Latest / Becker's Healthcare Behavioral Health / Transforming Behavioral Health: Innovations in Crisis Care & AI with Tammer Attallah

Transcript

Transforming Behavioral Health: Innovations in Crisis Care & AI with Tammer Attallah · Becker's Healthcare Behavioral Health

0:00
Hi, everyone. This is Erica Carbajal with the
0:03
Becker's Healthcare Podcast.
0:05
Thank you all so much for tuning into
0:07
this episode.
0:08
Today, we're joined by Tamara Atala, executive clinical
0:12
director of the behavioral health clinical program at
0:15
Intermountain Health. Tamar, welcome. Thank you so much.
0:19
Thanks, Erica. Excited to be here. Yeah. Excited
0:22
for it. We have a really, really relevant,
0:25
timely discussion here today. Be behavioral health trends.
0:29
I'll start with one topic that really comes
0:32
up a lot in conversations
0:34
with health system leaders today
0:37
related to capacity pressure. So
0:40
emergency departments, as you know, often serve as
0:42
the de facto entry point for behavioral health
0:45
crises, which is not ideal for patients or
0:48
sustainable for health systems.
0:50
So can you start by sharing a bit
0:52
about how Intermountain
0:53
is working to ensure that patients
0:56
receive behavioral health care in the most appropriate
0:58
setting.
1:00
So, Erica, first of all, I couldn't agree
1:02
with you more,
1:03
regrettably for numerous reasons.
1:06
Emergency departments have become our defacto entry point
1:09
in the system. And what that really means
1:11
is that people are receiving care when they're
1:14
in crisis
1:15
and then trying to connect to care moving
1:17
forward,
1:18
which is really
1:20
a a point in time where,
1:23
people need support,
1:25
and they don't know where else to go
1:27
except the emergency department.
1:29
And although
1:31
there are benefits to doing that, it's certainly
1:33
not ideal.
1:35
And so at Intermountain Health,
1:37
there's several things that we think about when
1:39
we think about managing people
1:42
when they're at the point when when they're
1:44
in crisis.
1:45
First and foremost, we have developed across,
1:48
multiple areas in our footprint, primarily in Utah,
1:52
access centers. And access centers are really located
1:55
adjacent to hospitals.
1:57
And what they offer is,
1:59
is support that really meets their needs
2:03
in the crises,
2:04
help address the issues sort of directly that
2:07
they may have, whether that be a medication
2:09
change. It is served by nurses,
2:11
psychiatrists,
2:13
licensed clinical social workers in many cases,
2:16
to really collectively work together to support this
2:18
individual,
2:20
in the short run and then connect them
2:22
into services.
2:24
This does a few things. A, it provides
2:26
a really supportive,
2:28
environment,
2:29
which is much different than the oftentimes
2:33
chaotic attempts,
2:35
and important work of the emergency department that
2:37
may not necessarily help support someone as well,
2:41
when versus being in these settings. So that's
2:44
sort of part one. Part two,
2:47
it actually offers them better support at a
2:49
lower cost.
2:52
Thirdly,
2:54
oftentimes,
2:55
and we've demonstrated this in our data that
2:58
that the support they need actually helps
3:01
them be in a position to sort of
3:03
deescalate
3:04
sooner.
3:05
And as a result of that, connect them
3:06
into care in their community
3:08
as opposed to unnecessary
3:10
admissions into inpatient psychiatric admissions. And so,
3:14
and and that, you know, that benefits
3:17
the patient, the consumer,
3:19
as well as the health system trying to
3:20
help support,
3:22
those patients and
3:23
ensure that the emergency departments are really taking
3:26
care of what they do best in an
3:27
emergency situation. When someone is has a medical
3:30
emergency,
3:32
we wanna make sure that there's access for
3:33
those individuals there.
3:35
So that's how we deal with it at
3:37
the point of service.
3:39
In addition to that, Intermountain Health knows that
3:41
we can't do this on our own. So
3:44
we partner
3:45
with state
3:46
jurisdictions,
3:47
which also has sort of crisis plan work.
3:50
And so, for example, in the pediatric space,
3:52
we partner with the state of Utah and
3:54
our Utah footprint specifically
3:56
in,
3:57
in mobile crisis work, particularly with youth. So
4:01
there's a program called stabilization mobile response. And
4:04
really it's an opportunity
4:05
for calls to come in from families that
4:08
are in as they define their crises.
4:10
And we have teams that'll either deescalate them
4:12
over the phone and help support those families
4:14
where they are, or they will deploy someone
4:17
into the community to help support them.
4:19
And if appropriate,
4:21
provide
4:22
some ongoing,
4:23
supports for them in in in a sort
4:26
of a targeted,
4:27
you know, weekly basis in an effort to
4:30
help,
4:31
mitigate unnecessary crises in the future. And for
4:34
us, when people
4:36
are able to mitigate their crises, they're actually
4:38
able to stay engaged in treatment
4:41
in the community.
4:42
So I I I think just, you know,
4:44
our own internal innovations and really important in
4:47
doing that work, but also partnering with the
4:49
states and other partners in the community, I
4:51
think is really critical for us moving forward.
4:55
Yeah. Absolutely. So true around just the internal
4:58
initiatives and tackling it that way, but also,
5:00
you know, not discounting all the value that
5:02
comes from working with local and state level
5:05
partners.
5:07
Tamara, when you talk about the
5:09
access centers and those that are adjacent to
5:12
hospitals,
5:13
sort of supporting folks who who are in
5:15
crisis and and need immediate
5:18
care,
5:19
When did those,
5:20
really take off? Can you share any any
5:22
context on kind of the scope of those,
5:25
how many there are, and and just the
5:27
value that that those have really brought since
5:29
they went live?
5:31
Sure. So we have currently
5:33
we have three,
5:35
access centers
5:36
soon to be four. Those access centers have
5:38
existed
5:39
in our footprint for,
5:43
for
5:43
for not quite a decade.
5:45
So we've had these for quite some some
5:47
time,
5:49
and they've really been a service and a
5:51
support,
5:52
for the community. So,
5:54
and and we've seen the benefit of that
5:56
over time. I mean, in some regards, they
5:58
are,
6:00
sort of specific crisis centers really designed for
6:03
those that are in crises. In addition to
6:06
that, we also more recently,
6:08
built an observation center for our pediatric cases
6:11
through our primary children's hospital based out in
6:13
Lehi, Utah as well. So,
6:15
that that's more recent. But but this work
6:18
we've had for
6:20
for for quite some time, and we continue
6:22
to learn from that in terms of and
6:24
and and sort of evolve in the work
6:26
that we're doing in that space moving forward.
6:29
Yeah. Definitely. Sounds like Intermountain really was an
6:32
an early adopter there. I know recently we've
6:35
heard some more systems,
6:36
doing, you know, similar things and opening urgent
6:39
care behavioral health urgent care centers.
6:42
Yeah. Sounds like a very valuable initiative to
6:45
to keep folks out of the ED.
6:47
Tamara, I know Intermountain
6:48
developed a sophisticated
6:50
behavioral health risk model that helps clinicians identify
6:54
behavioral and mental health needs really as early
6:57
as possible,
6:58
enabling them to just intervene more proactively.
7:01
Can you walk us through how this model
7:04
actually works in practice
7:06
and what effects has it had on clinical
7:08
decision making, patient outcomes, or just overall care
7:11
coordination?
7:13
Absolutely, Erica. Let me just share, first of
7:16
all, the problem we were trying to solve,
7:18
and how we we can't really solve this
7:20
problem on our own. So sort of the
7:22
theme for us is we're not necessarily interested
7:25
we're we're we're interested in using data to
7:27
understand information,
7:29
but also recognize where our skills are as
7:31
a health system and frankly, where our skills,
7:34
require partnerships.
7:36
So first and foremost, we've had a long,
7:38
long history of integrating behavioral health,
7:42
in our primary care settings for well over
7:44
twenty years now. And we're really, really we're
7:46
very much an early adopter in that work.
7:49
In doing so,
7:50
what we learned is it's really important when
7:52
a behavioral health need is identified or screened
7:54
for in a primary care setting, we need
7:56
to provide feedback
7:58
to the provider
7:59
and the patient
8:01
of what type of services
8:03
or supports they may need.
8:05
So this started quite rudimentally,
8:08
when we talk about,
8:10
using
8:11
a whole host of different screening tools as
8:13
an example
8:15
to help identify
8:16
the complexity
8:18
and the intensity of the need.
8:20
And from there, it gives,
8:22
it gives our primary care providers
8:25
an opportunity
8:26
to very quickly determine what supports they need
8:29
either within their primary care clinic or referral
8:32
out.
8:33
In addition to that, it also helps, the
8:35
primary care provider know when there's an emergent
8:38
need
8:39
that requires
8:40
crisis support services. So for example, if there's
8:42
an identification of someone that's actively suicidal, for
8:45
example,
8:46
it not only identifies sort of those needs
8:48
and and it but it also helps us
8:50
place and and helps provide the support to
8:53
the primary care provider that has very limited
8:55
time to get to to help that patient
8:58
get into the right place at the right
9:00
time. So this began over twenty years ago,
9:02
and we've done many, many learnings from that
9:04
and and have learned from many experiences.
9:08
As an integrated health system, we also not
9:10
only have delivery data, but we also have
9:12
claims data
9:13
and demographic data of our patients. And so
9:16
what we've learned over time is how can
9:18
we use all of that data to better
9:21
predict
9:22
where patients will have greater success. And we
9:25
had some early developments in that work,
9:29
as sort of the developments of sort of
9:31
AI has become quite popular
9:33
today, but the fact of the matter is
9:35
we had been sort of using different
9:37
algorithms to sort of help us understand and
9:40
use data that was available to us.
9:43
One of the one of the lessons that
9:45
we've that we've learned, however, is we're not
9:47
a software organization. So we partnered,
9:51
a little over a year ago with NeuroFlow
9:54
to really help us understand,
9:57
they they're a software,
9:59
organization that really focuses on outcome measurement and
10:02
a platform
10:03
so that we can continue to evolve what
10:06
we've learned over the years and continue to
10:09
iterate that so we can better and more
10:11
precisely determine that. So, again,
10:14
patient shows up in primary care.
10:16
Primary care offers and pushes out a set
10:19
of either questions to include their own data
10:22
and help us then identify where does that
10:24
patient need to go.
10:25
Matching is incredibly important.
10:27
We oftentimes
10:29
talk about
10:30
the importance of providing access.
10:33
However, what we oftentimes miss
10:36
is access to what? Access for access sake
10:38
or in fact, are we looking for access
10:41
to the most effective care possible?
10:43
And so connecting them to that care is
10:45
really, really critical for us. So it's really
10:48
finding the right match,
10:49
enabling our primary care providers in this particular
10:52
use case to help connect the dots,
10:55
to allow them to sort of to to
10:57
land in the right service provision, whether that's
10:59
in their own clinic
11:01
or referred out. And to also close the
11:03
loop to know that that patient's actually getting
11:06
the care they need, and we're able to
11:08
see measured improvement with outcome tools,
11:11
to help us do that work moving forward.
11:14
Yeah. I imagine those those clinical decision making
11:17
tools. I mean, there's really no way to
11:19
capture that value when,
11:21
where it really just it's it's invaluable
11:24
when it comes to, like, how
11:25
much they're able to support primary care physicians,
11:28
as you said, who are overburdened and helping
11:30
them be able to not just identify
11:33
that need, but then also get folks connected
11:36
to the right
11:37
place and and resources for additional care.
11:41
When it comes to
11:42
burnout, which I think kind of alluded to
11:45
in your in your responses already, but obviously
11:48
remains a top concern, especially in behavioral health,
11:51
where demand continues to to outpace
11:55
available availability in per providers.
11:58
How is AI helping to alleviate some of
12:01
the administrative burden on clinicians?
12:04
And are are there any measurable improvements in
12:06
efficiency patient outcomes that you've seen to date?
12:11
Let me Erica, let me continue with thread
12:14
related to,
12:16
the work that we've done that we are
12:18
embarking on with neuroflow and we've done historically.
12:21
So first, let me highlight
12:23
one of our two priorities, strategic priorities as
12:25
an organization
12:27
is simplification.
12:28
Simplification for the patient, simplification for the provider.
12:33
We recognize
12:34
that our providers today,
12:37
when they when they made
12:39
the incredibly important decision to go into this
12:42
field,
12:43
they had not envisioned the administrative burden that
12:46
exists in our systems that have sort of
12:48
built up over the years. And so it's
12:50
really critical for us to think about how
12:52
do we simplify the work? How do we
12:54
make it so it's actually easy for them
12:56
to do the work as opposed to adding
12:58
administrative burden? So for example,
13:01
and I wouldn't consider this necessarily
13:03
AI, but the tool itself has AI components
13:06
in them as it's trying to sort of
13:08
help predict what's gonna best connect them in
13:10
the care.
13:12
The the the first example I would highlight
13:14
is is the previous example. But what that
13:17
does for us is it actually what and
13:19
what we learned
13:21
that we can't have an AI
13:23
platform, for example,
13:25
integrate as a separate entity. It has to
13:27
be integrated in the workflow, therefore, it has
13:29
to be integrated in our electronic health record.
13:32
It has to be done in such a
13:33
way
13:34
so our providers are not even spending an
13:36
extra click to use a particular tool. It's
13:39
just part of their natural workflow. So that's
13:41
really important for us as we think about
13:43
this work. Actually, we're in the midst right
13:45
now
13:46
of moving our entire
13:48
program. We have some of our regions are
13:50
on EPIC, but we will be
13:52
by September 6, we'll be completely on EPIC,
13:55
and we're thinking a, like, a great deal
13:57
about that. So that's sort of one example.
13:59
The second one that I wanna highlight that
14:01
we've actually done outside of behavioral health, but
14:04
will be applied here is our use of,
14:07
you know, initially, we had embarked back in
14:09
2022
14:10
by using
14:11
DACS,
14:13
as a as a tool to help
14:15
address,
14:16
all the documentation requirements for our providers.
14:19
And and the first version of it was
14:21
really
14:23
an opportunity for someone to record
14:25
content and to have someone transcribe that. Well,
14:28
fast forward a year and a half later
14:30
to two years, and we're now in 2024
14:33
a year ago.
14:34
And now we have ambient listening
14:37
through AI. And so we've had almost a
14:40
thousand providers in our system
14:43
using
14:45
DAX in that framework, in that iteration that
14:47
has this AI capability where it's listening to
14:49
the conversation between the patient and the provider
14:53
and actually providing notes
14:55
for the provider to document that work.
14:59
The feedback that we've experienced,
15:02
has been significant.
15:03
Specifically,
15:05
providers, as we ask direct feedback for these
15:08
users, have have actually had positive feedback. Oftentimes,
15:11
when we provide new tools to providers,
15:14
adoption rates are relatively low.
15:16
The providers we had over 6063%
15:19
of our providers, which is a lot if
15:21
you think about early adopters,
15:23
actually using the tool more often, which was
15:25
really, really impressive to us.
15:28
Not only does that offer providers an opportunity,
15:32
to improve their administrative burden, but what they
15:35
also reported is the time they had in
15:37
engaging with face to face contact with the
15:40
patient
15:41
increased significantly.
15:43
And what we know is
15:45
when providers are working at the top of
15:47
their license and they are practicing
15:49
their craft,
15:51
it actually reduces burnout.
15:53
And so understanding that is really, really important
15:56
to us.
15:57
So those are just a couple examples,
16:00
of how
16:01
AI has sort of been been been woven
16:03
into the work that we're doing. We're excited
16:05
to see this roll out,
16:07
to include behavioral health, but we but we
16:09
know that's an important piece of doing that
16:11
work moving forward.
16:14
Yeah. Absolutely. I think it's been really interesting
16:16
to see the evolution
16:18
of ambient AI and how valuable,
16:21
physicians have seen it. And now, you know,
16:24
in into the year ahead, seeing so many
16:26
systems expand into different specialties. So certainly more
16:30
more to come there.
16:31
Tamara, to close us out, future looking here,
16:34
when you're looking ahead,
16:36
what do you see as the next frontier
16:38
for AI and behavioral health? Are there any
16:40
emerging technologies or policy
16:43
policy shifts even that
16:45
might accelerate adoption or effectiveness in supporting clinicians
16:48
and patients?
16:50
Absolutely. Absolutely. So as we're looking ahead,
16:54
I think for the next frontier, I see
16:55
a couple opportunities. The first is, how do
16:58
we support our providers,
17:00
in between sessions?
17:01
We think that there's some opportunity there. We
17:03
know that there's limited access
17:06
to to, you know, particularly
17:08
for episodes of care
17:10
given the fact that our workforce is not,
17:13
cannot literally provide
17:15
the frequency of care
17:17
that would be ideal for patients. And so
17:19
there are solutions that are available in the
17:21
the form of bots and educational content that'll
17:24
help support that work moving forward.
17:28
We think that there's some great opportunity there.
17:31
The second thing I'll add is is we
17:34
see opportunity when we think about the development
17:36
or the
17:38
providers using evidence based psychotherapies
17:40
with the right with the right intervention at
17:43
at great fidelity, meaning that they're practicing
17:47
those techniques, whether it's cognitive behavioral therapy or
17:49
whether it's,
17:51
whether it's,
17:53
dialectical behavioral therapy.
17:55
Feedback is very, very hard,
17:57
in the current the way that care is
17:59
delivered. There are now solutions
18:02
to actually use ambient listening to provide feedback
18:05
to the provider
18:07
to ensure that they're actually
18:09
practicing
18:10
within the scope of a particular
18:13
evidence based practice. And we think, although that
18:16
may be scary to think about at the
18:17
beginning for therapists,
18:19
it's an opportunity for them to continue to
18:21
improve.
18:22
And by that improvement, they'll actually deliver better
18:25
care.
18:26
What we know in general
18:28
that when folks just randomly get connected to
18:31
care and we're not quite certain whether that
18:34
practice is is evidence based or that there's
18:36
a mismatch between the need and evidence based
18:38
practice.
18:39
What we know is the efficacy of care
18:41
can be as low as thirty percent.
18:43
If we have the right match to the
18:45
right evidence, we see the efficacy
18:47
similar to
18:49
salicylates and aspirin is to headaches.
18:52
You know, you know, two thirds of patients
18:54
get better. So we see some opportunities there.
18:57
And lastly, from a policy perspective, I think
18:59
we're seeing a lot of state jurisdictions really
19:02
propping up
19:04
and partnering with health systems like ours to
19:06
think about
19:08
what's the best way we need to be
19:10
thinking about
19:11
supporting these innovations while maintaining safety. And a
19:14
big component of that is having consumer transparency.
19:18
Let's make sure
19:20
and really from a policy perspective, let's make
19:22
sure that consumers know when they're interacting
19:25
with a component of the organization,
19:27
they know whether it's AI or not, and
19:29
that we think that's really important.
19:31
Secondly,
19:32
let's make sure that there are mechanisms to
19:34
have escalations
19:35
beyond those interactions
19:38
that get to a live individual that can
19:40
help support them. So we think those two
19:42
areas really will help enable
19:45
safety while be while we continue to innovate
19:48
in the AI space.
19:50
Yeah. Certainly, Tamara. I think what you mentioned
19:53
at your first point there of just speaks
19:55
to kind of what we're hearing
19:57
in terms of that emerging focus on those
19:59
in between touch points
20:00
and how can we support patients, you know,
20:02
in between visits better.
20:05
So, Tamara, thank you so much for your
20:07
time today. Really appreciate
20:09
all of your insights and and great discussion
20:11
here.
20:13
Absolutely. Happy to have had the conversation
20:15
and and join you. And we wanna thank
20:18
everyone for listening as well. Thanks all.