April 19, 2024: In this episode, Dr. CT Lin, CMIO of UC Health, engages in a thought-provoking discussion about the integration of generative AI in healthcare workflows, sharing insights from his conversations with industry experts. He explores the nuances of clinicians' perceptions of AI-generated notes, the potential impact on patient care, and the delicate balance between automation and human expertise. CT delves into the challenges of customization, the importance of teamwork in optimizing EHR usage, and strategies for mitigating clinician burnout. As the conversation unfolds, listeners are prompted to consider the evolving role of technology in healthcare and its implications for clinician well-being and patient outcomes. How can AI enhance efficiency without compromising quality? What strategies can healthcare organizations implement to foster effective collaboration and innovation? And how can clinicians navigate the complexities of automation while preserving their expertise and autonomy?
Key Points:
Donate: Alex’s Lemonade Stand: Foundation for Childhood Cancer
This transcription is provided by artificial intelligence. We believe in technology but understand that even the smartest robots can sometimes get speech recognition wrong.
Today on Keynote
(Intro) my strategy is. Protect the wellness of the clinicians. Most clinicians come to practice trying to do the right thing for patients.
And so it's these mountains of optimizing patient care and optimizing teamwork so that we're as efficient as possible.
My name is Bill Russell. I'm a former CIO for a 16 hospital system and creator of This Week Health, where we are dedicated to transforming healthcare one connection at a time. Our keynote show is designed to share conference level value with you every week.
Today's episode is sponsored by Quantum Health, Gordian, Doctor First, Gozio Health, Artisight, Zscaler, Nuance, CDW, and Airwaves
Now, let's jump right into the episode.
(Main) , all right, here we are. It is Keynote and today we're joined by Dr. C. T. Lin, CMIO for UC Health, UC, University of Colorado Health System. Dr. Lin, welcome to the show. Hi, thanks for having me.
Yeah, I'm looking forward to our conversation. We caught up once before at a Chimefall Forum, I think it was. And we had pretty wide ranging discussion. I remember when we got done that. I looked at my producer and I said make sure you have him back on the show because it was Anytime I learned some new things.
I think this is this could be fun so let's start with a very basic question. What's a typical day? For a cmio These days.
There is no typical day, is there? There is
no typical, right? You just take the day as it comes. Unfortunately, I have recently set the record for sitting on my butt for 12 hours in a row and having about 15 meetings. Just one issue after another. And it's fun. And also terrible to have this whiplash from some meetings are about, Doc, if you turn the computer on, I'll show you some efficiencies that come from using an electronic system as opposed to your sticky note system that, don't you hate when you, those sticky notes get lost?
All the time. And so there's that conversation. And then the next conversation is how come my Bluetooth AI generated note doesn't show up the right way and the such and such. And I want to do the next thing with predictive analytics. And so in the space of 30 minutes, I'm going from turn on the computer and let's talk about electronics.
And then other times we're talking about, at what point do we retire? Because the Predictive analytics takes over all of my cognitive work.
It's funny because I hear that a lot. Is that really a concern at this point? We've got Not at this time. So far. It's not in your
term. But if you draw the line though, eventually you get to the point where you go, it's already pretty good at a whole bunch of stuff.
When does somebody finally Add the sophistication to connect all these together, we've moved from the idea of narrow intelligence, right, artificial narrow intelligence, which is I can land a rocket booster without falling over. I can, drive your car, if the conditions are right, all the way to artificial general intelligence, which is ask me anything, and I can give you a reasonable response.
And we're starting to see glimmers of that. So the dotted line gets there, but I'm not too afraid in the next few years
And there's, we could be wrong.
We could all be wrong too.
So what's the expectation of the CMIO at a major health system? Is the expectation to come alongside the physicians and help the physicians is the expectation to have one foot in the future and be able to guide where it's going?
Is it a lot of blocking and tackling right next to the physicians and the clinicians helping them to be more effective or. What is the, what's the expectation of the role?
I think you hit on a sore point, which is there is no expectation. You fit into the health system the best way you can.
You look for opportunities to make things better. I feel like I'm I'm an advocate for the well being of clinicians. That's at its root the core of my job, which is defending the clinicians against external regulations implemented poorly. We're great at that. The default is turn on all the tools, but big deal.
Clinicians will have to click through a whole bunch of alerts, but they ought to be reading those alerts. I'm sorry that there's 15, 000 hard stops in the course of taking care of one patient, but yeah, you should be thinking about that and you will never survive that. on the other hand we don't want clinicians to go I've learned that if I hit return three times, all the alerts go away.
then back to the, as if you didn't have a computer system in the first place, you're just trying to go fast and going fast. Isn't necessarily the best care either. So my strategy is. Protect the wellness of the clinicians. Most clinicians come to practice trying to do the right thing for patients.
And so it's these mountains of optimizing patient care and optimizing teamwork so that we're as efficient as possible.
So as you look back at your journey what sparked your initial interest in this field wanting to step into it?
I was completely misunderstanding the idea of IT in healthcare.
my first job was Chief Complainer. In 1997 I wrote a seven page complaining email about how terrible the rudimentary computers we had At the time, 1997, we're delivering foot high stacks of paper charts and the University of Colorado Hospital at the time had just bought Stat Land.
which is a DOS based laboratory computer system that would store, at the time, 90 days of test results. What more do you need than 90 days? Come on, really? Now, day 91, we ran out of disk space. We have to delete it. But in the meantime, 90 days is terrific. A the claim was this system is so usable that the login is.
And the password is lab. How much more usable do you want it? Alright, so I had a problem with that. I had a problem with the fact that the page down was the F1 key, and quit was the escape key half an inch away. So the usability of tapping next, often you would end up quitting. Right? And so I had page after page of complaint about how terrible the system is, and I thought great I'm gonna get in IT and I get to do, button design.
I get to move the screen around, change the colors. This is gonna be my job. I love this. And it was years before someone pointed out, change management is not at all about the technology. It's about something else. And so there's a curve in my learning that finally got to the point that, yes, technology is about 20 percent of the work.
80 percent is really the socio political skill that you bring to the table.
So what have you learned? Go deeper into that whole, talk about the learning. Talk about the learning about change management and taking technology, Into an organization and having it be adopted and used effectively.
My favorite story tell you a bad story and then I'll tell you an improvement story. early 2000s when I was first trying to push information transparency with patients. We were the first to publish the idea that you could give patients the entire chart. We had a system called SPARO, System Providing Patients Access to Records Online.
It's a very bad acronym that if you misspell S P A R O, you'll still get our research study. The idea being that we would show progress notes, what we call open notes now, and test results to patients immediately. And we chose one practice of seven doctors and 100 patients and did so successfully for a year.
studied in a randomized controlled trial and me thinking that most physicians are scientists and if I give them good p values, everyone will change their mind, of course. And I went and presented this first of all the seven doctors who we studied this on, one of The doctors was completely against the ideas, is the only reason I'm going to participate is because my colleagues are going to do and it looks like you're doing a rigorous academic study and at the end of the year, you will know exactly how bad an idea this is. And same guy at the end of the year turned around and said, This makes me a better doctor, because the patient's already probably seen the test result, instead of letting it sit for a day or two because I'm busy with other things, I'll get on the line and either send a message or make a call and say, Hey, I want to reassure you this result.
I've seen it. It looks pretty good to me. So I took that data and I thought this is rock solid data. It's high quality data. Patients love it. It's not more work for doctors. It's not more work for nurses. Let's go give it to the rest of the organization. This is back in 2001.
And I blithely walked into the leadership meeting with 40 other position leaders, gave my presentation. I'm just ready for the accolades to come in. Go, what a great study. And just like Machiavelli tells you, he says, if you don't already know how everyone's going to vote, By the time you get into the meeting, you haven't done your job.
And I had no idea what I was doing. And he also says, and the reason that change is so difficult in large organizations because at best, your advocates are lukewarm. And at worst, your detractors have All the passion in the world and it played out exactly in that meeting where the family medicine doc goes, you know That's really interesting.
We're trying to do more transparent. Why don't you come to our docs? Let's talk about it some more That was my best advocate and the next guy stands up and says CT. I don't know where you went to medical school But where I trained, the medical record is a dangerous place and patients have no business being in there.
I write about cancer. I write about terrible outcomes. I'm aiding my own thinking. The patient shouldn't be in here. This is a terrible idea. The next day I goes, yeah, I'm with Bill. Yeah, I'm with Dave and Bill. Yeah, Frank, Dave, and Bill have a good point. And pretty soon the attitude sweeps around.
I don't want to change. This is not good. I've never heard of information transparency. This is a terrible idea. And I got shouted out of the move. Yeah. Thanks for coming. That's a. Cool little pet project you did. You can go away now. And the lesson there was I ended up working on other projects for years because I couldn't get any movement.
And I didn't realize that there's a whole change management piece of it. Years later, having read this read Leading Change, which is now my Bible. I've read this book a number of times from John Cotter. It's a terrific book. If you haven't read it, getting the burning platform, getting the guiding coalition, making sure your messaging is right.
I didn't do half that stuff. And in contrast, when years later, I was trying to push a different project, APSO notes. So I don't know if you know that SOAP notes are Subjective Objective Assessment Plan, an electronic health record with note bloat, the plan is all the way at the bottom. It's a lot of scrolling when you get to any progress note, try to find The conclusion of the note, why don't you flip it?
So the doctor's thinking is the first thing you see. And when you explain it that way, eventually clinicians go sure, that's fine, but I had to change the way I document it. No, I don't want to do that. I, it's working fine for me as it is. And the way that I got this to change was recognizing that I failed with the same leadership group years ago, I came this time prepared.
I had done individual conversations with each of the medical directors and everyone hates. The emergency department notes, which were, 12 pages long and you couldn't figure out heads or tails of what happened in the emergency department because the assessment was sprinkled throughout all these 17 pages.
I would go in and say, you know how those ER notes are terrible and hard to read? I want to flip them so that the assessment plans the first thing on their note. They go, that's great. That's a great idea. would love that. And I'd say in order for me to convince them to do it, you're, I have to ask your department to do it too, right?
And just as you would expect, it was, there's nothing wrong with our notes. I know your notes are great, but from a change management perspective, I need your help to do that with them. And so they would agree. They would say it's not that hard for me to flip it around, but boy, if those notes are better, that's everything.
And so I'd had 40 individual conversations blaming the other guy that their notes are terrible and my notes are terrific But I'm willing to change to give you the benefit of getting those notes changed 40 identical conversations Everyone else's notes are the problem. My notes are fine, and I got 100 percent agreement to do right?
So you walk into the room and when you say hey C. T. Lin says when we go to EPIC We want to do absolute notes universally. No objections completely different way of thinking about things than before.
Cotter's book was our Bible as well for change management. And we even had things across, established burning platform.
Build a coalition of the willing. Essentially, we had a series of things. And when my team came in and said, Hey, this makes perfect sense. Everybody agrees. I love when they say everybody agrees that the patient, we can engage the patient more with this. And I go, everybody agrees. Do they, because even if it sounds perfect, if the patient has access to their medical record, they will be more engaged.
They will be more, all this stuff. And they go, no, everybody agrees. I'm like, I've never been in a meeting where everybody agrees with that statement even if it's the perfect statement, it doesn't, almost doesn't matter.
Exactly.
You know, we're talking about transparency here.
I'd love to hit on it a little bit. Elaborate on how transparency impacts patient care and outcomes.
can be difficult to objectively measure the benefit of information transparency. We know that even back in 2001, when you do a debrief with patients who participated in our SPARROW study, almost every one of them were like, I can't get this level of information from any other doctor I see.
I see the doctor's notes. I see their correspondence with other doctors. I get to look up the terminology for myself. Now I understand why they insisted on increasing that licinopril to its maximum amount because there's research literature that says I live longer when, I'm on the right dose.
So all of these things come out of patient conversations, but when you measure satisfaction with your doctor. You get a glass ceiling to start with, right? Most patients have a very high opinion of the interaction with their doctor, and you can't show a statistically significant p value improvement just because you have information transparency.
So it can be hard. The things we do know are that patients adherence to therapy does go up. We know that That there's no increase in workload for clinicians and nurses. And we believe there's more loyalty to a practice rather than jumping around finding the next doctor that, hey, if the doctor can trust me with all this information, I'm going to trust my doctor more and stick around longer.
And we think there's less churn in
those ways. So you and I talked about provider burnout the last time we were together. We talked about a lot of things. We talked about and you mentioned some of them earlier. You mentioned regulatory requirements coming down on the system and having to implement those things.
we talked somewhat about the technology and the technology burden. And also the pace of change just the amount of information a physician is expected to know. And the amount of new information, periodicals and studies and everything else just keeps coming. Well, in comes this thing called generative AI.
And right now, at this point, there used to be a hype cycle that sort of ended at a peak, which is if people are listening on the podcast, I have my hand apart by a certain amount. The hype cycle has been increased to a level I've never seen before. This thing is going to be able to interact directly with patients and you're going to be able to communicate with the patient at a fifth grade level.
Therefore, you're going to be able to take this very complex diagnosis and you're going to be able to explain it to them or the care instructions and explain it to them or take an ICU patient and do a summary Of 10 years of care and bring it all to, I'm hearing all these kinds of things and still feels to me like we're, we can see it from here.
It's, there's possibilities of it from here, but the hype cycle has really gotten a little out of hand at this point.
I absolutely agree. I absolutely agree. We raise our hands for innovation partnerships with our known partners quite often. In fact, we're working with Epic as our electronic health record vendor, and we're working with them on chart summarization.
It's still baby steps at this point, even though the claims are, yes, we can take, 10 years of data and give you a nice summary. We're still working our way towards, are you looking at the right progress notes? Because the early summaries we get from our alpha level partnership with Epic is.
We give it a whole bunch of progress notes and Epic or the AI mistakenly picks all the physical therapy notes. It turns out the patient has been seeing PT for the last 12 visits and it runs out of space to get to my first internal medicine note. So the summary of the entire patient chart is, patient's ankle is getting a lot better.
What about the heart disease? What about the Ran out of space. So we're still trying to teach it what to look for, what's important, that sort of thing. so yes, we can see that, we can see that there's something in the hype that once we get all of the blocking and tackling right, that chart summarization is going to be pretty amazing.
And we look forward to the day when the AI can write a discharge summary. Out of daily progress notes with no additional effort. That's something that takes docs, hours per discharge to do. And all that can evaporate. we're working on right now is a partnership on replying to patients.
in my organization, everyone knows that CTLin ruined healthcare by insisting that users use computers. And and so one way that CTLIN is trying to make up for that is knowing that during the pandemic, as our incoming patient message volume tripled, 350 percent increase from 53, 000 a month to 190, 000 messages incoming per month.
It's an unsustainable work that clinicians start that work after five o'clock, right? You have a full day of seeing patients in clinic or in the hospital and you go home and guess what? You, on the average, Primary Care Doc has 20 to 50 more messages to deal with every day. And it's an enormous burden and we thought applying generative AI to draft a reply might, in some fraction of cases, handle this volume.
And we've been trying this now for six months. And some of the best responses are when a patient says something pretty straightforward I hurt my ankle. What do you think? And it does RICE, Rest, Ice, Compression, Elevation spectacularly. And it's very friendly. Thanks for your message. Oh, I'm sorry, that ankle must be really uncomfortable.
Here's some good advice for you to try for the next few weeks. And then let me know if it's not any better. And the physician looks at the gendered response and goes, That's terrific. Send. No typing, and the AI has written a very reasonable response. There are the things that AI does terribly where the patient sends and says Hey, that CAT scan was really hard to understand.
What should I do about it? A, we're not giving the AI the entire CAT scan report. And so it doesn't have any context. And so the best it can manage is, It's important to see your, regular provider to come in and have a discussion, and that's completely useless. So we're now giving it the last progress note.
We're saying the last note written by me. And in some cases it's pretty astounding. One case of our rheumatologist a patient sent in a message saying, swallow knuckles. The message title was Swallow Knuckles. Okay, that doesn't make any sense. And then the entire body of the message was, I have more pain.
What should I do? And you would think there's no chance the AI can do a good job, but it read the last progress note from the rheumatologist who said, your rheumatoid arthritis. And so the reply says, your rheumatoid arthritis is probably the cause of your swollen knuckles. And, as was recommended at your last visit, you should take the anti inflammatory prescribed twice a day and let us know how that goes.
That's phenomenal. That's amazing. When you start to be able to put the pieces together just right, you can see glimpses of the brilliance that a tool like this can do. My favorite, I'll just end with my favorite example, which is let me give you a counter, not great example, and I'll end with my favorite example.
We had a, we also have our nurses engaged in using the AI, and it's very, we have to be very careful with our medical assistants and nurses that they don't go beyond their scope of practice and recommend treatment changes. You have to choose, Hey, this is something easy like a the patient wants me to re-fax a form, and the AI wrote that.
Don't worry, I'll re-fax that form for you. That's really great and it's perfect for MAs to use. One of our nurses had a complaint come in that, this cough and drainage for my nose has been going on for a long time. It's clear liquid, and most of my symptoms are better, but this drainage is, it's been two weeks and it's still going.
What could it be? Should I work? Be worried. And the AI says it, it's probably just a long recovery, but then it says. But there's a small chance this could be a CSF leak from your brain. Oh. Wow. And the nurse, having not read it carefully, sent the message to the patient, who immediately drove over to the clinic and says, this is an emergency, I need to see the doctor.
And we go we have to make sure our policies are, and we re educate our clinicians and nurses, be careful what you're sending. Because even though, strictly speaking, it's a correct answer for a neurosurgery practice, It's not necessarily correct for a primary care practice. And then I'll just say that in, our best example is a patient who said, I'm so grateful about the forms that you've completed, so this legal thing went through.
I would love to bring you pizza, bagels, whatever, to treat you and your staff. And their AI response is, we're so grateful that what we did helped you out. However, we have a policy for not accepting gifts from patients and your gratitude is all the thanks that we need. I'm not sure I would be that gracious in my writing.
No.
talked to Chris Longhurst about this and the study UCSD did. I think Stanford was involved. I don't know who else was involved, to be honest with you. And that's the one that everybody likes to cite of, generative AI is more. empathetic than clinicians.
Right. And it's funny, cause when I asked Chris about it, he goes, He said, most times the clinicians have just finished a long day and they're just getting to those notes. And he goes, it's just human nature after a long day's work. If you're asked to do, even more work, to use fewer words.
You're going to be more to the point. You're going to be very succinct. Whereas another word costs generative AI, nothing. It's right. And you can instruct it to be.
Here's the curve ball. You ready for this? So the curve ball is our usage rate of the generative AI notes is actually still pretty low in the 10 percent range in some clinics as high as 20, but in the 10 percent range, and the reason when we asked them, why aren't you using it more?
They go, it's too nice. it doesn't sound like me. My patients are going to know it's not me. And so they'll say, I'll just, I'll read it. I'll find it useful as a guide from what I'm going to say, but I'm going to pick up the microphone, say it myself. So that's interesting that clinicians feel like it's too polite.
They're going to see through this.
so two other things on this one is have you put a watermark on it? So the note gets sent, right? Because it went through the clinician, you then consider it sent by the clinician, or do you say this note was generated by the clinician?
We were
very careful about this. We actually engaged our bioethicist team and talked about this quite a bit. We know that in the EPIC library, there are organizations that do not reveal this, and I think in the long distant future, everyone's going to expect that any interaction with the EHR will have an AI component to it.
But right now we felt it critical and we followed the UCSD example of using the sentence that says automatically generated. and then edited by Dr. Lin. So we make it very clear there's both entities participating in the generation of this node, we think it's critical to do
I was talking to Michael Pfeffer as well about this, and similar usage rates, like 10 percent, maybe 15 percent The rest were reading it somewhere can you stop putting it in there I'm not going to use it.
I
think we have to be careful about how it's being measured, because right now we only have a very crude measurement. The measurement is, did you start with the AI draft, or did you start with a blank note? Either way, you're reading the AI draft. And many of our clinicians who think that, quote, the reply is too polite, We'll say, I like the idea, I just said it myself.
And so we're not actually capturing the rate of clinicians going, this was useful, don't take it away. All we're doing is capturing the percentage of people who pushed the actual button that says, start with the draft and I'll edit it. Because some clinicians say, it's harder to actually fix a sentence in the middle than it is to just mumble the whole thing myself.
So we don't really know.
you believe we're going to see I know with generative AI right now, we're seeing models that you can train it on your language. You can train it on your emails, your stuff, and it can start to mimic you. And
I think it's coming right. So right now, industrial strength work is super generic.
And my dream of the far future is it reads my entire oeuvre in Epic, right? I have 10 years of writing. Why can't it sound like me? Yes, it, I think eventually will. And, make it do the first draft and it'll be much better. And it'll have the patient's entire chart rather than just the last progress note, for example, these capabilities, I think will expand.
📍 📍 In the ever evolving world of health IT, staying updated isn't just an option. It's essential. Welcome to This Week Health, your daily dose of news, podcasts, and expert commentary.
Designed specifically for healthcare professionals like yourself. Discover the future of health IT news with This Week Health. Our new news aggregation process brings you the most relevant, hand picked stories from the world of health IT. Curated by experts, summarized for clarity, and delivered directly to you.
No more sifting through irrelevant news, just pure, focused content to keep you informed and ahead. Don't be left behind. Start your day with insight at the intersection of technology and healthcare. This Week Health. Where information inspires innovation. 📍 Increase
You look at the potential for this, where do you think it's going to have the greatest impact? And right now we're just talking about generative AI. We know that AI is being applied in a lot of different areas across predictive models and those kinds of things. But generative AI specifically.
Where do you see it having the biggest impact? Where they say, CT, you ruined my life as a clinician. And now all of a sudden it's nope, this is your penance and this has changed the game.
we can only see so far.
To the horizon and just what was it? 2009 when the iPhone came out, we're like smartphones has gotten as far as they can. Everyone has a Chiclet keyboard. Come on. And then, in the next year wow, there's a whole different paradigm shift. I think there are paradigm shifts coming that we can't see, but from where we can see, right, let's just draw the line out for when we can see which is that I would love to have an assistant sitting on my shoulder going, Hey, did you think about this?
And either it's a reminder of care gaps. It's very straightforward. Did you get the tetanus shot? Hey, they're overdue for a such and such, or reading my note and go, Hey, you're writing this thing that sounds like familial Mediterranean fever. Have you thought about that? No. Frankly, I didn't.
But I'm glad you spotted that in our interaction. So I could see different assistants sitting on my shoulder, giving me nudges and tips and going. What do you think about this as an idea? And that broadens my thinking, or deepens my thinking, or speeds my thinking. I like that idea.
Eventually, does it get to the point where it replaces me and it's faster than me? Sure, at some point. But, yeah, anyway, then you can say, I can focus not on the grunt work. There's so much grunt work, which is signing forms, putting dates on paper. Can that all go away so that I'm working at the top of my scope, which is thinking about really challenging.
What the heck is that thing? And trying to noodle my way
through it. I heard Bill Gates's talk right after he had seen ChatGPT for the first time. And he envisioned this whole idea of an assistant that's with you at all times that hears you talk about, yeah, going on a trip or in June, I'm going to go to here.
And the next thing you know, you're receiving information on flight schedules and hotels and that kind of stuff. And when I heard that, I thought, Cognitive load. For the clinician, there's so much cognitive load throughout the day. It would be nice to take some of that off so that they can focus on the handful of things that they really need to focus on.
And it's not a handful, it's greater than a handful.
So I would love to throw another curveball here, which is that one of my favorite books I've read recently is The Glass Cage. By Nicholas Carr and talking about the next phase and the blowback on automation. And he uses the aircraft industry as an example of the fear that pilots with a lot of automation and assistance on aircraft takeoffs and landings that you atrophy your skills because the routine stuff that AI can do for you.
And when you finally get, when you rarely get into a situation where suddenly all your skills are necessary, then those skills disappear because they've atrophied. And so the idea is this automation complacency, you just accept what the AI gives you. and then when finally push comes to shove and you have to take the controls, you're like, wait, I haven't done this in so long.
I'm not very good at it. Is that's going to happen to clinicians. And we're already seeing an example. It's not AI, but just with our alerts, drug alerts. I had a clinician say to me, I wrote for a muscle relaxer for my patient who was taking birth control, and your system didn't stop me for that drug interaction.
That's your fault. You're like wait, just gave up all your expertise to the system and you just hit enter and just, please stop me if this is wrong. the way that you're going to treat AI now. And think we have a big, I have a big fear. That you get to the point where you trust the system more than it's good for.
these are some potholes in the road.
Yeah, this is going to be where the rubber meets the road. It's like we're self driving cars. if a self driving car gets in an accident, who's at fault?
Yeah,
right. And, if a patient is harmed in some way. And AI was somewhere in that cycle who is at fault.
Now we'll leave that to lawyers because I have talked to some and who said, look, it's AI is a clinical device, blah, blah, blah. It still needs the clinician on the other side. And that's the category it's in right now. But still Those are maybe not issues to be dealt with, but they are issues to be dealt with.
But they're points that need to be well educated across the board.
difference between this tool and other tools is that it is so broadly available. In fact, I'm working with a medical student who said, I paid for a subscription to whatever AI service, and I uploaded my entire infectious disease textbook to this service for 4, something like that.
And she says, now I can ask questions. Hey, what would I do with an E. coli infection with, such and such that's a hyaline nephritis? And it would give me a paragraph, and it would give me citations out of the book because it knows the book inside and out. A student can do this. You can do that? Can you teach me how to?
So it's no longer an informaticist, top down control of the technology. It's groundswell of people figuring stuff out. And we have to be careful what
comes out of this. It's similar in IT. We used to have infrastructure. That was the purview of the CTO And then all of a sudden people were like yeah, I put my credit card in.
I got storage. I put the database out there and I'm doing research on it and I'm running through the stuff. And you're just like, what just happened here? Just because it's possible doesn't mean it's good. However, it's accessible. And so it democratizes access to these kinds of resources.
You bet.
You bet. It's amazing. We talked a little bit about InBasket redesign and your ambulatory sprint teams. The last time. And like to close on that because Maybe pick one of those that you think might be of most interest to the community out there.
Because
It was a transformative set of things that you're doing at University of Colorado and the team's doing theirs. So I'd love to hear more about it.
Yeah, could go on for a while. I think that I want to talk about Asymmetric messaging, in basket messaging.
And I give a talk that has nine big points about how you can blow apart the in basket. Everyone hates the in basket. Everyone hates our personal emails. It's unsustainable how easy it is to generate messages, per people generated messages. And then to try to reply to this thing, one of my colleagues said something I will never forget, which is the in basket and your email is someone else's to do list for you.
And you think, wow that's where I spend my whole day is on someone else's to do list for me. How interesting. And what that does is it takes away from if you have a career goal or long term non urgent projects that completely evaporates. When all you're doing is putting out tiny fires by email and in basket messaging.
Now for in baskets for clinical care, of course, patient needs what they need. But for email, you might say, if I spend my whole day in email and I never make progress on that long term thing that I was trying to accomplish, am I pointing it in the right direction for my career? So that's an open question that you have to personally redesign in their entire books about this getting things done and so forth.
From an asymmetric work perspective though A, I think it's important to take the incoming tool and make sure that your teams, your clinicians, and their nurses, MAs, and staff are optimally working together. In worst case, you'll have an ponging the same message 17 times because the one doesn't understand the other.
And typing little words is a terrible way to collaborate as a team. And instead, you should sit next to each other and say, all right. For the next 10 minutes, let's tackle the next 17 messages, and let's verbally talk about who's going to tackle what, and what do I intend. And that's going to more effective than the ping ponging of messages back and forth.
And people are like, you can do that? I can sit next to them? I don't have to use the No! Please don't the tool. Cool. Talk to each other and the tool becomes an augmentation of your personal relationship. Oh, it's just sometimes just because you have a tool you think you're supposed to do this way and to be a good soldier.
And so we try to break that mold by saying our approach is it's teamwork first. And then by the way, we have tools to augment your teamwork. And if you come at it that way, you go, Oh if I have permission, I'm going to redesign everything. And you should, because every time there's a new team member, the EHR, generative AI, you should say, welcome to the new team member that changes everything in our interpersonal dynamics.
Let's figure out the best way to use it. One of the things that one of my colleagues is doing is actually protecting the time of primary care docs in terms of reducing burnout and improving messaging. Let me explain that. Let's say in internal medicine practice, you're seeing 10 patients in a half day.
And in that time you have a full load of 10 patients and by five o'clock. 6 o'clock, whenever you finish clinic you've got 20 to 40 messages you have to tackle, and some of these messages could have been converted into a visit. These are complex enough questions. I have a patient sent 17 questions in one in basket message, right?
What's with my elbow? Is it a thing? I thought it could be gout. Can you help me diagnose this gout? I need a prescription refill on these 17 things. My blood pressure is doing this. My A1C is going crazy, and it could be the sweets I'm eating at night. that's one message I can't even, right?
And the right reply is, come in for an appointment. My next appointment's not for weeks. And so nobody's happy with that interaction. And what's the best way to do that? And one idea was, can we protect one spot per half day for every clinician? To say, that's your spot, doctor. No one else is going to touch it.
And you can do whatever you want with it. And if you're drowning under messages, take that half hour, take that 20 minutes, and all you're doing is processing your messages during that time. So at least during an 8 to 5 shift, you're doing messages during the day, and you have your team around you. Because otherwise, the 20 to 40 messages you're doing at night, nobody's around.
You're doing it by yourself, and you're spraying other people's in basket these messages, and there's just delay after delay. Why not say you can do some of this work and hand off some of it by sitting next, huddling with your MA and processing some. You could do it that way. And already that was very satisfying to our team members to say, you respect me enough to give me time back to do actual work rather than pushing this over to pajama time.
Number one. Number two, someone like me Where I'm in clinic less frequently than a full time clinician, I'm 20 percent in clinic. Then, my challenge is I can never get hospital follow ups in with me in a reasonable time, because my time to next is, eight weeks out, and I'm supposed to see these hospital discharges in seven to ten days.
I can never do that. With that protected time, I don't use that for messaging. I say, I have a spot where I can put you in tomorrow. I can, I have a spot where I can put you in Monday of next week. And for the first time in my career, I'm seeing my own hospital follow ups, the high risk patients that I really have to connect with.
I'm so happy with that. That's just a tiny bit of added control back to the clinician. And at the end of three months of trying this. Taking a spot out and saying, we're going to absorb the financial income loss of a 10 percent reduction. Turns out clinicians put other patients under their own control into that spot.
Raising the RVU again, and I'm getting to do things I want to do. And we went from a 90% anticipation of thinking at baseline, 90 percent of our clinicians said that I may cut back or leave medicine in the next three years. All the way down to 30%, dramatic reduction of intent to leave. Now, A, that's a terrible number, 90 percent in the next three years of my colleagues are thinking about cutting back or leaving, but, and 30 percent is still terrible, but what a difference that is for a small investment of time.
And I don't think a lot of clinicians or practices are thinking about this. What an opportunity to really partner with how clinicians are thinking about being professionals in their career.
Yeah, 90 to 30 is amazing. The 30 would indicate the other things that are still a challenge with practicing medicine and why not.
Ambulatory sprint teams we'll close on this. Just a couple minutes on this.
So in I will fully give credit to
I'll think of his name in a minute who was working at Sutter previously when he developed a team called Integrated Architects, where he took some 20 year olds and sent them into a clinic and said, I know you don't know anything about medicine, but we'll go in there, think about process, come out in two weeks and tell me what you did.
And he sent in a web designer, a database architect, a number of other folks, and he said you would be astounded with what people can do when they really take teamwork apart and say, here's some tools that I built because I watched you do your work, and there's improvements here. We stole that idea into our sprint teams.
We have now a team of 11. There's a whole hour to talk about how I finally got governance fund a team of 11. But this team of 11 is partly EPIC analysts who can rebuild tools, a physician informaticist, a nurse informaticist, a project manager, and four to five trainers who can reteach the EHR in a way that supports teamwork.
And again, unlike saying, here's a better way to use EPIC, it's how are your teams working together? How is prescription refills working for you? How is managing referrals working? Oh, that's terrible? Let's see how you do it. Oh, you do it this way, and then you hand this over to here, and then it's five handoffs?
Have you ever thought about doing a daily huddle? What's that? A daily huddle is where you sit and solve the problem together. Oh, that's a good idea. And by the way, if you have a daily huddle, here's a tool in Epic that supports your daily huddle. Oh, how cool. So this is the kind of thing that our sprints do is go into a clinic and take everything apart and put it back together.
And we get net promoter scores of minus 44, going all the way up to zero, which is a dramatic improvement of I hate you, all the way up to that's all right. And now, lately, we're getting Minus 12 scores going up to the plus 20s and plus 30s and plus 40s. And the net promoter for Sprint itself is a plus 70, that was astounding.
Can I have one of those again?
Does it lead to a highly customized EHR or do you somehow? So this is
the thing we actually learned our lesson between over customization and over standardization. And prior to having Sprints our team was even with hundreds of analysts on the team, we have. 4, 000 doctors.
It's impossible to even get down to the level of specialty customization. We hit the sweet spot of saying we will customize by specialty, but all the specialists have to work the same way. Right? One endocrinologist cannot be different from another endocrinologist. And so when we go do a sprint in one endocrine clinic, we will contact the other three or four endocrine sites and say, we're about to build a tool.
Does this sound good to you? Because you're going to get the same tool. So we have a customization for endocrine. We have customization for cardiology. We have customization for ENT. But it's system wide. And that's the sweet spot for us. It's not over customization down to the individual. We had order sets down to the individual surgeon before, and now we're saying the specialty has one set.
And that's worked well for us.
Did did the 11 people contribute an ROI to the organization?
I know that somebody's going to be listening to this and going, Yes. How did he do it? How did he get 11 people?
first sprint were people that I stole for two weeks. I just stole them. And I'm like, can we please have, no, can we please have, no. And I'm like, fine, let's just go do one and show what people, comes out of it.
And the satisfaction was skyrocketing. And the medical director of that practice was like, you gotta get you one of these. And by the time my team was disbanded the CEO, the CIO were hearing rumors that everyone wants a sprint thing, whatever that is. And and they're like what did you do? How did you steal these people?
And I eventually got permission to keep the people I stole, and then that we would work out other arrangements for the jobs that they were no longer doing. But from an ROI perspective, we went after the wellness number, which is that we think that out of every 500 doctors, seven on average quit.
And then we, our claim was we could cut that number in half. And if that's the case, two and a half people works out to, two million ish dollars to replace these people on a churn constantly. And if we are having more satisfied clinicians, maybe fewer of them leave. That's hard to argue from an ROI perspective, other than, this is our projection.
And to the leadership, what else are you doing to invest in physician wellness? We have some yoga classes and we have We think it's harder data than Yoga classes in terms of retention.
think I saw online where you used ChatGPT to write the lyrics for something that you played.
And I also saw, a fair number of posts while I was out on your blog, and doing a lot of stuff on computer generated art as well. is that just more of an interesting
thing to you? Yeah. So I write a blog at ctlin. blog and I've always worried about linking to images from the internet.
I try to make attributions, but it's not truly creative commons attribution. So I'm always looking for images that I don't have to worry about, stealing from others and GPT allows me to say, show me an image of Yeah, a doctor and a nurse arguing over computers and here's three images from Microsoft's generative tool.
And so I've been doing that lately. And then recently one of my colleagues and I did write a song to Taylor Swift's Blank Space that we use the AI to draft some of the lyrics. So that was a lot of fun. Just some of the lyrics. Just some of the lyrics. Yes. Yes. rewrote about 80 percent of it.
But. it's like the notes you were talking about earlier. It wasn't in your voice. It wasn't what you wanted to communicate, so you had to rewrite part of it. Exactly. That's awesome. Dr. Lin, I want to thank you for your time, and want to thank you for, sharing your wisdom and experience with the community.
It's always a pleasure to catch up. I appreciate it.
Thanks for listening to this week's keynote. If you found value, share it with a peer. It's a great chance to discuss and in some cases start a mentoring relationship. One way you can support the show is to subscribe and leave us a rating. it if you could do that.
Big Thanks to our keynote partners, Quantum Health, Gordian, Doctor First, Gozio Health, Artisight, Zscaler, Nuance, CDW, and Airwaves
you can learn more about them by visiting thisweekhealth. com slash partners. Thanks for listening. That's all for now..