Latest / Becker’s Healthcare Digital Health + Health IT / Dr. Eric Cheng, Chief Medical Informatics Officer at UCLA Health Sciences

Transcript

Dr. Eric Cheng, Chief Medical Informatics Officer at UCLA Health Sciences · Becker’s Healthcare Digital Health + Health IT

0:00
This is Rosie Talago with the Becker's Healthcare
0:02
podcast.
0:03
I'm thrilled today to be joined by doctor
0:05
Eric Chang, chief medical informatics officer of UCLA
0:08
Health Sciences
0:10
and professor in the department of neurology at
0:12
the David
0:13
Geffen School of Medicine at UCLA.
0:16
Doctor Chang, it's a pleasure to have you
0:17
on the podcast today.
0:19
Yeah. Thanks for the invitation.
0:21
We have a great conversation in store, diving
0:23
into some of the most exciting trends in
0:25
health care, the innovations happening at UCLA Health,
0:28
and how leaders like today's guests are thinking
0:30
about growth in the year ahead. Before we
0:33
get into it, doctor Chang, can you please
0:35
introduce yourself and tell us a bit about
0:36
your background?
0:38
Oh, yeah. Sure. So what first, again, thanks
0:41
for the invitation. I've really enjoyed listening to
0:44
the podcast of some of my friends and
0:46
former coworkers. I'll just call out Mike Pfeffer
0:48
at Stanford and Clara Lynn at Seattle Children's
0:51
who have done your podcast.
0:53
I've interviewed some applicants for our clinical informatics
0:56
fellowship program, and many describe
0:59
a, like, childhood hobby in computing or an
1:02
engineering background,
1:04
but I don't have that. So I'll describe
1:05
my background in a bit more detail just
1:08
to show there's another path to my role.
1:11
So
1:12
I developed an interest in outcomes research during
1:15
my neurology residency.
1:17
UCLA is a national leader in health services
1:19
research, so I did a fellowship in that
1:21
area. And I spent over a decade as
1:24
a grant supported
1:26
health services researcher
1:27
measuring quality of care
1:29
for persons of neurologic conditions.
1:31
And I also volunteer for my specialty society
1:35
in developing
1:36
quality measures for regulatory programs.
1:39
I had
1:40
ventured into informatics
1:43
even though I really didn't know,
1:45
what that word meant.
1:47
So I'll give a brief,
1:49
example. So we were analyzing the VA national
1:52
database and whether blood pressures were
1:55
well controlled the year after a stroke,
1:57
and we showed that a stroke patient's
2:00
antihypertensive
2:01
regimen
2:02
frequently wasn't intensified
2:04
even though the recorded blood pressure in the
2:06
clinic was high.
2:08
But then we
2:09
reviewed some charts, and we read that many
2:11
physicians would remeasure
2:13
a high blood pressure. They would document that
2:16
value, which is typically a lower value, as
2:18
free text in their notes.
2:20
So at that time, I call that an
2:21
administrative
2:22
database error, but that's actually informatics.
2:25
To fully understand data on the back end,
2:28
you need to understand how it's entered on
2:30
the front end. So using informatics vocabulary,
2:34
there wasn't a workflow for physicians who enter
2:37
blood pressure values in the same data field
2:40
where the automated blood pressure machines were transmitting
2:43
it. So when UCLA purchased an EHR,
2:45
they sought some physicians who could help configure
2:48
it. I did my clinical work at the
2:50
VA, and we had implemented an EHR for
2:52
years.
2:53
UCLA also sought someone who could help interpret
2:56
the federal
2:57
regulations about meaningful use, and I was pretty
3:00
familiar with government regulatory programs based on my
3:02
volunteer work. So
3:04
I guess they thought it was a good
3:05
fit. It's been said, like, many times that
3:08
life doesn't
3:09
always make sense looking forward, and these aren't
3:12
the career choices I would recommend to someone
3:14
who would like to be a CMIO.
3:16
But looking backwards,
3:18
yeah, I can see how these experiences
3:20
led to my position and my way of
3:22
thinking.
3:23
Absolutely. That's so interesting, and it's great to
3:26
see another path, like you mentioned, opposed to
3:29
that more exclusively or assumed
3:32
engineering technical side of things.
3:34
And I found it very interesting that you
3:36
said you veered into informatics without truly knowing
3:40
exactly what that word meant.
3:42
But it's sounds like you bring a very
3:44
valuable combination to the table, especially
3:47
in today's health care environment.
3:50
Yeah. It's it's a so there's,
3:52
again, not a single path. And
3:55
maybe one of the reasons why I was
3:57
elected to be one of the first physician
4:00
informaticists
4:01
was that I was a little different from
4:03
the other physician informaticists who did have a
4:05
different background,
4:07
maybe more business or may maybe more computer
4:09
science related.
4:11
Yeah. So with your
4:13
new or fresh perspective
4:15
perspective in mind, I'd love to hear how
4:17
you're viewing the current
4:18
landscape. So what are the top three trends
4:20
that you're following in health care today?
4:23
Okay. Well well,
4:25
answer number one has got to be artificial
4:28
intelligence, of course. Right? There's so much that
4:31
people have said, and I follow,
4:35
you know, the developments of both excitement and
4:37
skepticism.
4:38
I'll I'll just make this brief comment.
4:40
Based on my health services research days, I
4:43
know that
4:44
there are
4:46
methodologic
4:47
flaws, especially like self selection bias, that can
4:50
make it difficult to interpret observational studies.
4:55
So when you really need confidence in an
4:57
answer,
4:58
there isn't a substitute for randomization.
5:01
So I'll describe a project here for Ambient
5:04
Scribes, very hot topic. We randomized 250
5:08
physicians to
5:10
two vendors and a control arm and then
5:12
crossed them over at set intervals.
5:15
So then we
5:16
got some insights on how they were the
5:18
vendors were similar and different from each other
5:21
and when they did or did not differ
5:23
from the control group.
5:26
The overhead to conduct a randomized trial
5:29
is too high to run for every AI
5:31
implementation.
5:33
But when a project has a, you know,
5:35
potential seven digit price tag, it is worth
5:39
slowing down the implementation
5:41
just a bit so they can study it
5:42
properly.
5:43
So I definitely encourage others to do that
5:46
as well.
5:47
The second,
5:49
I think, thing I'm following
5:51
is
5:52
related to AI.
5:54
So I'm cheating a little bit, but I'll
5:56
call it,
5:57
systematic
5:58
data collection.
5:59
So there's a well known
6:02
epidemiological,
6:03
like, study that shows that about twenty percent
6:06
of early mortality
6:08
is attributable to health care,
6:10
and the rest is attributed
6:12
to factors
6:13
we don't systematically
6:14
collect, such as patient behaviors, social drivers of
6:18
health, the built environments, and genetics.
6:21
So I'll tell another story from my research
6:23
days. We were wrapping up a trial. We
6:26
had some difficulty
6:28
getting the final survey results from some participants.
6:32
So I asked a statistician,
6:34
can't we run some fancy
6:36
imputation
6:37
imputation methods to overcome this?
6:40
And he looked at me and said,
6:42
you know, imputation
6:44
is certainly better than no imputation,
6:46
but there is no substitute for primary data
6:49
collection.
6:50
So likewise right now, if we don't collect
6:52
the right
6:53
predictors,
6:55
then the outcomes will be inaccurate due to
6:58
omitted variable bias, and AI can't overcome that.
7:02
What AI does is generate better predictions for
7:04
the data that we already have. And to
7:07
fully leverage it, we just need
7:09
accurate and more types of data. So I
7:12
view that as complementary
7:14
or kind of almost more fundamental to get
7:17
AI to to work at its best.
7:20
And then the third topic, I'll mention a
7:22
topic that maybe
7:24
no one else has mentioned in your podcast.
7:27
But in 2025,
7:29
it's gotta be government regulations. Right? It's it's,
7:32
it's the world we live in. CMIOs have
7:35
always needed to monitor this space, you know,
7:37
meaningful use, the twenty first Century Cures Act,
7:40
there's TEFCA,
7:41
so all the quality measures.
7:44
But it's not just at the federal level,
7:46
but it's at the state level as well.
7:48
So I'll give an example. So the federal
7:51
twenty first Century Cures Act states that results
7:53
need to be made available to patients
7:57
without delay,
7:58
but our state government passed a law stating
8:01
that if you have radiology results that show
8:03
new or recurrent malignancy,
8:06
that should be verbally disclosed instead of through
8:08
a portal.
8:09
So we developed a process for the radiologist
8:12
to manually tag such scans so they are
8:14
held from the portal for a short time
8:17
so that the ordering provider
8:19
has the opportunity to reach out to the
8:21
patients.
8:22
But because that process is inconsistent,
8:25
we just supplemented it with a homegrown machine
8:28
learning algorithm
8:29
that analyzes the text of the radiologist report
8:32
to determine
8:33
whether it meets a criteria
8:35
mentioned in state law and then holds it
8:37
back from the patient portal.
8:39
Now did I need a state law to
8:41
tell me that informing patients of a new
8:43
diagnosis of cancer through the portal
8:46
is
8:47
now potentially problematic?
8:48
No. But
8:50
I couldn't champion and prioritize
8:52
the developments of a machine learning
8:55
algorithm
8:56
if I didn't have the weight of the
8:57
state law behind me. So it is
9:01
important to follow, like, what's going on in
9:04
government and
9:05
possible, you know, really leverage it to drive
9:08
change.
9:09
Absolutely.
9:10
I particularly liked your example of the connection
9:14
between that systematic data collection
9:17
and artificial intelligence and machine learning
9:21
tools such as those.
9:23
I feel like a lot of times it's
9:24
easy to think of this artificial intelligence as
9:27
this higher
9:28
higher thing that's all knowing and super smart,
9:31
which it is.
9:33
But we gotta pay attention to the data
9:35
that we are feeding it because at the
9:37
end of the day, like you mentioned,
9:39
if we're omitting
9:40
variables,
9:41
that's bias, and it's very important for the
9:44
results that it produces.
9:46
Yeah.
9:47
People, yeah, others have commented
9:50
it as well in a maybe a slightly
9:52
different way. They point to the quality of
9:54
the data. If we
9:57
if the notes are somehow inaccurate, then certainly
10:00
the predictions are inaccurate. But in addition to
10:03
the quality of data, I think the types
10:05
of data makes, should be expanded.
10:08
Absolutely.
10:09
So zooming in a bit now, UCLA
10:12
is known for being on the leading edge
10:14
of innovation.
10:15
So I'm curious to know what are you
10:17
most excited about right now at UCLA?
10:21
So I've listened to this podcast, and a
10:25
lot of people mention,
10:27
burnout reduction, and I agree.
10:30
And people point to several different factors, typically
10:34
patient message volume and pajama time and EHR
10:37
usability.
10:39
And I agree with that too.
10:41
But I wanna highlight maybe a personal interest
10:43
of mine. It doesn't have a financial,
10:47
ROI, but I still think it's just as
10:49
important.
10:50
I think one of the reasons why it's
10:52
so
10:53
cognitively taxing
10:55
to use an EHR
10:56
is because the notes are just too hard
10:58
to read.
10:59
They're long. They're repetitive.
11:02
It mixes both past and current data.
11:05
The information that's useful to the author is
11:08
frequently not relevant to the reader. But but
11:11
I get it. You know, the clinic has
11:13
such time pressure.
11:14
It's faster to start from a template or
11:16
your prior notes than a blank page.
11:19
Yet despite its long length, it doesn't always
11:22
capture the essence of the patient.
11:25
Like many medical students who chose neurology,
11:28
you know, I was drawn by Oliver Sacks,
11:31
narratives of patients.
11:33
So when I'm reading a note now, I'm
11:35
not expecting literature, you know, but it is
11:39
hard to hear the patient's voice in many
11:41
notes. So if the Ambient scribe
11:44
can capture the essence of the conversation
11:46
while the upcoming, like, summarization tools can bring
11:49
in the relevant parts of the note, that
11:51
frees the physician to focus on the activities
11:54
we were trained to do, and people cite
11:56
that as a reason for better satisfaction.
11:59
But on top of that,
12:00
it could lead to a better experience for
12:02
the reader as well. We could achieve both
12:05
a
12:06
detailed rich note yet relatively
12:09
concise.
12:10
It's so nice to read a well written
12:13
note, and I like to find a way
12:14
to make that easier.
12:16
That's a great point. I do hear a
12:18
lot about how it can streamline the process
12:21
for
12:22
the providers and the nurses and the doctors,
12:24
but that's a very interesting perspective to bring
12:27
up for the reader and the patient bringing
12:29
the patient's voice into the notes too. Very
12:32
interesting.
12:33
Yeah. There there's a tradition. Neurology at these
12:36
long form,
12:38
back to Luria and Sachs of, like, describing
12:42
kind of how how a patient is, you
12:44
know, thinking
12:46
and doing. And,
12:47
yeah, it's maybe impractical in everyday use, but
12:50
I do I do enjoy reading things like
12:53
that.
12:54
Absolutely.
12:55
And finally, looking ahead, growth is always on
12:59
everyone's mind, especially in this industry that's adapting
13:02
to new challenges and opportunities
13:05
so quickly. So I'm wondering how you are
13:07
thinking about growth over the next twelve months.
13:12
So
13:13
I'll think of it from the perspective of
13:15
the organization.
13:17
UCLA has always
13:19
seems to be expanding their outpatient footprints,
13:22
but they've recently
13:23
expanded in other ways as well.
13:26
I've been at UCLA now for thirty years,
13:29
and
13:30
we purchased new hospitals for the first time
13:33
in the in the past year since I've,
13:35
since I've been here.
13:36
We just implemented our EHR there. We're gonna
13:39
open a new psychiatric hospital next year.
13:43
We're implementing our EHR in our university student
13:46
health center this summer.
13:49
We just rolled out our
13:50
managed care medical
13:52
Medicare Advantage health plan.
13:54
So we're expanding
13:56
into
13:57
lots of different areas other than opening up
14:00
outpatient clinics.
14:02
So I think that's probably
14:04
a main kind of focus of activity. But
14:07
even if we weren't expanding,
14:09
there's a lot of work to keep up
14:10
with, you know, developments from our EHR vendor,
14:14
from other third party vendors, and from our
14:16
own community as well. And I I guess
14:18
I wanna mention that. An academic medical center
14:21
is in the position to build some of
14:23
their own tools, especially when it comes to
14:25
AI.
14:26
And sometimes that's led by
14:29
the IT organization itself, but sometimes it's led
14:31
by our own faculty.
14:34
I'll say it's a lot. I just wanna
14:36
add that no one can do this alone.
14:38
There are great people
14:39
under me, next to me, above me, whom
14:42
I interact in
14:44
everyday basis, and that is one of the
14:46
most enjoyable aspects of my job. There's kind
14:48
of a lot of variety of the things
14:50
that I'm exposed to.
14:52
That's wonderful. And that's very cool to see
14:54
that you've been there for so many years,
14:57
but you just now are expanding in new
14:59
ways you haven't seen,
15:01
aside from the outpatient that you've been doing.
15:03
But seeing this expansion and growth in new
15:05
areas is very exciting and
15:08
also very exciting to think of that possibility
15:11
or the capability of an academic medical center
15:15
to have faculty led
15:17
new tools,
15:19
be in that position to build some of
15:20
their own tools is very, very exciting.
15:24
Yeah. I think because of that, we are
15:27
in the process of hiring a chief health
15:30
AI AI officer to help
15:32
kind of what's the best way to put
15:34
it? Maybe organize some of those activities.
15:37
You know, faculty, one of their strengths is
15:39
that they are,
15:40
you know, fiercely independent, but it it would
15:43
help to have some way to connect them
15:46
all together. And I think that's that'll be
15:48
one of the
15:49
functions of this, new,
15:52
person.
15:53
Absolutely. Like you said, you have great people
15:56
all around you. No one can do it
15:57
alone. So bringing all those great powerful minds
16:01
together to do it together.
16:03
Yeah.
16:04
Well, thank you very much. That is all
16:06
the time we have for today, but I
16:07
wanna thank you, doctor Chang, for sharing such
16:10
wonderful insights and experiences
16:12
gained from your experience at UCLA.
16:15
It's been really fascinating to hear about all
16:17
the wonderful things that are underway, and we
16:19
look forward to collaborating with you again soon.
16:23
Thank you. So look forward to that.