[00:00:00] Christine Ko: Welcome back to SEE HEAR FEEL. Today I have the honor of being with Dr. Art Papier. Dr. Art Papier, MD, is Chief Executive Officer and Co Founder of VisualDx. He is a dermatologist, medical informatics expert, and Associate Professor of Dermatology and Medical Informatics at the University of Rochester School of Medicine and Dentistry. Dr. Papier graduated from Wesleyan University, the University of Vermont College of Medicine, and completed graduate medical training at the University of Rochester Medical Center. His interests are wide and span healthcare costs as related to clinical accuracy, clinical decision support systems, diagnostic error reduction, cognitive error, medical education, and empowering patients. So all things that I really love. I'm really excited to talk to him. I probably won't have the chance to ask him everything that I would want to ask. Thank you to Art for joining me today.
[00:00:59] Art Papier: Christine, it's a pleasure to join you.
[00:01:01] Christine Ko: Thank you. Could you first share a personal anecdote?
[00:01:04] Art Papier: I've been doing a lot of talking about dogs lately. I'm a dog lover. There's this wonderful place you can Google called Dog Mountain, which is the site of where well known Vermont folk artists lived. They had a 130 acre farm and a print studio employing artists there to create all this imagery that had dogs as a theme. Eventually the property became a non profit. The goal is to maintain the artwork and this beautiful piece of land. You can see online all this beautiful artwork and even buy a t shirt that has some of these wonderful images.
[00:01:43] Christine Ko: That's cool. I assume you have a dog at least, or multiple.
[00:01:49] Art Papier: Yeah, I have Hope, an eight year old yellow lab that keeps everybody calm that's near her.
[00:01:54] Christine Ko: I recently learned that being calm, um, sort of actually maybe a more "negative" state may be good for decision making. It helps us think more carefully. Related to that, can you talk about cognitive error?
[00:02:07] Art Papier: Sure. I am a dermatologist. And I realized that dermatologists really don't think about problem solving the way that internists do. There's this wonderful book called Thinking, Fast and Slow by Daniel Kahneman. And that's influenced me. And story really starts with meeting Dr. Larry Weed, who invented the problem oriented record in the soap note format 50 years ago, when I met Larry many years ago, he started talking about cognitive mistakes to me, like premature closure, anchoring bias, and at the time I didn't realize there's a whole field of this. And I don't think there's enough training in this for medical students at all, about how do you think. So when you mention, maybe you're in a sad, affective state, you're more careful in your thinking. I didn't realize that, but that sounds really interesting.
[00:03:04] One of the great problems right now is how burned out physicians are and how much stress they're under. And when we put our shoes into an emergency physician's shoes in a crowded, noisy, busy emergency department with 20 patients boarding in the hallway because there's not enough beds, and a six hour wait in the waiting room to get in to be seen. Basically you're asking a human being to see a different problem every 15 minutes, say in that ER, or it can be in the primary care office too. And that person can come in with one of 500 different complaints and then the clinician needs to ask the right questions to do the problem solving of analyzing the problem.
[00:03:55] That's rife for cognitive mistakes. And so I'm very interested in this area. And of course, dermatologists do pattern recognition and in Thinking, Fast and Slow, that's explained as System 1. And then your analytical skills using your cortex is System 2. And, Dr. Kahneman makes the argument that we essentially as humans are lazy, that we want to make quick decisions. We don't really want to think about things cause it uses a lot of glucose. And, from an evolutionary point of view, it's much better to just make these snap decisions. The snap decisions are rife with error.
[00:04:38] And the other reference besides Thinking, Fast and Slow is some of the papers of Dr. Pat Croskerry, an ER physician and psychologist. There's a whole body of literature around this. But, what I had spent a lot of time thinking about is the cognitive bias of representative bias. Representative bias is really teaching to the classic. And then the patient comes in as a variant.
[00:05:05] Christine Ko: Yeah.
[00:05:06] Art Papier: Dr. Weed, to his credit, was talking about these cognitive biases 50 years ago, and that it's a total myth that a human can see a different person every 15 minutes, ask the right questions, know what questions to ask, and then process that all and deliver highly reliable care.
[00:05:27] Christine Ko: Yes.
[00:05:27] Art Papier: So from the patient perspective, you expect your healthcare professional to deliver highly reliable care. Yet, let's step back and think about the environment. The environment is waiting in the ER wait room, and physicians under pressure to go faster with less resources. And of course, if it's a private equity company running the hospital or running the practice, then there's even more pressure because they've often understaffed the clinical environment. So that's a setup for failure, right? And then the other thing that we're missing is feedback loops. Busy doctors are unaware of their failures. We haven't set up a learning system. So there's so much opportunity. in medicine for the next generation to work on these problems. And unfortunately, I think we're working on the wrong problems.
[00:06:20] Christine Ko: What are the right problems?
[00:06:21] Art Papier: The right problems is this: people come in and have problems and they need help from us. And how do we deliver the best possible care? The myth is that we can run all this information through our brains and deliver highly reliable care. And we can do that for 80 percent of the patients, common things happen commonly, and the common presentation of the common happens the most commonly. But the next level of error, and this is the representative bias, is the variation of the common.
[00:06:54] Christine Ko: Yeah.
[00:06:55] Art Papier: Missing the variance of the common is the bulk of the error. When you study diagnostic errors, it's missed myocardial infarction, it's missed appendicitis, it's missed breast cancer, lung cancer, colon cancer, prostate cancer, missed aneurysm. Every disease I just mentioned, anyone that watches a TV show has heard those diagnoses. They're not rare diagnoses, but the patient with the MI doesn't come in with crushing chest pain, and pain down their left arm, diaphoresis and fatigue, like a classic MI patient. The patient comes in with jaw pain, a toothache, is female; all the things that would cause cognitive biases. So we teach to the classic, and the patient comes in as a variant. And that's really been the focus of my career, thinking about how do we move medical education away from teaching to the classic? And how do we teach to using information tools? in concert with our brain.
[00:07:55] Christine Ko: I love it. Augmented intelligence.
[00:07:57] Art Papier: Exactly. Augmented intelligence. Exactly.
[00:08:00] Christine Ko: So I think you're touching on that, with the right question being, how do we really reduce diagnostic error when, physicians, not just in the ER, I think more and more, we have less time in the clinic even to spend with patients. I think you're touching on maybe that your company VisualDx is trying to get at that representativeness and teaching people ahead of time.
[00:08:26] Art Papier: There's an educational component, but it's not just education. It's actually using technology. I don't know if you're aware, but we have a machine learning component to VisualDx. So a non dermatologist can take a photo of the rash with their iPhone and be guided to a differential diagnosis, a very intelligent search for them, which is way better than them going to Google and typing in what they think it is. So we've made a lot of progress. The other point is there are things that have broken for years. Let me ask you, Christine, have you ever seen a patient admitted to the hospital for bilateral cellulitis? You ask any dermatologist that question, they all say, yeah. That's a diagnostic oxymoron. And we're giving that person IV antibiotics, putting them at risk for C. difficile, and putting them at risk for a serious drug reaction. When they don't need to be admitted, and they don't need antibiotics. That problem has been going on since I was a resident. So that's 30 years. It still goes on. We've made this wonderful progress. In medicine, we have great therapies for HIV now that we didn't have 30 years ago. We have wonderful treatments for Hepatitis C, and the list goes on and on. But today, in this country, people are being harmed by being admitted for cellulitis, I guarantee it.
[00:09:50] Christine Ko: Yeah.
[00:09:50] Art Papier: And so we can't fix some of the lower hanging fruit because we're focused on the wrong things.
[00:09:55] Christine Ko: Yes. So for the non dermatologists out there, lipodermatosclerosis or you could think of really bad
[00:10:01] Art Papier: or stasis dermatitis.
[00:10:02] Christine Ko: Stasis dermatitis.
[00:10:03] Art Papier: There's all these cognitive mistakes. There's premature closure at the red leg. Instantly, it must be cellulitis because that leg is red, it's swollen, it's dripping because they're vesicles or whatever, and there's this immediate jump to, oh, it must be cellulitis. Dermatologists have that expertise to look at and say stasis dermatitis or lipodermatosclerosis.
[00:10:26] Christine Ko: Yes.
[00:10:26] Art Papier: And so how do we use information to reduce that error? That's what we focus on at VisualDx. Things like that. And that's just one disease. There's a half a million admissions in America for cellulitis every year; 140, 000 or so, or 160, 000 are wrong, in error. And then you think about fixing that, and how much less harm you'll cause, and how much money you'll save.
[00:10:55] Christine Ko: Yeah, do you think that VisualDx has helped move the needle away from making that error?
[00:11:02] Art Papier: We have this incredible study that University of Maryland ran. A randomized control study where emergency physicians going to admit a patient for cellulitis, and in Epic they got a best practice advisory that would ask just a few questions about, is it on a leg, arm, whatever, that you're suspecting cellulitis? Is it unilateral, bilateral? Is there edema? Is there scale? Is there itch? Just a few questions, and it would give them a differential. And of course, if they said bilateral, cellulitis would not be in the differential. They showed that they reduced unnecessary admissions by 38 percent. But so many people are unaware of this. Because we're focused on the wrong things.
[00:11:48] Christine Ko: Yeah. How do we get focused on the right things, though?
[00:11:50] Art Papier: With your help, with this podcast, maybe somebody's listening to the podcast, that will get in touch, and we'll work with them to peel back the layers of the onion here and really work at an operational level in the health care system. This is not just an individual problem where we just train doctors to be smarter. It's where the priorities are set in the health care system.
[00:12:17] Christine Ko: Yeah. How did the University of Maryland get that trigger in Epic? It was just an institutional...?
[00:12:23] Art Papier: They have some really talented emergency physicians who work in medical informatics. There's a doctor, Dan Lemkin, who knows Epic inside and out. He's an ER doc that's interested in these kind of challenges. And I don't know exactly what they did on their side, but it's not a lot of work. We embed VisualDx into EPIC, but certainly, I would say, a majority of ,dermatologists are unaware that VisualDx integrates into Epic.
[00:12:53] Christine Ko: Yes, I was not aware of that.
[00:12:55] Art Papier: Yeah.
[00:12:55] Christine Ko: So, that's important. Do you have any final thoughts?
[00:13:00] Art Papier: Yeah, I'm really interested in better information for people at home too. We're fascinated by ChatGPT and large language models at VisualDX, trying to figure out how we use the wonderful potential of the large language model, particularly with the way it uses natural language search to get at great content, right?
[00:13:23] The challenge with ChatGPT and some of the large language models is the hallucinations and not really knowing how you arrived at what you arrived at in the model. It's a black box. And to me, the piece that we need is the map. The analogy I use is imagine you have a GPS for your phone, and it's directing you to go somewhere, and the GPS doesn't have a map, and it's telling you to drive a hundred miles on US80, and then take a left at a certain exit, drive another 20 miles, and it's just directing you with verbal instructions, but there's no map. If it was a hundred percent reliable, you'd use it all the time. If you knew it worked, but it was only 90% reliable, you might drive outta your way 10% at a time. You'd want to know where am I, in a less reliable system.
[00:14:20] And what we're thinking about at VisualDx is this visualization of medical information. How do we create the map of medicine so you know where you are when you use these tools? So when you use large language models, and you use augmented intelligence. How do we create information technology that guides and teaches as you use it? Because I'm not a believer that we're getting rid of doctors and replacing it with computers. It's the healthcare professional that's going to use these tools. Patients are going to use these tools. And these tools have to make us smarter as we use it. Otherwise, it's going to de skill us. It's going to make us stupid, where people lose their skills and those skills are going to be needed at times. So, that's the story.
[00:15:10] Christine Ko: Thank you. Thank you so much for spending this time with me. I appreciate it.
[00:15:14] Art Papier: Christine, I really appreciate it as well.