Spend Money on AI or Just Give Doctors More Time?

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


September 04, 2019

This transcript has been edited for clarity.

Eric J. Topol, MD: Hello. This is Eric Topol. Welcome to our fourth episode of "Medicine and the Machine." Today we're continuing our discussion with Danielle Ofri and my cohost, Abraham Verghese. In the previous episode we had an enthralling discussion with Danielle about the gift of time and whether [artificial intelligence (AI)] could help improve the quality of time that physicians spend with patients. In this, the second segment of our discussion, we get into many other topics that are integral to the patient-doctor relationship, including errors in the medical record and administrative creep.

Welcome again, Danielle and Abraham.

I want to get back to the electronic medical record (EMR), one of our favorite subjects, to perhaps rail upon or at least discuss how to make it better. In charts today, 80% of the notes that we look at are cut and pasted from prior notes which are error laden, and patients never have a chance to edit that. I'm sure you have looked at your own notes as a patient and found mistakes. How do we get that on track and how much is that contributing to the problems and errors that you delve into?

Danielle Ofri, MD, PhD: It's horrific. I was seeing a patient today who sees a rheumatologist. Rheumatologists, not to cast aspersion, have a habit of keeping a running tally of the disease, and they copy and paste the past 20 visits' worth of what the x-rays show and the treatments. Now the note is 10 pages long and, of course, I skip it all and I want to get to the last line. I want to know what is happening now and where we are. In fact, I had a patient recently who had a diagnosis of rheumatoid arthritis (RA) in the chart, and it dawned on me one day that I've known him for a couple of years but I've never seen any manifestations of RA. I did what we inevitably call a diagnostic timeout and went and hunted back through the medical records. Of course, it took me an hour, but I don't think he ever had RA. He had a positive rheumatoid factor somewhere a thousand years ago and someone wrote down "RA," and it got kept in the chart like a liturgy and it never got out. That happens a lot—these diagnoses accrue and we just faithfully copy them. So I'm sure there is quite a bit of medical error.

The problem is, who has time to go back and prune them all? We hear that patients may edit their charts, but can all patients really do that? Some patients probably have the intellectual wherewithal, but I think many patients would struggle with the terminology and not know what to edit.

Topol: Right. Some of these things we're talking about perhaps will lend themselves well to machine learning, to be more accurate from processing the data that is in the real record rather than some of the things that don't belong there.

Ofri: But who edits the machine? Who makes sure the machine is not making a mistake?

Topol: Right, these are the challenges that lie ahead. There is promise, but a lot of proof points to that we are not there yet. It also goes back to the conversation that you have with the patient, whether it's the actual time of the history or, as you eluded to, the second history while you're doing the physical exam. It's got a lot of valuable information in it, and now there are many companies, including the likes of Microsoft, Google, and some 25 startups, that are taking that conversation and synthesizing a note which has all of the critical components in it. It's starting to get a lot of acceptance in the United Kingdom and China and in limited clinics piloting it in the United States. Do you think that that type of synthetic note from conversation, with elimination of keyboard, would have value?

Ofri: I have thought about that. Is it going to transcribe our schmoozing? As we approach depression by talking about what is going on with your family back home, can it really [filter through that]? It has potential; I just need to see it in practice. Maybe doing the actual nuts-and-bolts of the history, review of systems, all that stuff, I could see. But in actual conversations, will it know how to filter out the chit chat from the [important content]?

Topol: From what I've seen, it learns from each doctor, so it would have to understand the schmooze stuff and ex that out. It learns with some number, like 40 or 50, of these notes per doctor, and then the patient edits their side. It shifts the burden to machines and to patients to some degree, and the machine learns for doctors. We will see. This is a promise that lies ahead that is unfulfilled but has a chance.

Ofri: But it seems that the doctor has to go back and look at what the machine has done with the note, and then correct that to make it clear before the patient sees it and might see things that could be construed as insulting or negative.

Topol: The way it's worked so far is that there is only a limited number of notes that the machine looks at to understand what should go in the note. It tries to find the patterns. The note goes to the patient, and then it is signed off on ultimately by the doctor. It's a work in progress; it's a theoretical thing.

Ofri: Who is liable for mistakes?

Topol: Mistakes? Mistakes—what are those?

Ofri: When the machine puts something in and something goes wrong, does the doctor get sued? Because you are right—the doctor will sign off. Has the doctor had enough time to reread every word of every note when they sign? No. We have to sign off on them kind of in batch, but somewhere along the line something will happen, and who is going to be up on the stand? The computer? The developer from Silicon Valley? I don't think so.

Digitizing the Physical Exam

Topol: Recently there has been a movement for more patients to record their whole visit, and that is part of this as well—that the whole thing is archived. You can listen to the entire interview or conversation during a visit. This is in play; it's not settled yet at all.

Abraham Verghese, MD: We have the capacity to so easily record what actually happened in the room with our state-of-the art cameras and to translate that into an accurate description of the physical exam. Right now, everybody does these perfect physical exams theoretically and in practice they probably don't. I think there is some way to make us more honest about what we're actually doing with the patient and then record it legitimately. What do you know in that area, Eric, in terms of latest developments, and Danielle, what are your thoughts?

Topol: You bring up an interesting point. Kind of like a body cam on the police officer for the victim—I don't know if you want to go there. You have written quite a bit about this issue, the cursory physical exam (or even less than cursory). It's what I always refer to as a WNL—not for "within normal limits" but for "we never looked." This is a serious problem. We will never be able to digitize a patient's life story and nor should we try. That is different from having a conversation in the visit. I don't know if we want to try to digitize the exam. Danielle, what are your thoughts?

Ofri: If I had 45 minutes per visit [as we talked about happening at Geisinger Health], I could do it all for most patients. I could do a history, a reasonable physical; and write a cogent, coherent, concise note. But I'm trying to fit that into 15 minutes and that is not possible. So maybe rather than doing all of this jumbling of machinations that cost gazillions of dollars, spend the money on letting us have adequate time for our patients and we would do a better job.

Verghese: We're working with a medical ontology fellow. We have all of this precision when it comes to diagnostic codes—you break your left foot in a mobile home while cleaning the toilet, and there's a nice code for that. But the best we can do about a patient is say that they are a 63-year-old male with a history of hypertension. But this gives no sense of where they are on the arc of their life and who they are. Bringing in ontology with that type of precision to the patient and the substrate of their life is a worthwhile endeavor; it's not to digitize them but to better capture variation between patients.

Topol: I totally agree, and as Danielle pointed out in the New England Journal of Medicine essay,[1] there's no CPT code for contemplation. But what I was getting at is the patient's life story, which I see as something different: what the person is—not just what the person is saying, but their nonverbal expression. What really worries them? What is going on in their life that might not come up in a short visit? To try to really understand and get in the head of a person and know their environment. Trying to come up with an ontology of coding for the real story that is them would be hard to do.

Ofri: What about a little extra time so I can free-text a social history and write a paragraph or two that captures the aspects of their life that are crucial to their current pressures? Where they are from, how they got here, who they live with. It doesn't have to be a novel and I don't need a code for it; I just need enough time to ask it and to write it. In 45 minutes I can probably do a reasonable job.

Topol: I think you could; you're a gifted writer.

Ofri: Thank you, but I think any good internist who wants to ask about their patient can write a few lines about the crucial issues—if they had the time. Right now everyone is just struggling to keep up with the basics. No one has time to document.

'Administrator Creep'

Topol: That is a really good point, and that is why in this podcast we've been talking about trying to get the gift of time from various ways of tapping into the promise of AI. One other thing I wanted to discuss is the "administrator creep," as you've written about and which I think is widely known among physicians. Over the past 35 years, there has been an over 3200% increase in administrators and managers who don't see patients, and there has been a relatively constant 40% or 50% increase over those same 35 years in physicians—a couple of orders of magnitude difference—in nonpatient care personnel.

Ofri: It used to be a ratio of about 1:1 and now it's about 10:1 of nonclinicians per doctor.

Topol: What do you make of that and what should be done about it?

Ofri: When I pointed that out in my paper, I got a few aggrieved letters from administrators who talked about how busy they are and how much work they have, and how much they are helping patients. I certainly recognize that many regulations have been put into place that require administrators to attend to them, and I'm glad they are there because I don't want to do it and we need that. But I feel like the resources at some point are inadequately allocated. The first thing we need is enough nurses and doctors on the floor or in the clinic to take care of the patients, and as long as we're so short-staffed of that, we really have to divert the resources clinically, and at some point we have to prioritize.

I don't want to call it quite rent-seeking behavior, but there seems to be a lot of people. I don't know what they are doing, but it is not moving the needle. I just want to give vaccinations, not write a report. We've just gotten to the point of absurdity where there are so many more administrators than clinicians, and we haven't fully staffed the clinical side yet.

Topol: Abraham, what about you?

Verghese: I marvel at the amount of administration it does take to keep abreast of all the various regulations and all the various payers. It's an onerous task for anybody. I can't profess to any understanding of what is happening up in the executive suite. I know that they are doing a lot of heavy lifting, whatever it is.

Ofri: In Canada, the ratio is about 3:1, so somehow they manage to give quite good medical care, certainly reasonably comparable, with many fewer administrators. I think a lot of it has to do with our patchwork of health insurance coverage in the United States. You need administrators because there are 25 different insurance companies with different regulations and the like. But you can do it with fewer administrators.

Topol: That is another question about medicine and the machine: Can some of these folks be replaced by machines, or at least part of their efforts? Abraham described the heavy lifting. You have coders and all of these back-office operation people. You would think that much of that could be picked up by machines, but I guess we'll have to see.

Forget the graham crackers.

Ofri: Don't unemploy them; retrain them as phlebotomists and nurses and certified nursing assistants. We need the people. Now that patient satisfaction is a big factor in reimbursement, it matters to administrators, but their response is usually, "Let's get valet parking, or a nicer coffee machine, or graham crackers in the waiting room." To me that is absurd. Ask the patient what they want. They want to not wait so long for the doctor, enough nurses on the floor, and the phone answered when they call. Do those things and give them 40 minutes with their doctor instead of 20, and you will have more satisfied patients. Forget the graham crackers.

Verghese: I think one of the big challenges for anybody in administration, and I really feel for them, is trying to predict the future. There is so much uncertainty out there. They are dealing with huge budgets, and the slightest error in projection can just send them right into the red. I think it's the instability of healthcare reimbursement in general and the very uncertain lobbying efforts that undo things that seem reasonable. It's a very difficult task for anybody.

Ofri: Right, but some things are pretty constant. Illness is pretty constant—it never goes out of style. No matter what is in and out these days, everyone has got their chronic illnesses. We're still going to have diabetes and hypertension and obesity; those are not going away. We can at least predict that we still have a lot of chronically ill patients and they need to see their doctors and nurses. That much I think we know is a constant, and we should be able to work on that.

The idea of giving more time to a doctor is a financial decision. If you want to double the amount of time, you can just have twice as many doctors and I think we could pull it off. I don't know how much it costs for a hospital to switch to a new system...gazillions of dollars. We could have hired a lot of doctors and kept our other system which was okay, but twice as many doctors would have had a huge impact on medical care.

Verghese: But then you're up against [other issues]. For example, in psychiatry, a 15-minute medication visit is reimbursed at a higher level than a 40-minute psychotherapeutic encounter, so no one is doing any of those.

Ofri: Right, but that is our society and that is a decision we have to make. We have to decide what is of value, and our profession has really weaseled out of this. We let the standards be set a half-century ago, and who was at the table? It wasn't therapists and primary care doctors; it was mainly specialists and procedure-oriented people, and that is why procedures are reimbursed so much more.

What is diabetes management? It is mainly education and talking with your patients about how to take your medications and how to eat. If I talk with a patient about how to cook brown rice instead of white rice, which I do about six times a day, I'm probably reimbursed $40 on a good day. But if, while I'm talking to the patient about how to cook the brown rice and broccoli, I simultaneously thread a tube into one of their orifices—and you can pick any of their orifices you want—that reimbursement goes up by 10-fold. And if, while I'm threading a tube through their nostrils and talking about kale, I run a CT scan over them, now you are talking big bucks, and that keeps the hospital afloat.

But what do we need for diabetes management? We need a doctor or nutritionist to talk to a patient for 45 minutes about how to manage their disease. Our reimbursement scheme is really kind of flipped on its head.

Topol: That goes to corporatization and also what contributed to those relative value units. The theme of a lot of what we talked about today is time. Could you ever foresee that radio, TV, or billboard ads for health systems would say, "We give patients time"?

Ofri: I try not to watch TV at all.

Topol: Do you think there will be a time when [places will advertise that] doctors will spend twice as much time with you?

Ofri: If you gave patients and doctors a taste of that and administrators a taste of patient satisfaction with that, I bet it would.

Time as a Factor of Burnout

Ofri: If you look back at this whole epidemic of "doctor burnout," I think we've mischaracterized it. Most doctors aren't burned out from medicine; most of them love medicine and would not want to trade it for anything else. All they would love to do more than anything in the world is have time with their patients to take care of them in the way they think is clinically appropriate. But they have so little time, and the burnout is really all of the extrinsic stuff. I think most doctors would be thrilled to have twice as much time with their patients.

Topol: I agree with that diagnosis about what is the real cause of burnout. What do you think, Abraham?

Verghese: Completely. People want to spend time with their patients, and one of the reasons we call the center Presence that we run here at Stanford is because that word kept coming up—it's the one thing that patients said they wanted more of. They wanted their doctor to be present and not distracted by the medical record, and conversely, physicians were saying, "We're doing everything but being present with the patient."

I don't want to take Pilates; I need an assistant.

Residents complain that they are often in a bunker looking at a screen, and that is not where they want to be. We put them there. That is our doing. It happened on our watch that they wound up spending so many hours there, so I completely agree.

Topol: My hope is that the whole medical community, led perhaps by doctors, will stand up for time, stand up for their patients, and get their burnout rates markedly diminished, because otherwise we're going down a spiral here with peaking burnout, depression, and suicide. I don't know what else can turn this around. Do you have any other remedies in mind, Danielle?

Ofri: It's also the patients taking a stand as well. I just got an email from a woman who wrote a piece for STAT, who has seen the effect of her doctors who are so depleted and unable to handle their load, that it's affecting her. She does not want her doctors being drained down by the EMR. I think patients recognize that it's not working in their interest. They don't want their doctors depressed and suicidal; they want their doctors to be present and to be there as much as the doctors and nurses want to be. Maybe we need to have an alliance with our patients that we all go on strike. What if we didn't code for 1 day all across the nation and the patients simply refuse to fill out their forms for 1 day? Maybe some people would pay attention.

Verghese: I think we see that the traction that has come with burnout has actually made an impact on the way administrators have reacted to the EMR by not adding more keystrokes for us. I think it took burnout and attrition and [clinicians] getting recruited and leaving for people to realize that this is a very costly business. I don't think our suffering alone is enough, but there are dollars involved when people leave medicine, so I think we finally have attention on this topic from the powers that be. But clearly it's not enough.

Ofri: I'm not quite so sure. I feel like this whole emphasis on resilience and well-being [is unfair]. Take a mindfulness class or meditate. I don't want to take Pilates; I need an assistant. The idea that it's the doctor's fault for not being resilient enough to handle all of this—that is really not the case. Doctors are the most resilient people I know. They survive and take pretty good care of their patients in the current environment, so they are quite resilient. It's an unfair reversal.

Topol: I agree that there is resilience, but so many people are leaving the profession and sincerely telling their kids, "Whatever you do, don't go into medicine." These are not good signs. We'd love to see that turned around.

What a privilege it was to have you on board with us, Danielle. It was really fantastic. I'm going to turn things over to Abraham to wrap things up on this "Medicine and the Machine" podcast.

Verghese: I will indeed. Danielle, you have been such a wonderful and important voice in medicine. One of the burning needs in medicine and science in general is that of a commentator, someone on the inside who takes the time to reflect, takes the time to ask the difficult questions, and voices the angst of the profession. You have done that so beautifully for many years now, and we hope that there is much more coming. Thank you so much for being on this podcast.

Ofri: It's my pleasure, and I would also like to give a shout-out to your writing, Abraham, and also yours, Eric, because you bring new ideas to life through fiction and nonfiction. As you always say, "Fiction is the great lie that tells the truth."

Verghese: Oh yeah.

Topol: Thank you.

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.