicine, and the literary journal.

Ofri: Certainly no one was expecting to be involved in a literary journal when it started out. I was an MD/PhD student. I did a PhD in biochemistry and was fully on track to be a laboratory scientist. In fact, I was headed toward neurology since I was a neuroscience researcher, but I did my 1-year preliminary internship in medicine at Bellevue and completely fell in love with the patients and the stories, so I stayed with internal medicine. That was during the height of the AIDS epidemic which, as you all recall, was quite a difficult and dark time. So when I finished my 10-year of stint of medical school and training, I really wanted a year off. During that time, I began to write down the stories of those patients and I remember that it was a singular moment. I knew there was something historical about those patients we were meeting and that I should be writing things down, but I didn't have the time and I also think we were too close to the emotional bone. Only when I stepped back could I really start to think about the patients in more detail. When I came back to Bellevue, I wanted to incorporate this somehow but didn't quite know how.

Marty Blaser came on as a new chair, and an astute colleague had us meet because he was having his students write a 1000-word essay on their medicine clerkship about anything inspired by patients, and I was asking my students in the clinic to ask a patient what it's like to have their illness, and write me an essay about that. We compared our stacks of essays and thought initially that maybe we'd make an in-house journal, kind of a mimeographed book of student writings. But as we talked more, it became apparent that there was a wide societal interest and fear about our health and vulnerability, so we decided to make a literary journal to use more creative ways to explore these issues. We began the Bellevue Literary Review on the 16th floor of Bellevue, right off the medicine wards. We took out a two-line call for submissions and had 1000 submissions in the wink of an eye.

It was clear that what people think about medicine and health and illness is not served by the "top 10 tips for osteoporosis" kind of medical writing that is already out there. Fiction, poetry, and creative nonfiction are ways to grapple with the issues that are not necessarily addressed by typical medical journalism. That is really where it started, and we're coming on to 20 years now. We get about 4000 submissions a year, so it has not let up.

Time to Think

Topol: That is extraordinary and another reflection of the human side of medicine—certainly not one to be mimicked by machines, at least in quality of writing. I am interested in going back to a couple of your recent pieces. I read everything you publish; you are so talented and able to articulate things that people in the medical community identify with. Your piece "Perchance to Think," in the New England Journal of Medicine,[1] was quite exceptional. The metaphor was that medicine has become a petri dish for medical errors, patient harm, physician burnout, unnecessary tests, unnecessary referrals, and more. Can you elaborate?

Ofri: It's interesting; one never knows what will resonate with readers. For me, I was just writing about a really rotten morning. A patient had walked in with a message from another doctor: "Rule out adrenal insufficiency and rule out rheumatoid arthritis," based on some random blood tests that had been sent. I was trying to take care of that while taking care of his other 10 chronic issues, and realized that I simply could not process adrenal insufficiency at the same time as handling his diabetes and obesity. What I really needed was some time to think.

I just couldn't do it in the moment, and part of me felt like a dolt because I simply could not, off the cuff, remember which way ACTH [adrenocorticotropic hormone] diurnal variations run, and in the end I sort of scribbled out "Refer to endocrine" to let them figure it out. But I felt like I punted because I know an internist should be able to work out the basics of adrenal insufficiency and rheumatoid arthritis—that is our job, after all. But in 15 minutes, it is not possible to do it with any amount of thoughtfulness. So many of our referrals and tests are because we don't have time to think. If I had a half an hour or an hour to think, I probably could have gotten by without a lot of the tests and referrals that I ended up doing.

The book I'm working on now is on medical errors, so I was attuned to the genesis of medical errors and where they start. I could feel them starting at that moment because I knew I was giving short shrift. I was trying to read UpToDate while he was talking, and I was trying to write what he was saying and read the things he was bringing in at the same time, and I knew that I would easily make mistakes. It really does feel like a petri dish for error, and that if we want to tackle medical error, we can't just keep making checklists, though they're certainly helpful in some situations. We also need time to think and organize our thoughts, especially when it comes to diagnostic error.

Verghese: You wrote about that so beautifully, and I think it echoes that famous series of photographs of William Osler at the bedside where he's inspecting, palpating, percussing. But then at the end, there is this pose of him with one leg up on a chair just looking at the patient with his chin in his hands, really contemplating. Hypothetically, do you think artificial intelligence (AI) will save us one day by taking all of those many facts that were whirling around the patient in your scenario and sorting it out, rather than forcing you to do it?

Ofri: I could envision AI helping with some of that. There was a lot of preparatory drudge work that wasn't my time well spent. If somebody could've said, "Here are all of the tests laid out, here are the possible tests you need, here are four reasonable options you can do, here are the ones that are validated," then maybe I could use my 15 minutes more efficiently instead of madly going hither and thither trying to pull the information out. I could certainly see that.

What I don't think AI could do is help [counsel] a patient. "We have six medical issues here. What is the most important for you and what is your goal in our medical care?" Even if I put all of his attention on this possible adrenal insufficiency at the expense of his diabetes that was really wearing him down, maybe I was not doing his overall health a benefit.

Time as a Determinant of Problems in Medicine

Topol: It's so remarkable. You have this situation where you were kind of self-deprecating and said, "I'd managed to come out of this single case feeling ignorant, inept, and disgusted all at once." In fact, that is very, very common because we have so little time. David Feinberg, who was a CEO at the Geisinger Health System before recently becoming a leader at Google Health, told me about a program at Geisinger that he had initiated, where the doctors and patients were not allowed to leave the room for 40 minutes. He said it was the most popular program they had ever initiated. The patients loved it because they had time to communicate, and the doctors thought it was extraordinary because it was three times as much time as they previously would have ever had with a patient.

The gift of time—that is why not only the title of the piece but its content really struck home. How much do you think time is a determinant of the problems we have today, both on the side of doctors and patients?

The problem is that time keeps getting filled.

Ofri: I think it's an enormous part of it. The problem is that time keeps getting filled. As we get certain tools to make life easier, such as the EMR, other things come into their place, and so it's the unopposed time that we need to actually discuss. What is it that we do well as doctors? What do we do better than computers? Because Lord knows that the computer knows the 25 types of vasculitis better than I will ever know them; there is no point in memorizing them. But figuring out what the patient is most afraid of, or wants to focus on, or what is happening in their life that is really impacting their ability to take their medications, are things that we do better. We need to have the time, but time that is not filled in by other tasks that are easily piled on.

Topol: In the New York Times, you had an op-ed about the healthcare system exploiting doctors and nurses.[2] You were basically getting into the fact that healthcare has become corporatized. There was an exceptional paragraph, a metaphor again like the petri dish, but this time it was about a factory. You wrote, "In a factory, if 30 percent more items were suddenly dropped onto an assembly line, the process would grind to a halt. Imagine a plumber or a lawyer doing 30 percent more work without billing for it. But in healthcare there is a wondrous elasticity—you can keep adding work and magically it all somehow gets done. The nurse won't take a lunch break if the ward is short of staff members. The doctor will 'squeeze in' the extra patients." Can you amplify on that?

Ofri: One of the things I love about working in medicine is my colleagues. Nurses and doctors are incredible people who, by and large, want to be there for the right reasons. No one goes into medicine now for fame or fortune. People go to Wall Street for that if they really want to make money a lot faster. But people who go into medicine generally want to be there and want to help patients, and so they do the right thing. By and large, we do. No one clocks out without getting their work done. At some point the system has "realized" that. I don't think it's a plot to take advantage of doctors and nurses but a recognition that if you add extra work in or squeeze in more patients and require more work, it will get done because doctors and nurses do not shirk their duty. If they don't do it, their patient will suffer. In what other profession would that happen? We don't think much about it. At some point, it becomes an exploitation of the professional ethic of clinicians.

Verghese: Do you think, in some measure, that the failure is on our end because we have not come together nationwide or as a profession to say, "These are the imperatives that we must have in order to take care of people well"? I think it is our fault that, unlike the assembly-line workers, we don't protest or organize in that fashion.

Ofri: I would say partly, but I think there are reasons why we don't. One is that we feel like it's a profession, not a job. We feel like it is part of our duty, and we take call after hours, and we interrupt our kids' ballet recitals to take call. We're also reasonably well paid and we accept that, and we get respect from society, so it feels a bit unseemly to sort of kvetch about taking all of these calls after hours, so we mostly don't feel comfortable doing that. But you are right—at some point we've reached a tipping point where now not only is it destroying doctors' lives, it's impacting patients and patient safety.

The EMR and the Demise of the Patient Relationship

Topol: I totally agree with that. There is a long history that you eluded to, Danielle, for doctors not thinking about anything like that. They are head-down and busy with their work. A lot of things happened along the way to promote corporatization—not just relative value units, but in more recent times, the EMR. There wasn't a whole lot of fight against that, particularly because they were designed to promote billing and had nothing to do with promoting the lives of patients or doctors or clinicians in general.

The electronic record is another part of the story that seems to have really led to the demise of the [patient] relationship, and unfortunately that represents the most digitization that has occurred so far in healthcare. You've written quite a bit about that, Danielle.

Ofri: When the EMR first came, it was a miracle. And it still is. I tend to write a lot about the faults, but I well recognize the benefits and I don't want to go back, and I think no one's going back. The old days of trying to find the chart that was stuck in the dermatology clinic or the x-ray that was in the surgeon's back pocket were terrible for patient care, and I think the EMR has helped in that respect. However, you're right; it was really arranged for billing, and that is problematic in a number of realms.

One is that it fragments how we think. We've just switched to a new medical record system so we're kind of learning a new system from scratch. It's incredibly complex, which to me reflects how complex medical thinking is. The paragraph you wrote in the chart by hand was the result of innumerable thought processes that will intersect: the history, physical, labs, your sixth sense for the patient, the patient's nonverbal communication. All of that came together and you had a differential in your head. With the EMR, you have to do that in a very rote way, in 100 different places. So replicating the doctor's thought process is quite a complex task, and the EMR is not up to standard for that. We don't think that way, with the labs here, the consults there, etc. We put them all together seamlessly. Fragmentation is one problem.

The second problem is that it dominates the clinician's attention. There is no way to focus on the patient and the EMR at the same time. But if you don't write in the EMR while you see the patient, you will never finish. It is not humanly possible. So most doctors are now staring at a screen. It's like the 1950s secretarial pool, the electronic indentured servitude now, and it's really impacted the ability to communicate with the patient and hear what they are saying.

Communicating With Patients

Verghese: So much of what is written in textbooks just assumes that everybody is on the same page and the patient will deliver their history and we will record it. Danielle, you have written beautifully about the challenge of taking care of people who speak a language other than English. At Bellevue you have a unique opportunity to reflect on what that is like. Talk a little bit about how that magnifies all of the issues we're talking about in dealing with people who may not have the language that you have.

Ofri: This morning I had a patient in clinic who spoke Sylheti, a dialect of Bangladeshi, which before a year or two ago I didn't even know existed. I realized that many of my patients who I thought spoke Bengali or Bangladeshi actually spoke Sylheti, and using a Bangladesh interpreter was quite limiting, and so I realized that this was an error for many of my patients. Every layer of barrier between communication certainly changes things. An interpreter boils things down to the most basic parts of the conversation—the actual facts—and you miss so much nuance. It also takes twice as long, so I always worry that I'm missing much of what is really going on with patients.

The other aspect is education and literacy level. I had a patient with diabetes who had terrible medication "noncompliance," and every time he came in I would spend an hour reorienting his medications. I would print out this gorgeous medication list and feel so proud of myself. It took about a year or two until he confessed to me that he could not read. Here I am giving him this list that he couldn't read at all—it could have been wallpapered on his wall—thinking I'm such a great doctor. But I could never find out, because we were not speaking English, that he literally could not read and that was why he kept mixing up the medication.

There are so many components that go into it. You have the patient sign up for MyChart and do all their stuff online. Well, a lot of patients either don't have access or can't navigate that type of thing. Maybe they can't make the call to arrange the radiology appointment because it's just too much.

Verghese: Sometimes, sitting in Silicon Valley and listening to the next best thing in healthcare that somebody is trying to dream up, I feel that they often lose sight of the reality of someone like you on the frontline in that kind of situation. I'm sure that in all of the things that you do, your humanity comes through to the patient, and the relationship you build must mean a great deal even with a different language and interpreter.

Ofri: That is probably the best thing about medicine, and I think the joy still comes through from both the patients and the doctors. Many patients love coming to the doctor. You usually think of going to the doctor as a terrible thing, but I find many patients very upbeat. They will dress well for it, bring a gift, prepare for it, and I think for many, they get an attention to the issues of their life that they don't get elsewhere.

Often it's just an issue of respect. Many of our patients, I sense, get disrespected in many aspects of their lives, especially patients who are immigrants or in marginalized communities—transgender, undocumented, or who are facing poverty. Many get treated terribly in so many situations, and this is their one chance. They can come to their doctor who says, "I'm here to help you; what can I do?" That is a rare thing for them. It's so nice to be able to bring some joy into a patient's life, and you might be able to make them feel a little bit better. Maybe you can't cure their condition but you can move the needle a few degrees, and what is more amazing than that?

Topol: You've outlined more of an ideal scenario, but from other points that we discussed, so many patients feel that the time they have with their doctors is so limited and they are not having eye-to-eye contact because of the screen and keyboard. There is a lot of dissatisfaction—the sense of being roughed up with a lack of attention. We hope that we'll be able to restore what you just described for all of the patients in the future.

The physical exam remains important for diagnosis but also for the second round of the history.

Ofri: I now use the physical exam in a different way. I still use it for its physical findings, but more and more I use it as a refuge from the interference of technology, because the physical exam is one of the few places where we can talk unimpeded by a screen or a recorder or a keyboard. It's just two people talking and touching with nothing else in between, and that is so rare in medicine and in life in general. So often it is during the physical exam that the patient says what is really on their mind, so I use the physical exam as my second shot at a good history. That is when the patient will talk about domestic violence, depression, sexual symptoms, the cough they forgot about until I auscultated their lungs. The physical exam remains important for diagnosis but also for the second round of the history.

Verghese: That is well said and you won't find me arguing with you.

Topol: Danielle, I think the average time, when you take a history, is no longer than 18 seconds before the patient is interrupted.

Ofri: At best. It's very hard, and I talk about this a lot when I speak with medical students and even with the general public, who are appalled when they hear that statistic and think, "How rude those doctors are." I try to explain that we're not rude, or at least most of us, but that we want to find the answer. The patient says, "I have a pain over here" and we jump right in. "When does it start, when does it stop, what it makes better, what makes it worse?" [We jump in] not because we're rude but because we want to help them feel better. Of course, if they had a second thing to say, like, "I think I might have had a stroke last week," we'll miss that.

When you think about diagnostic error, which I think is among the hardest aspects of medical error to fix, it's right there in those first 18 seconds—when we jump in and divert the conversation and not get to point two—that I think we're making our biggest error. One of the things I try to tell my students and myself is that for the first 1 minute, don't say a thing, don't write a thing. Just look at the patient. Provide full frontal listening for 1 minute. It's not a lot of time but it feels like a lot, and most patients will get out what they need to say in that 1 minute. Then I'll say something like, "I don't want to miss what you're saying. Do you mind if I take notes while you speak?" and I pull out the computer and they are forgiving. They understand that we have to do that. That 1 minute is a huge investment in trust but also in improving your diagnostic accuracy.

Time for Diagnosis

Topol: That is great. Here is another thing about time, the early time with a patient. It's thought that if you think of the diagnosis in the first 5 minutes of evaluating a new patient, and it could be quite a complex patient, the chance of being correct is 95%-plus. But if you don't think of what turns out to be the right diagnosis, the chance of error is exceptionally large, so it's very limited time. It's reflexive thinking, System 1 thinking, to get the answer; otherwise the chance of error is as high as 70%. Do you have any thoughts on that? You have a book coming out next spring on error. How can we get the error rate down in diagnosis?

Ofri: When we're medical students, we write that whole exhaustive differential diagnosis. And when you read those medical student write-ups, they are torturous because they have a huge differential for every organ system and every possible etiology. Of course, we can't do that in real practice, but I think, as Mark Graber [founder of the Society to Improve Diagnosis in Medicine] would say, the discipline of the differential is to at least ask yourself, "Is there anything else it could be?" Push yourself. Is there anything else it could be and is there anything I can't afford to miss? Try to be rigorous about it. Questioning yourself even for 20 seconds will help with the diagnostic error.

We often never know if we're right or wrong. Patients don't come back and we assume that they got better, but maybe they died or maybe they went to the next hospital up the block and got a different diagnosis. We try to get some feedback if we can. The other thing is being comfortable with uncertainty. We are so lousy with ambiguity. We rail against it because we're programmed to be certain and confident, but you sort of make peace with uncertainty and say, "I don't know what it is. It could be one of these three things, so here's what we'll do and here's how we'll watch it." Be comfortable with not having an answer and helping your patient be comfortable with uncertainty, because that is the reality in general outpatient medicine. We often don't know, we often never know, and that is just the way medicine is.

We've been trained to be perfect, and anything less than perfect is complete failure.

Verghese: Going hand in hand with that, Danielle, you spoke very eloquently at the TEDMED meeting about our needing to recognize that we can fail and be willing to fail and to admit our failures—that a lot of our errors come from sort of a dogged desire to not lose face, so to speak.

Ofri: We've been trained to be perfect, and anything less than perfect is complete failure. If you're not perfect, go work for a drug company or go off to the dark side. But if you're going to make clinical medicine, you should be up there—that is the way we pimp on rounds. And that is really terrible because it then encourages us to lie, to make up what we don’t know. "Fake it till we make it." That is terrible for a patient. We should never fake it till we make it. We should say to the patient, "I don't know" or "What I first said was wrong. I was wrong." In studies that have at least given patients hypothetical scenarios, patients are much more likely to stick with a doctor who tells them they have made a mistake. They think, "Okay, I can trust that doctor; I know they'll be honest." I think patients are very willing to hear it if you tell it honestly, because they want to be able to trust you and they are worried about not knowing what's going on. Admitting an error generally instills the patient's faith in you rather than the opposite, which we usually assume.

Topol: We will have to pause it there for the moment, but we'll continue our discussion with Danielle Ofri in our next episode of "Medicine and the Machine." In the meantime, thank you both, Danielle and Abraham, for yet another enthralling discussion.

And to our audience members, please join us for the second half of our discussion with Danielle in episode 4. We'll talk about many other topics that are integral to the patient-doctor relationship, including errors in the medical record and administrative creep.

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.