23From Chart Chaos to AI Symphony in Healthcare
This week I am talking to CT Lin, MD, (@ctlin) System CMIO at UC Health who tells the story of his early entry into the world of healthcare informaticists as the “Chief Complainer”. CT delves into the complexities of electronic health records (EHRs), the pressing cybersecurity concerns, and the overwhelming in-basket conundrum faced by physicians.
As healthcare systems grapple with mounting patient messages and the burden of communication, Dr. Lin offers insights into the delicate balance between using AI and preserving human judgment. The application of AI to draft responses to patient queries emerges as a promising solution, easing the load for clinicians while ensuring human oversight.
CT Strikes a Unique Chord
We discuss the nuanced challenges of integrating technology into healthcare, presenting a vision where AI can be a supportive tool for physicians while emphasizing the importance of human engagement in the design process. Listen in to hear the culmination of the conversation as Dr. Lin’s breaks out his ukulele for a signature performance, where he playfully reimagines Elton John’s classic to humorously critique the intricacies of EHRs.
Listen live at 4:00 AM, 12:00 Noon, or 8:00 PM ET, Monday through Friday for the next week at HealthcareNOW Radio. After that, you can listen on demand (See podcast information below.) Join the conversation on Twitter at #TheIncrementalist.
Listen along on HealthcareNowRadio or on SoundCloud
Raw Transcript
Nick van Terheyden
And today, I’m delighted to welcome Dr. C. T. Lin. He is the chief medical information officer at UC Health CT. Thanks for joining me.
CT Lin
Thank you for having me.
Nick van Terheyden
So as I do with all my guests, always important to get a little bit of the backstory. You’re a physician, very accomplished physician, but also an informaticist. I think that’s been one of your passions, certainly, through time. Tell us a little bit about your journey to this because I imagine similar to mine, probably not a straight line because informatics wasn’t really a specialty even although now it is.
CT Lin
In brief, I started in my role as chief complainer. In 1997, I was a junior faculty member at the University of Colorado. And there was talk that in the future, we might take our patient charts out of the storage basement and put them in the computer. And I was all about that I served on a number of computer committees. And at one point, based on a response to one of the administrative leaders saying all you doctors need to calm down, because look at all the stuff we’ve purchased on your behalf. We have a dictation service. And we have all these, we have a lab test lab system, where you can access up to 90 days of test results. Now on day 91, we’ve run out of disk space, we have to delete stuff, but what do you need more than 90 days worth of, of data? And And oh, by the way, will tell you how usable this system is so usable, that the login is L A D, and the password is L a b? I mean, how much more usable? Can we make a system for you? Well, I wrote them a seven page rebuttal on how terrible their design philosophies were. And as a result, I was hired on as a 0.1 FTE the title being because informatics hadn’t been invented yet physician liaison to the IT department, because you seem to have a lot of opinions. And so I served as that and grew the job from point one, four hours a week, up to what I have now, which is an 85% appointment, and a team of 20 informaticists. But it took quite a long time to journey out of the wilderness on that one.
Nick van Terheyden
So, you know, lots, lots of pieces in there of, you know, some of the challenges that we’ve seen, certainly in healthcare and technology and the application of technology. One of the things that I say frequently is, you know, walk a day in the shoes of the individuals using some of this technology to really get a sense of what’s necessary. What’s appropriate. You talk about that from a security standpoint. I mean, let’s be clear. That’s one of the pressing issues for informaticists and healthcare facilities, they have the biggest target on their back. When it comes to the securing of this information. It is a rich, rich target. Is that something that keeps you up at night?
CT Lin
Yes, we have plenty of things keep us up at night. The cybersecurity piece is not directly under my informatics department. The IT department works on that and every time I walked by my cybersecurity colleagues, they show me the map of how many hundreds of 1000s of attacks we undergo every day. Yeah, and I just am flummoxed by how much how many layers of work that we have to put into our defense just to keep our doors open and not be one of the victims. And you know, cross our fingers. We keep one half step ahead of the hackers because, you know, everything is improving. So quickly on both sides of this battle.
Nick van Terheyden
Yeah, it’s interesting, I feel it’s a little bit like a car alarm. I mean, I have the same thing in my own home network. I track it, I watch those attacks. And I’m always hoping that, you know, I’m I’m the network that has the alarm in it. So oh, well, let’s move on to somebody else that doesn’t, which is not, you know, the most compassionate way of thinking about it, but you’re right. It’s a constant battle. It’s a nuclear arms race. So you’ve seen a lot of change through the course of this. I mean, we’ve seen the implementation of these medical records. You know, there’s there’s some challenges, especially with the way that that’s been rolled out. Let’s talk a little bit about the sort of process and some of the technologies that you see physicians trying to use with communication. So one of the things that we hear is this in basket and the challenge that physicians have constantly with burnout, we’re sort of processing all of this information they’ve layered on all these things. I always look at any EMR screen doesn’t you can name any one in my view, and it looks so complex, so challenging. Even as a technologist, I feel don’t want geeky, I can’t absorb it and use it. What are we doing to try and fix that? Because it just seems to be overwhelming everybody as they come into the profession?
CT Lin
Well, let me take the first half of your question before we get to in basket about design and complexity. Because we’re a victim of our own request. I’ll tell you that when we first were looking for an EHR back in the early 2000s, many of my physicians were very much drawn to, and I forget the name of this EHR vendor that has gone out of business a decade ago, that their big claim to fame was, you can be up and running with with half an hour of training. And our EHR is so easy to use, we designed it with physician usability in mind, and you go and use the system and then you go to talk to doctors who’ve purchased the system they go, and it doesn’t hardly do anything. It doesn’t do anything, you know, it will write a prescription. But can you customize it? No. Can you save a favorite? No, there’s only one way. But boy, it’s easy to use, you walk up and it’s self evident. You go here you go here you write the prescription it transmits, or it prints. But no, you want to make it print to two locations at once. Because you want the patient wants a 30 day supply here. And then a 90 day supply to mail order. Nope, doesn’t do that. But it’s easy to use. And so you have to be careful what you ask for. And we asked for complexity. Well, on Tuesdays, I’d like it to work this way because I have a nurse, but on Wednesdays, I don’t have a nurse. So I want you to do and the EHR vendors are very happy to design all of those possibilities. And what happens is your screen gets more and more cluttered because they’re responding to our requests. So I can’t say that it’s you know, this them versus us, it’s us asking that you need to fit it to my to my workflow. And so they come up with more and more versions to try to meet that. And so as informaticists, we have to take two steps back and go. All right, it will inevitably get more and more complex, because we asked for it. What’s the right balance? How do we tell the asking physician, no, you can’t have that you have to learn a standard way, we can’t give you five different ways. And the sixth way because, you know, your, your colleague likes it the other way. So it’s a hard balance to strike. And that’s a large part of our work is simplification, from the work that we’ve from the complexity we’ve actually asked for.
Nick van Terheyden
Interesting, you bring that up, because one of the things that always, you know, humorous joke in any setting, but you know, gather seven doctors in a room and ask them for an opinion. And you’ll get eight, nine, whatever. And we do have a tendency to come up with alternatives. I want to push back a little bit because as I hear that, you’re right we do we asked for these things, but we asked for it in the absence except for those that have sort of followed this and understand the impact and the customization challenges. And it reminds me a lot of the renovations through the pandemic, where if you had a contractor and you say, oh, I need this done? Sure, I’ll do this. But that’s change orders and it’s more money. I feel like there’s a responsibility on the part of the vendors as well to sort of push back and say, Well, no. Is that something that’s even viable? Are we willing to accept that?
CT Lin
I think that we’ve worked with easily a dozen vendors over the years. And our current vendor epic at this point, has done a reasonable job trying to standardize, and yet they’ve got hundreds of customers asking for improvements. And they’ll they’ll you know, put it up to a vote, you know, how many organizations want one of these in, so you have to participate and vote, and the ones that get the top votes get 10 tend to get implemented. So it’s a balance. And are we there yet? No, we’re on version 17 of something that will take 20 3050 100 versions to get close to where we’d like it to be. And with AI, I think there is an opportunity to step back and go many, many of those buttons that can maybe go away. And we can ask the AI to do some of this work for us. And we’re starting to see the beginnings, the glimmers of that now.
Nick van Terheyden
So I think fair point. I mean, it’s it’s a balance between all of this, it’s easy for us to sort of point the finger and there’s an awful lot of that that goes on in healthcare. You know, I always have the sense that physicians abdicated a lot of the responsibility. They said, No, no, I’m not gonna get involved in this. But thankfully, you know, we’ve developed this whole medical informatics you’ve led, you know, a lot of
CT Lin
Let me interrupt you because as you say that I actually have one of my colleagues speaking in my ear that I can I can hear when we first went live in 2011. I will say that a neurosurgeon in our organization who will go unnamed, during the Go Live had had been very unhappy with one of our trainers who was trying to teach him how to use the EHR had to write progress notes that their preference lists and basically sent the trainer back to the command center crying. Don’t Don’t send me back there. I don’t want to talk to him again. So whose job is that at CT Linds job to walk down to have a physician, a physician conversation, I sat down with him and he says, Listen, this thing’s not intuitive. And he says for $27 million, it ought to think like I think. And my response to that was, first of all, it was $54 million. And, and second of all, we had planning meetings for the last year, you came to none of them. But your colleagues did. The thing thinks like your colleagues do, because they’re, they were at the meeting. Where were you? Then he was? Well, I was busy, you know, saving lives? Well, well, sure. And yet, you expect to come, you know, drive this fancy sports car, and you’ve never taken a driving lesson. And you didn’t help us design this thing for you. So it’s a matter of participation, and engagement. And I think that the clinicians who are most engaged in our work are the most satisfied, hey, you built it the way you know, I think and the way I take care of patients, and now I can see my way through to the other side. But if you are hands off and say, when I walk up to it, it should be completely self evident. No one walks into the bar and says a new piece of or equipment should be completely self evident, you’re going to practice with it. And I don’t think clinicians necessarily think that the EHR is a sophisticated tool they need to spend time thinking about, and that’s still a problem today. Some of our clinicians are, don’t come to me with training. No, no, I’m busy seeing patients, but boy, when I want to go to use it, it’s terrible. Well, there’s a reason for that.
Nick van Terheyden
It always reminds me of the commentary, I don’t have time for that. And that’s never true. It’s, it’s not a priority for me, which I think is exactly what you’re saying. And you’re right. It’s combined contributions. With some guidance and some handholding, I think, joint responsibility, which you know, is important. For those of you just joining, I’m Dr. Nick the incrementalist today I’m talking to Dr. C. T. Lin. He’s the chief medical information officer at UC Health. We were just talking about the EHR implementation, the challenges of combined contributions, the importance of everybody getting involved and participating both in the design phase, but also in the ultimate implementation training. We talked a little bit before about the in basket, I know you’ve got lots of thoughts around that area of overwhelming challenge for physicians. Tell us a little bit about that.
CT Lin
Sure. You know, I, I was gonna say that I have a reputation in my organization for being the guy who ruined healthcare, there is that that’s the guy, he’s the guy wrote out, he wants us to type in his effing electronic health record that he wrote some weekend ago? Well, sure, no, not mine. It’s, it’s from a vendor. But nevertheless, we’ve been trying to encourage information transparency and online communication with our patients since 2001, you know, several EHRs ago. And we’ve shown that patient satisfaction not only with connecting with their provider, but with overall satisfaction with their healthcare, in general writ large. It has, there’s an 11% improvement just by having patients feel like I can connect with my nurse my doctor online without going through the phone tree and being on hold and so forth. So we’ve grown our, our patient portal, even before the pandemic, up to about a million patient accounts online. And it exploded during the pandemic. You know, as you might imagine, when we shut the clinics down, patients found another avenue to contact us, we went from 53,000 messages a month, just prior to the pandemic, and three months into the pandemic, we hit 186,000. And that number has sustained in the 180 to 190 every month since then, that’s a 3.5 fold increase in message volume. And none of that translates to protected time to do that work. That’s just net new work on top of full clinics and full hospitals. And unpaid right.
Nick van Terheyden
What’s an unpaid,
CT Lin
unpaid, right. So after you go home at night, you open the computer backup and the typical primary care doctor will have 40 to 50 messages to handle and Yep, see, there it is CTN ruining healthcare, because, you know, we he’s insisting that we respond to patients in a reasonable turnaround time. And it’s difficult to how do you do that? It’s all it’s the right thing to do to connect with our patients and provide care. And yet there’s no billing back then there was no billing code. There is now and we’re stepping very carefully through our way of doing so because there’s also a public perception that the larger health systems are out to get the last few pennies in my pocket, and they’re going to nickel and dime us with every message telephone calls used to be free and now you’re going to charge me for this. So we’re very carefully Stepping into this idea that it is possible to build for these messages. But how do we do so in a way that’s public friendly and patient friendly. And we haven’t solved that yet. So over time, we’ve tried to figure out how to redesign the way we think about the Alaska we have a project where we are we calling it blowing up the in basket, and sort of putting it back together. One thing that we did to ourselves, which back in 2011, was a great idea we thought was improving communication between our specialists and our primary care doctors. We knew that back then only about 20% of our specialists would carbon copy their progress note, back to the primary care doctor forwarded. And we often were feeling left out, well, you’ve changed something with my patient, you didn’t let me know. And we thought, well, this is brilliant, we will automatically when we can identify the referring physician send a carbon copy without without anybody’s interaction. And it turns out over the decades, that is exploded into hundreds of 1000s of messages per month. And it mostly falls on the backs of primary care clinicians who are already overburdened with work. And so one thing we did that we stole from another organization, the the acronym was DC, the CC, get rid of the carbon copy. And that dramatically reduced the in basket volume, I’d say 19, out of my 20 doctors out of every 20 doctors were in a were approving of this, Hey, this makes my volume go down. And then of course, we have the 5%, who say, Well, you’re ruining health care, because I can’t see what my specialists are doing. One of my colleagues I talked to said, I save up all my carbon copies for a Saturday. And I’ll spend eight hours just reading the hundreds of CC messages. And there’s usually out of four or 501 or two, that mentions me by name and asks me to take a task. And usually those people will have already sent a manual forwarding. And so the automated forwarding really is just noise to me. And by the way, I’m going to see this patient again anyway, should I read the message now? And then again, in three months, when I see the patient and try to put everything together? And is that just not today’s work today. So that’s one of the principles we came up with is, let’s design the in basket for today’s work today, and not try to do yesterday’s work or work that we’ll be thinking about again, in a month or two.
Nick van Terheyden
Yeah, as you’re talking about it, it feels very much like a an ideal kind of task for an artificial intelligence or augmented intelligence to say, is this an appropriate message? You know, so you talk about it? Is there an action point for me? Or is this just informational? If it’s informational? Let me pull it as opposed to pushing it. Is that something that we can start to apply this, you know, wonderful, predictive capability that everybody’s talking about the AI? Generally, they say, but you know, there’s much more complexity in there.
CT Lin
Oh, I completely agree that there’s lots of small little opportunities for an AI to cut to improve signal to noise, we’d have to train it appropriately. And because of all of our concerns with confabulation, and hallucination, how do we trust it? And that’s a very difficult question. So we are stepping gingerly into this pool of AI assistants, because we want to make sure we verify every step of the way that it’s more helpful than potentially harmful.
Nick van Terheyden
And you talk about the capacity to communicate. I mean, that’s how we do with most other organizations email, you know, we’re not sort of pushed into a telephone tree, although in some places we still are. It’s, you know, it’s a frustrating experience. I like the sort of process of finding that balance, and particularly around the billing. I mean, it’s hard to sort of attribute a cost to it. But there is a cost for the clinicians. That’s time it takes more time to do some of these responses, perhaps, and some element of automation. So as we shift, talking about AI, there’s plenty of technology. We’re using it in specific areas. Tell us a little bit about your experience with that and what your thoughts are.
CT Lin
We are just starting to dip our toes into this space. One of the things we’re doing I know that a handful of other organizations are ahead of us on this, San Diego and Wisconsin and Stanford are is applying AI to draft a reply to online message from patients. And so the idea is that, for example, that poor primary care doctor is getting 40 to 50 messages a day. Could the AI draft a reply? No, we want to make sure that it the human is always in the loop. There’s no such thing as the Chatbot just picking up and running and having a great conversation and making stuff up about for the patient. But at this point, my colleagues have demonstrated some pretty good success at something like 20 to 30% of incoming messages. They can use something from the AIS generated reply. You can imagine that if somebody indicates you know my ankle hurts, I think I sprained it I twisted it this way. And what should I do? Well, the AI having read, the bulk of the internet can say, you know, Rest, Ice, Compression elevation that’s all over the public internet. And then you the clinician looks at that and says, Send. That’s beautiful. I know. And it’s and it’s nicely packaged, and it’s empathetically worded. And it says, please contact me if you have any questions, and in science, my name. And so for messages like that, that’s, that’s spectacular. Other times, it’ll be the incoming messages, Hey, I saw my cat scan results is quite confusing. Can you tell me what’s my next step? Well, there’s no chance that the AI is going to read the entire chart and figure out how to respond very nuanced way to the clinicians judgment and their specialty in the way that they would advise the patient. And for that, nope, you toss away the automated reply and you write your own. So I think we’re in the beginning of thinking, Where is it useful, or is it not?
Nick van Terheyden
Okay, so I think, you know, excellent insights, but I want to make sure that we leave time for your trademark that we will, I don’t want to say most well known for but certainly very well known for. You are a ukulele player that creates all sorts of songs and content around healthcare and music, and you’re going to share one with us now before we finish.
CT Lin
Thank you for the chance to sing for you. So this is a song called epic man and it’s updated for chat GPT make sure that
she checked my templates last night pre flight Zero Hour ad. And I’ll be deep in my in basket by then. I miss paper charts so much I miss my life it’s lonely in hyperspace on Sci Fi and I think it’s going to be a long long time before making another ever again I’m not a man they think no no no epic epic man writing open pay 12 pages ain’t the kind of way to write your note.
In fact it is full of bloat and there is no one can make sense of it if you did
and shad GPT I don’t understand it’s just my job seven days of the week
for another another man they think I’m in no no no
writing open notes
think it’s gonna be a long long time and it’s gonna be a long long time
Nick van Terheyden
Well, that was fantastic CT I really appreciate it it’s certainly a first for this show. Not a first for me I’ve had the privilege of actually listening to you and seeing and hearing you both present but also play a ukulele in person. Thank you for the privilege and a little reminder of some Elton John. I’m super impressed with your both your melodies but also the the clever use of the words and taking songs. It reminds me a lot of another Singer that I’m quite fond of so thank you for joining me on the show as usual we’ve run out of time unfortunately but what a great pleasure thanks for joining me CT
CT Lin
thanks for having me
