This transcription is provided by artificial intelligence. We believe in technology but understand that even the smartest robots can sometimes get speech recognition wrong.

[00:00:00] Today on Town Hall

(INTRO) there's the benefits that ambient AI brings, but there's also the, what does it do to my thinking? Because the AI is doing all the notes, do we think anymore?

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 town hall show is designed to bring insights from practitioners and leaders. on the front lines of healthcare. .

Alright, let's jump right into today's episode.

Welcome to AI Insights, a production of This

Week Health. My name is Dr. Al Valerian, and I'm your host for AI Insights. For this month's podcast, we're honored to speak with Dr. Mark Weisman. He's a board certified internal medicine physician and in medical informatics. He's a certified CHIME executive.

I'm trained at the Harvard School of Corporate Learning, FHIR, Centura, ACMIO, and currently [00:01:00] the CIO and CMIO at Tidal Health. Mark, welcome to the program. We appreciate you being here.

Hey Al, thanks for having me on. Appreciate it.

Tell us a little work about your work around title health and what you're doing today.

Yeah, so I'm the CIO there and I'm also the CMIO, which is great. I consider myself a CMIO with a budget. I have a team of about 160 people and I get to really focus on that clinician experience, which is always what drove me into this. That's what, that was my passion is. Years and years ago, I noticed that the doctors were struggling with epic and I wanted to get in and help them.

And that was always my motivation. And I've been able to continue doing that now. But with title health, I can do it now for nurses and respiratory therapists and just. Across the board and even in the back office. I can make try to make life better for finance people and for rev cycle people and other areas that I just never even thought about before.

And you didn't even see them. But now [00:02:00] I get to try to make them more efficient to try to bring a new insights to their work. really love what I do. And glad I get to do it for title help.

That's terrific. Tell us about Tidal Health. How large is it? How many states, etc? And how much territory do you cover on patient life, etc?

So we're on the eastern shore of Maryland. So I was at Sentara, which was like Virginia Beach, Norfolk area. If you take the Chesapeake Bay bridge tunneling, go north. That's the next territory. All that territory up the eastern shore of Maryland, Delaware and a little bit of Virginia in there.

We're right now 2 hospitals and we're looking to merge with a 3rd. By May 1st, I believe we will be 3 hospitals. And that's always fun because I get to take another hospital up on Epic. So it's a fun project and get to add them to our ERP system and it's all exciting stuff.

450 beds. I think roughly when we're done, roughly 6000 employees and a mess load of clinics and residents and [00:03:00] medical students and all the things you'd expect in a kind of a community academic center,

a very complex environment for sure to work in, especially to bring in new technologies.

And now our discussion about AI. How is that going to influence your applications and healthcare today? And how are they transforming your perspective of patient care can be for the future?

It's been an interesting journey. There's we've had our successes in our and our flops as well. And I'll be transparent here.

I'll air our dirty laundry on the models where I'm like, yeah, that really didn't go like we had hoped. think many people have experienced the epic sepsis version one model And it just made a lot of noise if part of the model is designed so that, hey, if you gave the person antibiotics, it's more likely to say this patient had sepsis.

Thanks, Captain. Obvious. I've already given him the antibiotics. I knew they had sepsis already But to epic credit, they do have a version 2 of their sepsis model, which we are installing now. And so they've recognized it's a more advanced model.

It's not just a [00:04:00] linear regression model. I do think that they've come a long way, but yeah, we've got our lumps. And yet we also have our successes, particularly I would say in the ambient AI. I know a lot of health systems are looking at this, and I think we'll have a good conversation about what do we do with this new ambient AI tool and how is it going to change our world?

But we'll get to that. think but that's some of the exciting things that we're doing and some of the lumps and bumps that we've had.

From an AI perspective utilizing it to improve diagnostic accuracy and speed, what are some success stories you've had utilizing AI to improve the diagnosis or the speed to care for your patients?

I'm not sure I have a lot of successes here. I got more lumps than successes.

I think that's a good message for other CIOs and CMIOs to be aware of is. Hey, check your vendors, although they say it's wonderful when you, the actual implementation and the informatics behind this, it really matters. Just putting in these algorithms is not successful.

Clinical deterioration index is [00:05:00] another one where. We put it in and it's Pinging in nurses saying, Hey, your patient sick and the nurse is no kidding about 30 years experience. I know what a sick patient looks like. And yet, for the newer nurses, perhaps. Okay. You would hope it would help them.

And they're also going just got this patient up from the 10 minutes ago and now you're hitting me with an alert. It's the informatics on how and when the timing should go and who gets notified and what's the next step that they take. None of that's built. It's here's a model.

Go implement it. And there's an alert that's going to fire and we expect magic to happen and there to be less codes on the floor. It just didn't happen. We looked at our numbers like it didn't move. So now we're looking at reimplementing this and going back and saying, okay, what's the right informatics on this?

What is the workflow that needs to happen? And I think a lot of my CMIO colleagues are going to go. No, duh. This is what makes these programs successful. We took approach more of let's put it in and let's have it hit nurses. And we know that if it's fired, it's going to do great things.

And [00:06:00] that just isn't how this works. And that's the mindset that's out there. I've got a colleague who's pinging me saying we have to get this new AI tool for diagnosing detecting cerebral aneurysms. And the data says maybe that's the right thing to do, but hey, we're going to detect some three binaurisms that probably were never going to become symptomatic.

What about that side of the equation? And so one of the things I'm most proud of is actually setting up some AI governance. That's really been a huge advance for organization. I took 12 doctors. One PhD and we all took the Harvard AI and healthcare course. was a eight, ten week course over the summer we got an education on some of the basics of AI.

About the practical application and looking at the bias. And how these models could be trained. And does the patient population that we're dealing with look like the one that's in the model that was used to train?

Those questions matter. And we never asked those questions before. we [00:07:00] brought in an AI model. I'll give you the story that stimulated my journey into governance. The cardiologist wanted this tool and it helps them look at what percentage block to the coronary artery is calculates something called an FFR and today they do that by putting a catheter across the lesion and measuring pressures and takes time and there's some risk and they don't do it every time.

They're like, wow, this is a tool. It's going to give us the answer. It'll give us the answer. Every time it's going to be perfect. great. The tool was trained in about 350 Israeli men. And the eastern shore in Maryland. It doesn't look like that. Our cardiologists were sold. This is going to work. And after it's been in a little bit, they're like, yeah, maybe it's okay. Tells us some things that we already know.

It's put the brakes on these things and really assess. Is it going to do all that? It's toting to do because we're not seeing now. On the other hand, I think some of the computer vision things around radiology. I think those are going to be really good. We've seen computer assisted mammography for a long time.

I [00:08:00] think that's been out there lung nodules and some other of the incidental findings that using the AI or the radiologist. They're really not keyed in on that, but the AI may say, Hey, look over here. There's something over here you probably want to look at. I think it's helping with the speed and efficiency of radiology.

So we are looking at those. Our challenge is I want to stop buying these point solutions. This tool finds this one ditzel. This other tool finds this other ditzel from a different vendor. It's I want a platform. So I'm starting to look at this more from the vendors that are putting together the platforms.

And starting to reach out to them. And I think other CIOs are doing the same.

I love that. Presented two great points here. One is the governance. Everybody's working to facilitate that structure bringing in compliance and legal and clinical and research and data governance as long as you are within the parameters of caring for the patient and you're looking at the potential for hallucinations, misdiagnosis, et cetera.

From [00:09:00] a realm, I think and I agree that AI is really an augmented intelligence, like what else can we help you with? To apply for the patient's care, and then how do we get there? So the other point you made, which I love is about education, right? How are we educating the clinicians of the future?

How are we bringing that into the medical school, the graduate school, residency programs to allow them to facilitate new technologies, just like the stethoscope was brought in 100 years ago, whatever it was, today we're bringing in AI and how we want to enhance the care of the patients we deliver.

So that is all, we're all learning how to do that as well. And I think It's our responsibility as CMIOs and CIOs to support the GME education within our institutions to align technology with best practice and develop great doctors for the future and nurses and whoever else is taking care of our patients is utilizing this new technology on a regular basis at the bedside.

my message to other CIOs is invest in the education, send a group off [00:10:00] for more education on AI and maybe you bring someone in or maybe you send, there's plenty of training programs out there. I'm not sure you have to spend the 3000 per. Doctor that I did. But you know what?

I needed to attract the doctors. This was the Harvard course was very well put together when they put on a very professional product. And I really didn't want to do a Coursera for this. I really wanted the doctors to have a valuable product. My limitation here. I didn't have a lot of nurses.

I didn't reach out to enough nurses. And so now I need to go back. I need to do the same thing in nursing and in other patient care services areas. Because AI needs to be more than just for the doctors. And I also haven't done it so much for the administrative side. And so as we start to look at back office tools and Microsoft Copilot and does everyone get a license on that?

it really adding value? It just has expense. Does it add efficiency? No one really knows. No one's got brilliant use cases that I've seen. We're like, Oh, my God, this thing we were able to eliminate FTES. No, that's not gonna happen. It might [00:11:00] make people more efficient. But one of the things that we're seeing the governance team really should look at this people who are very experienced with clinical medicine, we find are using the AI tools very effectively.

They are the ones who though, maybe the most resistance to it are more seasoned doctors are sometimes like, I don't need this thing, but they're the ones who recognize when the tool is wrong. I'm more worried about the first year resident using AI. All gung ho about technology has no clue if the tool is right or not because they just don't have that clinical experience.

So that's one of the areas where the governance team really needs to dive in and decide. Okay, that's going to be your most energetic user, but perhaps not the best user. So just food for thought for my colleagues out there.

Great points. I think we're all trying to find what's the best practice around who we're training and how they're being trained involving pharmacy, nursing, clinicians, are all part [00:12:00] of the group that are taking care of these patients. So let's apply that appropriately with it. governing parameters around how they use it, when they use it and why they use it. Why are they looking it up? Is it that they don't know something and want to apply to a patient?

Or they want, they've got a good diagnosis started. Is there something else that they want to look up and utilize for a different or a slightly different diagnosis that they don't want to miss? So really it's understanding using the tool on a regular basis, but not relying upon it. To do the work

that's what's so fascinating to me in terms of education and medical education in general.

So you and I had to memorize a ton of facts that we most of the time didn't need and were useless 5 years after we were less, but we got out of medical school like today when we're training these doctors and nurses, are we training for the memorization of facts or do we rely more on AI and now we have to train more on.

Is the AI right or not? Which requires a different [00:13:00] skill set. whole realm of ambient AI and just having AI write our notes. Do you learn and think when you're writing your note? If the AI writes my note for me, am I thinking or am I thinking better because I'm paying more attention to what the patient is saying?

I'm not trying to memorize or scribble notes down or type. And I'm, not a great typist, so I'm looking at the keyboard, I have no eye contact, so there's the benefits that ambient AI brings, but there's the, also the, what does it do to my thinking? Because the AI is doing all the notes, do we think anymore?

There's some literature out there that says you think a lot when you're creating your note. I don't know the answer, I don't have that brilliant, no one does it to the best of my knowledge, but it definitely will impact medical education. And hopefully we don't have to memorize as much facts if the AI is listening to our conversation that between me and the patient, then the AI can suggest a differential diagnosis.

Great. think that's phenomenal. I would [00:14:00] also like it to go ahead and take care of, hey, it heard me say, I want this test, go ahead and free off that test for me. And go ahead and schedule it with the patient. As soon as I'm done, the AI can start talking to the patient and schedule it.

Great. So there's lots of use cases I think that we're getting there. We're not there yet. That it should just be seamless though. I finish talking to the patient, the AI takes over, That would be wonderful.

It's great. Just like any new technology is the trust factor, right?

The patients need to have that trust. But how do you support widespread adoption and trust? Any limitations around utilizing AI and healthcare that you see currently that we have to overcome to make it more ubiquitous and further access to improving care?

There's a couple and I'll tell a story from one of my colleagues.

He's an endocrinologist and he's using our ambient AI tool and he's evaluating this patient with an adrenal gland nodule and he goes through the visit. And he's not always the most conscientious at reading the note after, but this time he's you know what I'm just [00:15:00] gonna, I want to see what the AI said, and it's talking about counsel patient on their thyroid nodule that they have trouble swallowing or.

Difficulty speaking to let me know. It's I do say that. That is something that I would typically said. I absolutely did not have that conversation with this patient though. Where the AI came up with that from, he has no idea. It hallucinated that entire part of the conversation. He was strictly focused on the adrenal nodule for that visit.

that does impact adoption. Cause there are some doctors like I can't take that risk. I know I'm not the greatest at proofreading my notes. Now they use dragons or one of the dictation tools, and they probably should be more conscientious because it's going to misspell or mistake things, but humans make mistakes too.

But that's one of the barriers to adoption is that trust. And I think the tools will continue to get better. And I know there's A. I. That checks up on the A. I. Model. So you have a I. The computer is checking up on itself. And I think that could [00:16:00] be very valuable to add that level of confidence in it.

The other challenge that I have is really with the vendors and it's the model which they are delivering these A. I. Models to us right now, particularly the ambient A. I. Where it's a named user That really inhibits adoption in my mind because I've got a surgeon who's in the OR two to three days a week.

The primary care doctor that's in there five days a week. They have more value to it. So do I not give it to surgeons? Do I not give it to residents who their clinic rotations may or may not have them in the clinic or on a ward where we're using the AI that limits adoption.

The most in my mind, most brilliant company in the world is up to date. They gave me as a resident up to date for free, right?

We're all hooked try to take up to date out of your health system and CMIOs who've done that, but not many.

If the companies had that mindset of let us [00:17:00] Get these tools in, let us adopt, give us an enterprise license. Number one is a CIO. I don't want to have to provision this one license at a time. I just want to deploy it and then work on the adoption from an informatic standpoint, not as an it standpoint going, am I going to get my money's worth out of this?

Is it, if they're going to charge me 450 per provider per month. For a named user and they're only using it once a week. Do I get enough return on investment? Because there's no return on investment, patient care. It's right. And that's the other question that we have to answer as CIOs is.

Are we a I first? Is that our mantra now? Is that okay? If there's a I out there, we're going to invest in it and try this. And because we know it's going to bring medicine forward. Or are we more? We're gonna stand back with some others. Take that leap, I think that's a conversation the CIOs have to have with their CFOs because the R.

O. I. really isn't there ask any doctor who's using [00:18:00] ambient AI on a daily basis to try to take it away from them. There's something there. I can't tell you that it's a return on investment in hard dollars, but they're retention. They're eventually these doctors who are going to join our health system aren't going to come unless you say that they have access to ambient day.

I the pajama time reduction, maybe not seeing that universally in our data, but there may be something there or. Maybe they're spending the same amount of time, but they're doing additional things. They're able to focus on the patients while they're in the room. Maybe they're doing more quality work after hours.

not exactly sure of the return on this, but all your CMIOs are going to tell you it's there. It's just soft and CFOs hate that.

that's happening quite rapidly. There's change going on. The industry has to prepare. How are we CIOs, Executives and healthcare. How do you see future advancements in AI helping us?

And how do we prepare for the future? That's really [00:19:00] within two years, a ubiquitous utilization of AI across the healthcare market.

If we're training residents, do we have an obligation to prepare them for practice in which AI is ubiquitous? Or do we have to say, no, there's no return on investment giving residents these tools.

So therefore we don't do it. We're a teaching institution. There's a cost of doing business. So I, me personally, I'm saying put the tools in the hands. Of our clinicians and I want our nurses to have it too. I think ambient AI for nursing is the next major change in health care and that nurses clicking in flow sheet rows needs to go away.

They come in, they do the central line care, either the AI sees them doing it using computer vision or the nurse verbalizes. Okay. I've just done the X, Y, and Z to your lines and it's now populated in the E. M. R. with all the correct. Mhm. Fields that need to be documented in and the nurse doesn't have to go click, and they can just talk to patients just like the doctors are now doing in [00:20:00] the exam room.

That is going to revolutionize nursing. And it's close I'm hearing of some beta pilots that are out there haven't seen it. It's not quite available commercially yet. I am hearing some announcements are coming out during HIMSS. So we'll be looking forward to some of those.

I think we have a very exciting near term future to embrace, adopt, manipulate, and enhance for the care we give, but then there's the future.

There's a next generation and next generation in our intelligence really is within a year to two years away of that next goal. And we have to help bring everyone to that future, not just the one who are want to adopt, but it's going to be common practice to point that potentially legal. If you didn't use an AI engine in the future, and you had a misdiagnosis, why didn't you do that?

Because you had the ability to have access to the MRI. Why didn't you get the MRI for that patient? When that came out, same thing with new technologies such as AI, embedding itself [00:21:00] within the construct of healthcare itself. Any final words, Mark? We appreciate time. Anything else from your perspective, and any final thoughts for our audience?

Just as the C. I. O. I think it's our obligation now to prepare our organizations for that new wave. So I'm focusing on getting the cameras and the microphones into the hospital rooms because that ambient technology is going to be needed there. The computer vision is going to be needed there. The ability to see.

Who is in a room and how often they're in the room and not so much for, oh, we're going to beat on people because they didn't get to the bedside. But you can help understand what staff you need and how many texts or aids you might need based upon data. And today we make these decisions without data. As CIOs, this is our obligation.

This is our charge is to start to get that technology. into place. will get pushed back. There are nurses who have come to me saying, why aren't we hiring more nurses? You're putting cameras [00:22:00] and microphones. One, it's hard to find the nurse. Number two, it's hard to scale that. Whereas I think the technology component, we can scale and it's a matter of getting people used to that technology being in the room.

That's something as CIOs that we can start to do.

I totally agree, mark. I think, utilizing the triad of clinical practice, always taking care of the patient first. The technology and where it fits into the clinical practice, but advancing technology and maintain the educational process moving forward so that trainees.

Residents and young attendings can facilitate better care, but also the change management. We have to be great change managers as we move technology forward in explaining the why around innovations. And that's another responsibility, another hat we wear. And I really appreciate you bringing all this material up to this this week.

Thank you.

Thanks, Al. Appreciate the opportunity.

Thanks to Dr. Mark Weisman CMIO and CIO of Tidal [00:23:00] Health for participating in this month's Al's AI Insights hosted by This Week Health, Town Hall. Thank you, and we'll catch you next month.

Thanks for listening to this week's Town Hall. A big thanks to our hosts and content creators. We really couldn't do it without them. We hope that you're going to share this podcast with a peer or a friend. It's a great chance to discuss and even establish a mentoring relationship along the way.

One way you can support the show is to subscribe and leave us a rating. That would be really appreciated.

Thanks for listening. That's all for now..