Healthcare Providers & ChatGPT Will See You Now: Our Brave New World

Announcer (00:00)
The following is an encore presentation of Conversations on Health Care.

Mark Masselli (00:08)
Is this the future artificial intelligence that tells surgeons what to do, AI that can diagnose what ails you, will we
even need healthcare providers anymore?

Cris Ross (00:18)
These technologies are, value-neutral, but usage is not necessarily value-neutral. Bad people can use good
technology for bad purposes.

Margaret Flinter (00:31)
Here to help us understand the future of healthcare is Cris Ross, the Chief Information Officer at the Mayo
Clinic, and a board member of the global nonprofit that serves the health IT community.

Cris (00:42)
I think everyone is trying to figure out how do we take advantage of this stuff that could very well be
transformational without doing damage? And I think everyone that I know is taking a responsible stance of
trying to figure out how do we learn by application of these technologies?

Margaret (01:01)
And this is conversations on Healthcare.

Mark (01:12)
Well, Cris, welcome to Conversations on Healthcare.

Cris (01:15)
Thanks, mark. Thanks Margaret.

Mark (01:17)
And Full disclosure, none of these questions were developed by using ChatGPT. They're, they're all, all ours in
original. So, in the, in the new level of disclosure, you know, Cris, you're, you're helping lead the tech
revolution, we're all experiencing right now. So let's, let's start with an overview. Generative AI is a type of
artificial intelligence that can create a wide variety of data, such as Texan videos, it's ChatGPT's underlying
technology. Can you tell our listeners who's behind it and how does it work?

Cris (01:53)
Well, large language models have been developed over a number of years by various, companies in different,
threads. So there's a clearly a large thread that came out of open AI and its sponsorship from, Microsoft that led
to ChatGPT, Google has a similar kinds of models in things called, Palm, for example, and Meta has Llama.
And then there's many other, things that go by similar acronyms from, smaller companies as well.

Margaret (02:26)
Well, there are reports, that healthcare systems, of course, interest of all of ours, are eager to use ChatGPT, but
also a realization that right now it and other AI chatbots, are not good at everything. And one, somewhat to my
surprise, was that they're not good at spatial reasoning and math tasks. What's your take on this, and where are
we going with this innovation? And is that going to be resolved, do you think? Will those concerns be resolved
in the near future?

Cris (02:55)
The technology is advancing very rapidly, but look for right now, even though these models work well on
domains in addition to language, they're primarily being applied to language. All of these models are highly
fluent. They're very good writers and speakers. They speak in full paragraphs. They use good grammar. They
don't leave hanging sentence fragments, but they hallucinate in the absence of clarity. Sometimes they will
simply make things up. And that's where at least one class of problem arises. In addition to the other kinds of
problems, that you indicated, there's a lot yet to be done before these are ready to be exposed for clinical care.

Mark (03:39)
And yet, organizations like the World Health Organization recently issued a statement expressing concern about
ai. And the healthcare had said that, caution would normally be exercised for any new technology, but it's not

being exercised consistently with, with the new ChatGPT and other AI tools. And, and we've also heard across
the board that, a number of tech leaders have come out and called for a six-month, moratorium on the AI
development. Not sure that's practical, but certainly it's a reflective of the societal fears. Tell us, should we be
worried?

Cris (04:21)
Well, perhaps worried. You know, these technologies are, value neutral, but their usage is not necessarily valued
neutral. Bad people can use good technology for bad purposes. So I think there's a, a very robust debate about
whether these technologies should be regulated, whether they can be regulated, and if they are regulated, how
would you do that? So, to a large degree, these are being managed at this time in a coalition of the willing, if
you will. Now there's different regulatory and even statutory regimes that are emerging in different places, and
we'll see what happens. But at the moment, this is a bit of a, we're being governed by our good conscience kind
of world.

Mark (05:14)
I noted that the, Senate, was holding hearings the other day. Thinking through some of the issues, it seemed to
be more coming up to speed on the, on the issues. Are our policymaker capable of, managing through these
issues in such a new area of technology? That's a good question. I mean, it's a huge lift for even, you know, the
most capable of policy makers. And as you know, we're living in a somewhat polarized world nowadays, where
reason is short and heat is high.

Cris (05:50)
I, I don't know. I think I'm somewhat persuaded by the argument that says that, if one nation, you know, Europe
or the United States, regulates, that's not going to stop activity in other countries as well. And these technologies
are fairly borderless. You know, the things that will come into place will be things like liability law, that, even
though we may not be able to regulate what happens with AI in other parts of the world in the us, there could be
liability laws that would hold someone responsible for, appropriate and ethical use of these technologies.

Margaret (06:34)
Well, Cris, you're on the board at HIMMS, H-I-M-S-S for those, outside this, arena, which stands for Healthcare
Information and Management Systems Society, represents the big players, in health IT. You know, health IT has
been around for a long time. I'd say, you know, the last 25 years in particular, we've seen the transformation
from a rarity and practices to pretty full adoption across the country. And if there's big innovations coming, I'm
pretty sure they're on your agendas, and talking about it. What are some of the big, innovations that you're
tracking, beyond, what we've just talked about that you think is really gonna take healthcare forward in terms of
health information health IT?

Cris (07:17)
Sure. So, HIMSS is a membership organization with a hundred and twenty-two, a hundred twenty-five thousand
members around the world in about five dozen chapters and communities. And those chapters and communities
are active with their own local areas of interest. And HIMSS in particular has been very active in Europe in the
last several years. And we've been really pleased to see all the thought leadership coming from our colleagues,
in Europe. I think everyone is trying to figure out how do we take advantage of this stuff that could very well be
transformational without doing damage? And I think everyone that I know is taking a responsible stance of
trying to figure out how do we learn by application of these technologies into things that might, not be in the
clinical space, might be the administrative space. And that might be things like, in my organization, we're
evaluating whether our IT help desk could be, enabled with chat tools, for example, or maybe other kinds of
things that aren't rendering a clinical judgment, or guiding someone in a direction, but helping us with small
utility tasks that make our lives easier and allow human beings to augment, be augmented with those tools so
they can focus on the things that human beings are best at, human being clinicians that is, and what they're best
at in treating patients.

Mark (08:47)
You know, speaking of your organization, Cris, we're seeing, from Mayo a platform that breaks down language
barriers and allows for sharing of large and complex types of data. I take it the goal here is to detect diseases
early on. I'm wondering if you could just fill us in on how, how all of this is happening and what you're seeing
as, any of your, positive results.

Cris (09:12)
Yeah, so there's a couple things I could highlight. I'll be very brief. One is we're applying AI tools and not just
these generative AI or large language models, but some of the other models have been used for a while, like
convolutional neuro networks and, and deep learning models and so on, to things like diagnostic and
identification of heart disease, of, neurological diseases through speech patterns, other things like that. And we
found that we can get to pretty high level of diagnostic, accuracy. And there is a, regulated process of bringing
those technologies through the FDA approval process as software, as medical device. We’re also using it for a
variety of, as I said, helper tasks. So for things like, dictation, for example, you know, now it's become
commonplace. We're all using GPS, we're all using dictation. but those are AI-based tools that let us work more
effectively and with, less hassle. Mayo Clinic has a broad, approach under our CEO's, vision, to create a world
of, Mayo Clinic platform, which is a, a, a fairly sophisticated idea about how can we become a platform that is a
place where these ideas can be transacted for the benefit of many people beyond our brick and mortar walls,
who might want to, who can benefit from Mayo Clinic Care, where they can't come to one of our facilities. So
there's a, a variety of ways in which we are enabling these digital cures and digital diagnostics, and spreading
them, beyond, the confines of our traditional, four walls.

Mark (10:55)
You seem to be of the size where you could develop your own language learning model. Any, any effort in that
area?

Cris (11:03)
Maybe. So, you know, the ones that have gotten the most, public, view have required tens or hundreds of
millions of dollars to build. There’s some really intriguing evidence that says that highly focused large language
models could be trained on smaller data sets and have a high level of fidelity. They would just have to be more
for focused use as opposed to more broad use like ChatGPT is intended to answer any question from anyone.

Mark (11:32)
More of a proprietary system that could focus in on, on, might, might be of interest. Thanks so much.

Cris (11:39)
Possibly.

Margaret (11:40)
Well, we appreciate, we appreciate your national, spread, and certainly you're providing expert guidance to
many healthcare organizations out here on the East coast in our, neck of the woods, and appreciate that talent
being shared. But I wanna ask you, Cris, with your, your, long and deep experience with electronic health
records. Public health emergency is over been deemed to be over the emergency period anyway with Covid.
And we learned so much throughout the pandemic, but we also saw how many weaknesses we had in our
healthcare system. And while people did an amazing, extraordinary job of stepping up to the challenge, we saw
how hard it was to share information, electronic health records, often still can't share information with each
other, and sharing information with government public health, entities was really challenging. What does
HIMSS and, and you personally, I think it's going to take to achieve universal interoperability for healthcare
data and health records?

Cris (12:41)
So you're absolutely right that we saw lots of weakness, and in some cases failure points with public health. And
a lot of that has to do with the fact that we have systematically underinvested in public health infrastructure for
decades. People might think of public health as being like the Centers for Disease Control, as if there is some
national authority. And in fact, the CDC does amazing things, but most public health work is really delivered at
a state level and in some states at a county level. And, they've just not gotten along end of the financial stick
over a long time. So infrastructure was lacking. I think the majority of the problems that we have seen are
slightly less technical than they are sociological, financial, political, and so on. We have the means to
encapsulate data in effective ways, to make it, readable in multiple locations. We've figured out how to develop
new protocols for managing, that data. And it's now, been widely adopted. The remaining problems are, are a
little bit more operational, political, financial, sociological. Not to say that those are easier, in some ways,
they're harder in almost every way. They're harder. and, and I think that's where our focus should be next.

Mark (14:03)
Well, it's a great observation about the public health system of its spread. It's not just located in Atlanta. It’s, it's
all over the country. You know, I was thinking about, the role that you have. You mentioned your great, CEO

leadership and always focused in on both the mission and the margin, right. Of the work that we're doing. So tell
us a little more about, tax ability to save money, and are, are we getting cost savings that have, have been
promised? And I think we hear this all the time, this expenditure will help, both our mission and our margin.
What, what's your sense?

Cris (14:40)
Yeah, I think a lot of your listeners might remember this idea that was pretty popular in the 1970s and 1980s. It
was this idea of the productivity paradox. And it was this observation by Herbert Simon economist and, and
computer scientists who said, you can see, the, computer age everywhere except for in the productivity,
statistics. And it's, it's true. We had a, a point in time in which we were pouring lots of money into technology,
and we weren't seeing the requisite or, or expected lift in productivity. Now, that began to change in the 1990s
and the two thousand, essentially through accretion of technology over time. So I think applying that lens to our
current healthcare world might be helpful, because while we've invested in a lot of first generation technologies,
they haven't become interoperable. They haven't created full digitization, they haven't gotten to the level of kind
of transactional liquidity that one associates with leading tech, industries. So I, unfortunately, I think we're still
more in the investment stage right than we are in the accrual of benefit stage. But I think we can be hopeful that
we will get there.

Mark (16:02)
Hope springs eternal.

Cris (16:04)
Always.

Margaret (16:07)
Absolutely. And Cris, if I may, it seems you bring a newfound personal view to your work. You've spoken
publicly about your own healthcare and, cancer diagnosis. I’m wondering, speaking as, a patient as well as the,
health information technology expert in leader that you are, how has that influenced your thinking and your
observations, as we're seeing the, the big artificial information breakthroughs and, and really just the state of
healthcare, as it is in this country right now?

Cris (16:38)
Boy, do you got a minute? A colleague and I, Ed Marx, have been, are in the final stages of completing a book,
around our patient experiences. I've had cancer twice. He's had a near fatal heart attack, which he got while
running a tri triathlon go figure. And a serious cancer. And look, we know the patient experience can be better
and should be better. I think. You know, I tried to really learn from my passage through cancer. You know, I
didn't want to get cancer. The only thing you can do is re decide how to respond, not decide whether you're
gonna get it. And my decision to respond was, well, one thing was I was gonna try to learn from this, figure out
what, you know, what can I learn that I wouldn't get, if I wasn't lying in the hospital, better in the chemotherapy
infusion center and so on. So I've been trying to pay attention, and I think, there are so many ways that we do
miracles every day. But there are so many frustrations and agonies and disappointments that come with
healthcare. So what I'm hoping to do is, in a humble way, share, you know, Hey, this is what happened to me
and maybe this can, perhaps be a little bit of a guide to the next person down the road. And then whoever comes
after me will write the really good book, that'll be the guidebook, that people will use maybe to make their
healthcare journeys a little bit better.

Mark (18:05)
Well, we were just saying before the interview, I, is there a book out because, with your vast experience, there,
there, there are many stories to be told, and we look forward to that and sharing that with our, audiences when it
comes out. Cris, you know, one of the things in our practice that we've been focused in on is certainly remote
monitoring of patients and what a Mayo Clinic's innovation is at home hospital care program. I'm wondering if
you could take us through how that works, and again, sort of getting to the KPIs, what, what outcomes are you
seeing there?

Cris (18:37)
Yeah, so Advanced Care at Home launched, you know, sort of just in time. We've had things like a telescope
program that go back 20, 25 years. But we were able to, with partners, put together some infrastructure through
Mayo Clinic platforms, to take care of patients with relatively high levels of acuity at home. And we really tried
two different pilots in parallel. We had been sort of working on them in, in, in parallel, and so pursued both, one
of which was employed, in, Wisconsin for treatment of covid patients. And, the analysis, was, was quite strong

that the quality of care for patients who were treated at home was the same or better than hospital care. But
patients certainly, appreciated an opportunity to be at home rather than in a hospital bed. And it made bed space
available to people with particularly high acuity. So that was very promising through our Advanced Care at
Home command center, which runs out of our Florida operation primarily. We are now treating patients with
reasonably high levels of acuity. It's fascinating to watch. doctors and nurses round in the morning, just like they
would in a hospital facility. There's screens on the wall and they move from patient to patient to patient. They
have a conversation with the patient and their caregivers, as needed. There's ways to do outreach. There’s
monitoring in place at all times. And of course, there's an opportunity that if acuity requires that a patient can
come into, one of our hospital facilities. So we see more and more and more of that. The data is really pretty
interesting.

Mark (20:29)
That's Interesting.

Margaret (20:30)
Well, I think that is absolutely fabulous. And I, there certainly are patients for whom wouldn't be appropriate,
but an awful lot of patients would say, absolutely. Sign me up. This would be great. Right, right. You know, on
a less dramatic level, maybe, telehealth in primary care, has really boomed, since Covid caused, so many
patients and providers to move to online visits. We see this in behavioral health overwhelmingly, but also in in
primary, medical care. And I'm curious of your, work at Mayo at this point, and also what you're hearing at
HIMSS. Most of the outcomes seem positive as long as you're appropriately for what's an appropriate telehealth
visit versus what really should be in person. But there's some studies that suggest there may be clinical
outcomes that aren't so obvious.one was that patients who, went to the office for the care were more likely to
receive and adhere to some medications such as statins for cardio, vascular disease. I know you're, a tremendous
research institute. What are you seeing internally and how do we get that nuance about really where are we
gonna get the best outcomes now that we don't have the threat of, covid emergency hanging over our heads and
can really just focus on what's the best care and where?

Cris (21:45)
Yeah, that's fascinating about, differential compliance rates. I have not read those papers, but it makes intuitive
sense in some ways. So, you know, Mayo is not, our focus is on the treatment of complex and serious illness
and, doing diagnostic procedural work and other kinds of things a little bit less than our primary care, practice.
We do primary care, but it's not really the, our, our major in some ways. So I, I don't know if I can comment as
much on sort of that routine basis, but what I can say is we have found, video visits to be highly effective on
things like surveillance or WellCare or regular, maintenance kinds of activities where there's already an
established, physician patient relationship.it avoids someone having to travel to one of our locations. And we do
have a lot of people who travel to come to a Mayo Clinic. You know, I think about, you know, my
circumstance, for example, you know, occasionally I will come in and get a CT scan or a colonoscopy ‘cause
that's what's needed for, for my cancer care. But a fair amount of time, you know, I might have a question from
my oncologist about aftercare or something. Well, there's no reason why I would need to travel for that purpose.
I have a relationship with my oncologist.so it's a pleaser for patients in, in some ways I wonder if it might
improve compliance in some ways. Margaret, the other side of the coin, I don't know. Because people don't,
wouldn't delay care because it's easier for them to do a video visit at their convenience as opposed to having to
take time off of work or travel or something like that. So, much more to come. But I, I would probably also say,
you know, I made the point about interoperability being less of a technical question and being more of a
sociological one. You know, I think part of the reason why virtual care took off is ‘cause we all got used to
being on Zoom calls like this absolutely pandemic. Right? And so it doesn't seem weird to do a video call where
maybe it would've in 2018 or 20.

Mark (24:00)
Absolutely. The world really changed. And, Margaret, we see it in our practice, large, population, with financial
struggles. The, the notion of taking time off for work to drive an hour to wait 15 minutes to see somebody for 15
minutes to drive an hour doesn't make a lot of sense. And we've seen a lot of uptake in the behavioral health
side. You know, Cris, I want to get back a little to the issue of AI and also the future role of doctors, nurses, and
healthcare providers. Some of them for many years resisted having to fill out electronic medical records and
now there are reports of AI evaluating surgeons in the, or maybe helping on some of the filling out of, of, the
electronic health record. What do you see for the future of the medical professional, given that we're on this
cutting edge of AI and yet we might have a group that's, I won't say Luddites in this, but they've been some
resistance over the years that we've seen that have, there were certainly great leaders that we know about, but as
a field, they've, they've been behind the times.

Cris (25:07)
Yeah, it's a great question. You know, I think most of our physicians, those who are might be Luddites,
suddenly get religion when they find something that can make their life easier and more power to 'em, that's the
way it should be. The tools we've given them are not very good and we need to keep improving 'em. You know,
one of the things that's really exciting are the voice vendors that are now doing this idea of ambient listening,
which is to have a machine listen to a conversation between a doctor and a patient. And from that potentially
generate a visit summary or an after visits summary that, describes the dialogue, between the physician and the
patient. You know, they, they've been in highly constrained, somewhat artificial test conditions, but nonetheless,
there's enough there that suggests that could really be promising. Because here's the thing, no doctor or other
clinician likes to click boxes, but all that data that's collected is used for public health, for research, and for the
downstream treatment of that patient. If I'm a physician, I'm not clicking all those boxes for me, I'm clicking
those boxes for the patient and for patients like them and for the other physician who's gonna treat that patient
who's gonna benefit from the time that I took to record the data. So, you know, it's a little bit of a social contract
provider passing patient from one to another, but how can we make that easy enough and, and straightforward
enough so that physicians feel like they get joy back in practice and that they can feel like they can be in
dialogue with their patients and they can use their cognitive capabilities and not be scribes and I And, and, and
that's what we have to aspire to.

Margaret (26:51)
Well, that is such an important point, and you know, you, something you said there, queued up a question I've,
wanted to ask you about. All that information in the little boxes actually can be used to good end, when we pull
it together. You know, as healthcare providers, we've had our MPI, our national provider identification, for
years, where do you stand or maybe HIMSS stand on, having a national patient identifier that could be used for
uniquely identifying patient information, you know, on the theory that we would be vastly, better prepared to,
draw, informed, conclusions across large, enormous data sets.

Cris (27:34)
Yeah. HIMSS has consistently come out in favor of some form of secure and private patient identifier. You
know, it's, it's a, it's a difficult subject because people feel like they're giving up some privacy should they have
a number like that. And there's a number of kind of, concerns that raise from quite legitimate to somewhat, you
know, paranoid around those, concerns. And I get it. The fact is that we've accommodated that by using, instead
a combination of things we know about people that are in the public record that kind of help us with that
identification question. But it's expensive, it's cludgy potentially there can be errors associated. So, HIMSS has
been in favor of, patient identifiers.

Mark (28:24)
You know, speaking of patients, they're expecting a lot, of, of the work. They, in this sort of new error. I'm
wondering if we can keep up, with, the demands for efficient, effective, and elegant systems, that they're looking
for. What's your sense about it as you think about it from having been a patient? We're all patients, but, from
that lens, do you think our institutions are really, as nimble as they should be? They have the, ab ability, at least
in the healthcare space, he seemed probably a little more efficient in the consumer space, but how, how do you
think we're meeting patients' demands?

Cris (29:04)
Well, you know, mark, I used to think that healthcare was about the slowest and worst digital industry until I
recently tried to get a, I'm in the process of getting a, a mortgage to buy a second property. Holy smokes. Oh,
no. I didn't think it could be worse than healthcare, but it turns out it can.

Mark (29:22)
Finally, We, we have somebody we beat.

Cris (29:25)
Yeah, exactly. There’s so much more that we can do to make our systems more effective and efficient for
people, especially because healthcare is emotional. It's complex, it's personal. You know, it's sometimes
difficult to, you know, encounter to, to, to take on board everything that you need to know about your health.
And, and I know a little bit about your work. I've been an admirer of CHC and, and the kind of work that you
do. And, and it doesn't matter if it's behavioral, if it's a common chronic or if it's some sort of acute thing.
Health, it's hard for most people to understand and they don't fully comprehend it in lots of ways. So I think the,

the bar needs to be higher, and we can do better in making good systems for patients and for the dialogue
between patient, provider and other caregivers.

Margaret (30:20)
Cris, thank you for joining us. And thanks to our audience. There's more online about conversations on
healthcare, including a way to sign up for email updates. Our address is chcradio.com. And Cris, again, thank
you so much for joining us and sharing all your insights and experience with us.

Cris (30:37)
Thanks, Margaret. Thanks, Mark. I enjoyed it.

Mark (30:39)
All right, great. And, and make sure we get a copy of that book so we can profile it.

Cris (30:44)
You got it.

Mark (30:44)
Alrighty. Have a good one. Thank you so much.

Margaret (30:51)
This copyrighted program is produced by conversations on healthcare and cannot be reproduced or remitted in
whole or in part without the express written consent from Community Health Center, Inc. The views expressed
by guests are their own, and they do not necessarily reflect the opinion of conversations on healthcare or its
affiliated entities.