Junaid Kalia

I am so excited to welcome Dr. Kim. He is in our backyard. He's at Bellaire's Court and White and he actually does lead physician informatics team in the Plano Hospital. He has published some really amazing LinkedIn posts and I just wanted a real person who, who's actually implementing AI to hear from that exact physician who is doing it and leading it.

This change in charge which Edward Marx continue to speak about is going to be extremely important.

Andrew Kim:

So, hi, my name is Andrew Kim. I'm the director of physician informatics for the heart hospitals in Baylor, Scott and White. So you know, I work in with everything and anything with the ehr, as well as any sort of, like you said, healthcare and AI. Right now AI is obviously one of the biggest things in healthcare. Five, ten years ago, we heard nothing, but we heard nothing about AI. We heard everything about blockchain. Do you remember that every vendor was blockchain and none of them could explain what blockchain did.

But now every single vendor, you go to any conference and you just see A.I. everywhere. I think it has a lot of opportunities, but I do think right now the market is extremely crowded because every single vendor is working on their AI solution and not really working on them.

The main thing that you know, how does it service the patients, how does it service the clinicians? Right. But I think there are some gems in there. You know, I think, our main goal with, I work with one of our chief wellness officers and he has a main focus goal which is actually addressing physician burnout. So one of the things we're looking at is our current ambient listening product that we're utilizing in the clinics and in the inpatient setting as well.

And we are currently actually pulling data to see is it truly mitigating physician burnout? Is it truly reducing the workload? And also what are the, outcomes? Are they writing shorter notes or longer notes? Are they, are we addressing any pajama time or work outside of work hours and things like that?

Junaid Kalia:

So grateful that you are doing such amazing work and actually finding real world evidence because, there have been studies for ambient listening which are different. And then honestly I had to redesign my UI for my ambient listening portion to actually ensure that, the documentation review burden is not equivalent to documentation burden.

Edward Marx:

Yeah. Hey, Dr. Kim, thank you so much for, for those insights. It's always great, as, Junaid said, to have an actual practitioner as opposed to just, a strictly academic, if you will, talk, about this topic and, and you know, actually doing things. So, yeah, I, I am super curious. I know you're early, as you discussed in the analysis, but, but what's your sense and maybe for yourself and the, your closest colleagues, like, is AI truly helping yet, or, you know, where do you think we are on the maturity curve in terms of, you know, enabling a better lifestyle for clinicians?

Andrew Kim:

I think it depends on who you ask. One of the challenges that we run into, as in informatics, is that every physician has their own workflow, right? And you try to kind of set this, this workflow. A lot of vendors will create a workflow for the physician and hope the physicians will adopt to it. And if, you know, and as anybody knows, if you try to get a physician from practice in the exact same way in the past 20, 30 years, they're not going to adopt another workflow. Their workflow works the way it does because they've been doing it for so long. Their notes look exactly the same because they've been doing it exactly the same way for years. And if you try to give them some sort of a, A workflow where, oh, you're, you need to change your template to match our model, it's just not going to work. Yeah, so that's where I come in and I try to kind of marry the two together and make sure that, can we get the vendor to modify their model and can we get the physician maybe not to modify their model, but try to tweak the workflow just ever so slightly so that everything works a little bit better. So we're not. We really can't say much about how that's turning out in terms of AI tools and adoption. I think we're at the very beginning of it all and I think we're slowly kind of shedding and trying to figure out which ones are actually diamond in diamonds, you know.

Edward Marx:

And yeah, I think that's good insight. I kind of hear the same thing when I travel around the country, talk with different college. It depends who answers that question first. Interestingly, right. If you're in a conference room and you say, hey, how's, how is X or Y or Z working? Whatever the first answer is, everyone agrees. So they could say, oh, oh, this AI is working great. It's really helping with you know, clinician burnout. And everyone's like agreeing, but if the first answer is it's not helping, then everyone agrees with that. It's very interesting. But what would your recommendation be for? What would be your advice? Working with IT people or chief digital officers, chief information officers, you know, the people that you might interact with and how to best work with clinicians to partner to make sure you get the best solution and not this fragmentation that you were talking about, you know, these kind of best of breed silos.

Andrew Kim:

Yeah, I mean I think that's, that's the main goal, right? You have the tech people that are developers, analysts and code writers that have never really stepped foot inside a hospital. So they might have a brilliant idea, but they have no idea how it may fit to a clinician's workflow. And then you have clinicians who clearly have a problem and they have a, they want a solution, but you really can't marry the two. I think the main thing is it's just like any sort of sales, right? Find out what the problem is first we'll learn their workflow. And like I said, every physician has their own different workflow. Although the industry really has to try to make everything, you know, size fit one, but you have to try to modify it so that it may service 80% of the population. And if that's the case, I think you gotta winner. But I think a lot of the companies, their main number one problem is they try to find a solution that to a problem that may not exist or that may not address the actual problem itself.

Harvey Castro:

Yeah, doctor, Kim, thanks for coming. And I just want to, I'm curious from your institute, if you're able to share what, what kind of things are you doing to change the AI healthcare culture? Are there, Is there workshop? Are they going over stuff, you know, what, or what have you noticed?

Andrew Kim:

We've set up AI governances within our institution, where a lot of our leaders sit, as well as, ethics committees, legal compliance and clinicians as well, and tech people. So you kind of want everybody to pitch in and make sure that this is beneficial not only to our clinicians, but obviously patients number one, and our clinicians as well.

Junaid Kalia:

So the question for me is that one thing is culture where you can actually say, okay, you know what? This is the new Insight, Advocate, Advertise, etc. But at some point in time, what do you think is happening in a year, two year, or five years from now?

Andrew Kim:

I distinctly remember, a neurologist there that was kind of on the older side side. He told me, he said, Dr. Kim, I don't use a computer at home. I don't have a computer. I don't know how to use the mouse. And I knew that it was going to be challenging. And I think in the end, he did retire, after spending a few months trying to adopt a new system. System. I, I, I don't think it's, I do think there is an age component, but I'm not so sure it's an age component. More so that it's the willingness to learn. You know, we have older clinicians that are completely into the new tech. They want to learn more about AI. They want to, you know, learn more about how this can help them and benefit their patients. So I think it's just like with anything.

Junaid Kalia:

We always talk about patients, which is important without a doubt. I mean, and then physicians, of course. But then there is the care systems aspect of it, which I do a lot because I was the director of neuro, icu, director of stroke. Get with the guidelines. I'm looking at it. What I'm saying is that, that the third one, which is the informatics part, where you actually shine. Give me your vision of current capabilities that you think and then future how important that will be, including rcm. Like whole hospitals revenue cycle can change. And then what Edward always asks is the AI wars between the hospitals and the insurance companies.

How are that gonna adjust as you see forward, the back office in a clinic or the, you know, all of that.

Andrew Kim:

It should theoretically cut down on chart review. Theoretically. you need to look at who gave what at a certain time period, within the certain time, admission stay. And I think that's a little bit more nuanced when you're looking at very specific details. You know, one of the things that I think we find challenging is that I, you know, in the future, and I'm afraid to say this, but maybe you can actually just ask some sort of LLM or GPT into your EMR and say who did this when and it'll spit back an answer. And I think we're kind of halfway there.

Junaid Kalia:

In my opinion, given every single day, there's something new coming. Harvey, what are your thoughts on it as far as, you know, you talk to ministries, you talk to, you know, countries, not just, you know, low level people like me. I'm grateful that you always put me in high regard. But what I'm saying is that, have you seen that? That the back office, the back end, which is massive, by the way, if you look at it, there's 1, 10, 9 administrators for one physician.

Harvey Castro:

Man, that's a good one. Okay, so big picture, I want to go back a little bit to what Kim was talking about, because I think it's important for people out there because we have the Bell curve people listening. So I want you to understand the practice of medicine is the art of medicine. And so there's a lot of variability among us, even though there's guidelines. And even with those guidelines, they're shifting and where AI helps. And so my point is this. You're creating a model that it's fluid, that's also timestamped. You're creating algorithms that say, here's the guidelines.

But then not everybody's keeping up with the guidelines because it changes. what I'm trying to say is, what I'm seeing right now is this tension. And going back to summarizing what Andrew said on the age gap, it's not necessarily an age difference.

I agree with you. I think it has to do more with early adopters and late adopters. And as a result we're starting to see, see different ways of doing things I do think what I start seeing is there's going to be a shift and that shift will start looking like a better consolidation of understanding that whole AI and healthcare platform.

Edward Marx:

So yeah, there's a lot of, oh, it's clearly there's, there's tremendous opportunity Right. You see the headlines all the time. More and more health systems are letting people go. There's a lot of risk happening, because the cost structure is too heavy and how can we leverage AI? To use a computer. If you have to drive somewhere to use a computer is not good business. So, but they have these factories of revenue cycle people as well as IT people.

And I think obviously you can use a lot of AI today to displace a lot of that. And again, nothing against great people, wonderful people. We obviously mean no harm. It's just, it's just an alert that we should re. Educate ourselves, refocus. You can actually, you know, leverage all the things that we're talking about here and other things to really create a new opportunity for yourself. But yeah, obviously again, revenue cycle is a big area where a lot of automation could take place and significant reductions. IT, is another place that you could, leverage a lot of AI. I think all of our systems, you know, we're pretty bloated.

Right. Over time we just kept adding, adding, adding and not really making those hard decisions, that we can make.

Junaid Kalia:

This is, this is my prediction that in the next five years we're going to decrease the ratio from 10 to 5. Thoughts given that. Now, do you're on the first. As you clearly say, we're very early in this whole phase. People don't understand that. People think that AI is in here already. Then you have to change, you know, the, sorry, the what do you call leadership then physician. And then it's going to trickle down to back office and information, information that you are at. So talk, walk us through this.

What is, what is your sense of things moving forward?

Andrew Kim:

I think maybe, you know, I think the shift will be essentially their, their roles and what they do.

But you know, what if essentially those job roles shifted from working on pre auth and denials and fighting the insurance companies to maybe utilization and making sure that each single procedure or admission actually has the necessary documents and has the necessary filings before it even gets to that phase.

Not to mention, you know we've had this, I had this chat with a smaller AI, company that basically told us that they would fight the insurance companies using their AI. And my number one question is so your AI, is going to fight their AI and we're going to burn up a million trees and we're going to see a denial.

And then what? And they said we don't see denials. I said there's no way, you know, this is based. You're AI versus AI. Somebody's gonna win and somebody's gonna lose today. So, Dr. Kim, this is one thing that I ask everyone.

Junaid Kalia:

That if you had one wish and you would have a magic wand and you will only fix one problem, what would that be? Oh, man, that is a tough one.

Andrew Kim:

I think one of the biggest challenges that we run into with at least EHRs, and, you know, that's why we have all these different tools now, is that, the EHRs have gotten so complicated, right? And they're so costly and they're so complicated, and you need to hire an entire team in order to run and fix and troubleshoot and train. If I had one wish is to make the simplest EHR, that with the cleanest UI that even you know, that requires no training and yet does all the things that a complicated EHR does.

Junaid Kalia:

Just to give you some history background, there are 1300 EHRs currently in the United States. 500 of them are actively, mind you, 500 EHR is being actively used. I think, you hilt the nail. There are very few. Harvey, your floor.

Harvey Castro:

Yeah, I'm going to take on that question. For my. For my magic wand would be a. I know you guys are going to cringe when I say the word. It'd be like the, the her movie where I can just speak into the phone and be like, look what's going on with my health. And then the phone's like, dude, you haven't slept you haven't been doing xyz, you didn't work out this week, you know, and I predict you're going to have a heart attack if you keep this up in 2.7 years.

And then I'm like, all right. And then that, that kind of feel where now healthcare is very seamless. And my whole point is it's more proactive medicine instead of reactive.

Junaid Kalia:

So you are right that it is not just about the physical activity, the you know, obstructive sleep apnea or anything. We are seeing multi omics in place soon. Love it. Love the answer, Edward. Sorry. And how would you say what is your magic one?

Edward Marx:

First things that come to mind are things related to patient experience and clinician experience. Like, and no matter where you are, whether you're poor or rich, whether you're in sub Sahara Africa or downtown Rochester, that you're going to get the same level of quality care.

So I think AI is the great equalizer to make that happen. So that's my magic wand because right now it's very uneven. It's tragic, right?

Junaid Kalia:

So Dr. Kim, one more. What is the difference between personalization and McDonaldization and why physicians feel like that AI is going to McDonald and not personalize.

Andrew Kim:

So the whole idea is if you can find that this is truly beneficial for patients, clinicians, and not just anecdotally, but you can really show that, hey, this actually works great for everyone else and it works best for our patients.

It has great satisfaction for patients, and clinicians, then I think we should adopt it. You know, and I think AI, the ones that remain will be the ones that really do produce great evidence, produce great results as well as, it's able to be a little bit more moldable than like you said, the great equalizer.

Junaid Kalia:

Beautifully put. You have to pick your battles. So I'm going to start with Ed, do you actually see when you're sitting in a board meeting, somebody presenting evidence or just just some loud mouth like me?

Edward Marx:

No, I, I think over the years it's become more evidence based arguments for sure. But you know what? One thing we instituted was the only evidence accepted came through our official data capabilities as an organization. So you couldn't just make up your own stuff like you might have in the past because that's what people would do. Well intentioned people, nothing, nothing nefarious going on, but you just didn't know if you could trust the data. But the short answer is yes, it's more and more evidence based data approach decision making. But the secondary thing is you have to ask yourself where did that data come from?

Junid Kalia:

So now the question is that when I update a policy, how are you going to do that? So I'm just saying what, what are you, what are your experience been in that regard

Harvey Castro:

It's very interesting in that it's very central in the sense, like, here's Singapore here, the studies, here's the evidence. This is what we have. And it's pushed from top down. And that has a lot of weight because it can be said to the entire country, this is what the vision is, this is what we're going to do.

And so there's pros and minus on that. But for me there's a lot of FOMO because I'm like, man, they have this widget, they can pull it off.

Junaid Kalia:

Dr. Kim, last question. Evidence generation through AI for use, case scenario policies and decision making from a, you know, governance point and evidence generation, real life for clinical guideline improvement. Thoughts?

Andrew Kim:

I think clinical will always be tougher, because that's going to obviously require a lot of review. But I mean part of my, my role is to create, or I guess generate and retrieve data for evidence for our operational needs, any small, like you said, the local operational changes that require, our leaders come to me and say, hey, I would like to know how this is going and I would like to know, to get a baseline idea of where we are. And then, we trial some sort of methodology and then six months later we look at it again.