Latest / The Operator | Healthcare Business / AI Should Give Clinicians Back to Patients

Transcript

AI Should Give Clinicians Back to Patients · The Operator | Healthcare Business

0:01
All right,
0:02
everybody wants to talk about AI,
0:04
or maybe they don't wanna talk about it,
0:05
but they sure are talking about it.
0:07
And the questions come up.
0:08
Will it replace clinicians?
0:10
Will it write our notes?
0:11
Yes.
0:12
Will it make healthcare more efficient?
0:14
Hmm, remains to be seen.
0:16
Larry's latest article asks a completely
0:18
different question.
0:21
What if AI isn't here to replace
0:23
clinicians?
0:24
What if it's here to give them back
0:25
the part of the job they fell in
0:27
love with,
0:28
in the first place.
0:30
So Larry, let's start there.
0:33
Why do you think everybody's asking the
0:35
wrong question?
0:36
Because you open and you say,
0:37
forty years ago,
0:39
I took an oath to care for people.
0:41
You started with an oath forty years ago,
0:44
not with AI.
0:45
Why did you start with that oath from
0:47
long ago?
0:47
Yeah,
0:49
I think it's because the conversation
0:51
about AI has already gone backwards and I
0:55
wanted to reset
0:58
and level set around why we got into
1:01
this profession.
1:04
Everyone is now entering it asking what
1:07
the technology can do.
1:09
I wanted to start with what I promised
1:11
to do as a caregiver, four years ago,
1:16
to a patient as a caregiver,
1:18
not to a vendor.
1:20
And so the oath sort of frames the
1:22
tool, not the other way around.
1:23
If you start with AI,
1:24
you're already conceded the argument to
1:26
the people selling it.
1:29
I'm just not interested in that
1:30
conversation for the purposes of call to
1:32
care.
1:33
I'm interested in what we owe the person
1:35
on the table, the patient.
1:37
So the oath is the standard,
1:39
the code of ethics,
1:40
the whole professional reasons why we got
1:42
in here.
1:42
AI either measures up to it or doesn't.
1:46
Alright,
1:46
so one of the strongest lines in the
1:48
article, I thought anyway,
1:49
was the only question worth asking is
1:52
whether we have the discipline to use this
1:55
tool for the one thing it's actually good
1:58
for good for giving the human being back
2:02
to the patient.
2:03
What do you mean by giving the human
2:05
being back?
2:06
What does that mean to you?
2:08
Yeah, so if you think historically,
2:10
as your time as a PT and healthcare
2:13
provider,
2:14
what we've really been robbed is with time
2:17
with patients,
2:18
particularly in physical therapy where
2:20
everything has been, you know,
2:22
remediated or, you know,
2:24
gone down to a billable unit.
2:27
It's all about documentation,
2:28
pre-authorization, productivity mandates,
2:31
the nine-minute, eight-minute rule,
2:33
not twelve-minute visit windows.
2:35
Every one of those looks like another
2:37
minute of attention redirected at a
2:40
screen.
2:41
So giving the human being back means
2:46
reversing that theft.
2:49
Eyes on the patient,
2:50
listening for the thing that the chart can
2:53
never capture,
2:54
noticing what the patient isn't saying,
2:57
the verbal and the nonverbal.
2:59
That encounter fully present
3:02
undistracted is what got engineered out of
3:06
workflow and maximizing productivity and
3:08
efficiency we need to engineer it back in
3:11
and that's the whole argument everything
3:13
else is footnotes i started writing an
3:17
article and tim flynn was helping me with
3:19
it and i think i actually sent it
3:21
to you to review a bit and it's
3:22
about the five agents everything in ai now
3:25
is about agentic
3:27
movement and moving to agents.
3:28
I love that.
3:30
And I've identified what I believe are
3:32
five agents that can help us in physical
3:35
therapy.
3:36
because at the end of the day they
3:37
bring humanness back to health care and
3:41
the one we're going to talk about in
3:42
this article or what we did talk about
3:44
in this article and it'll transgress a
3:46
little bit also while i'm reintroducing
3:48
call to care is about the whole idea
3:51
of bringing the human back and using ai
3:54
scribe for the documentation pieces but
3:56
don't get me wrong scribes and agents can
3:59
do many other things they can do pre-cert
4:02
They can do outcomes.
4:03
They can give us feedback loops and a
4:05
daily report of how we conducted.
4:08
They can listen to our notes and give
4:09
us constructive criticism on how we better
4:12
have verbal and nonverbal skills.
4:14
So I'm all for AI,
4:17
but I believe its function is really
4:20
around humanity and allowing us to be
4:24
fully human and present for our patients
4:26
as opposed to be fully engaged in a
4:28
technology.
4:30
So let's bring some humanity into it.
4:32
You brought someone into this particular
4:34
article.
4:34
You named her Sarah.
4:35
You said, you know, painted a picture.
4:38
She's not a bad clinician.
4:39
She's been around for about twelve years.
4:41
She's not burned out because she stopped
4:43
caring.
4:44
She's burned out because the system slowly
4:47
pulled her away
4:49
from the work she loved.
4:51
And I think a lot of people will
4:52
say, well, these kids today,
4:54
it's very easy to fall into it.
4:55
These kids today don't.
4:56
But when you look at it that way,
4:58
because she's burned out because the
4:59
system pulled her away from the thing she
5:01
loved.
5:02
How common do you think a Sarah is?
5:05
um well i've known a hundred of them
5:07
i've employed probably hundreds of them in
5:09
the last forty years when your burnout
5:11
rate runs from forty five to seventy one
5:14
percent just on pts and i cite in
5:16
the article dentists and dental hygienists
5:19
and others um and when you have a
5:21
one year arguably fifteen thousand pts
5:24
walking out the door she's not the outlier
5:27
You know, she's she's the norm.
5:29
And and we have to recall the data,
5:32
which shows that on a full time equivalent
5:34
basis,
5:35
only twenty five percent of all physical
5:36
therapists that are licensed are actually
5:38
practicing clinically.
5:40
But what makes Sarah's story so worth
5:42
telling is that she's a great clinician.
5:45
Her outcomes are strong,
5:46
her patients like her,
5:48
her colleagues respect her,
5:49
and she's counting the years until she can
5:53
afford to leave.
5:54
And that's the tell.
5:56
It's not the weak clinician burning out,
5:58
it's the good ones,
6:00
the ones who came in with passion and
6:01
skill that the system,
6:03
the ones who worked hard in school to
6:05
even get accepted in PT school.
6:07
Now, they didn't lose their ability,
6:08
they left their sense of calling.
6:10
They lost their meaningful purposeful work
6:14
And once that goes, the clock is running.
6:17
It's just a function until they're looking
6:19
to get into a non-clinical position.
6:22
I feel like one of the distinctions that
6:24
you made is two sides to a coin.
6:27
People go back and forth and they argue
6:29
over it.
6:30
That burnout isn't this personal failure.
6:33
It's a system that stopped replenishing
6:35
its people,
6:36
what I just said a minute ago.
6:37
Why is that distinction so important to
6:39
focus on where that fault actually is or
6:43
where the solution could be?
6:46
Yeah, it's like everything, anything else.
6:47
You know, once you have the diagnosis,
6:49
then you could determine the prescription.
6:50
Right.
6:51
And I would also argue the system includes
6:52
our education system because we know
6:55
statistically that by the time new PTs
6:57
land in our clinic for their first full
7:00
time job,
7:01
a good portion of them are already burned
7:03
out.
7:04
Now, we can argue about the term burnout.
7:06
There's some medical things around it.
7:08
I've written pretty extensively about it.
7:10
But let's forget about the word for a
7:12
minute.
7:12
Just think of it as people leaving the
7:14
profession or being disengaged.
7:16
When you're disengaged and working with
7:17
patients,
7:18
you're less likely to have very good
7:20
outcomes.
7:22
um if it's a personal failure then the
7:24
fix would be resilience you know wellness
7:27
you know a lot of those kinds of
7:28
things but we've been running that
7:30
experiment now for a long time you know
7:33
burnout has actually gotten worse even
7:35
though there's been a more sensitivity to
7:36
it it's actually gotten worse so the inner
7:39
intervention is is failed because it's
7:41
treated it's treated the wrong problem
7:44
when you name it as a system failure
7:46
you're forced to redesign the system
7:48
that's much harder
7:51
And no one in leadership really wants to
7:53
talk about because it lands accountability
7:57
where it belongs, the operators.
7:59
And again, it's all back to operation,
8:01
the people who built the workflows that
8:03
the very clinicians were trying to keep.
8:05
So that's the distinction.
8:07
It's not a semantic description, Jimmy.
8:12
It's really what determines whether
8:15
anything actually can change or happen to
8:17
improve this.
8:19
Let's talk about AI scribes.
8:21
I feel like everybody markets them when
8:24
you see an ad for them.
8:25
If you get on a call with someone
8:27
who's wanted to talk about them,
8:28
they lead with efficiency.
8:31
That's their sell.
8:32
It's an efficiency tool.
8:33
And then they'll hit you with a statistic,
8:34
usually in time or money saved.
8:37
You argue that's almost missing the point.
8:41
Where are they missing the point on this?
8:43
Yeah,
8:43
so let's just level set here a bit.
8:48
It took us a long time to get
8:50
adoption to EMR,
8:52
especially in physical therapy.
8:54
The good news, I think,
8:55
is that there's been reasonable adoption
8:58
and quick adoption by AI scribes in PT
9:01
and in healthcare in general.
9:02
I've put dentistry a little bit behind in
9:04
that regards.
9:06
But why has that been the push?
9:08
The push is because we want to make
9:10
the system more efficient.
9:12
and efficiency is the wrong goal.
9:14
If you deploy a scribe and tell a
9:16
therapist that the point is that you can
9:18
now see more patients per day,
9:21
you've simply designed a tool, you know,
9:25
you've designed a tool to restore humanity
9:27
to work and used it to accelerate the
9:29
very thing and the very reason why Sarah
9:32
is burning out.
9:33
You haven't solved the problem,
9:34
you've automated it while leaving
9:36
everything else intact.
9:37
And that's sort of the, in the article,
9:39
I talk about Dr.
9:40
Blaine's calcification with a software
9:43
upgrade.
9:45
And the other thing is,
9:46
if the win is, oh,
9:47
we're doing this for our therapists,
9:48
this is how much we care,
9:49
and this is why we want to attract
9:51
more PTs because we have such shortages,
9:54
because you'll be able to,
9:55
by using a scribe,
9:57
you'll have less time on notes
10:00
you know, in the office,
10:01
so more time at home.
10:02
And, you know, that's real,
10:04
and that's a quality of life improvement,
10:05
but that's just optimizing Sarah's
10:07
evenings.
10:08
I'm not interested in what happens to the
10:10
patient, you know,
10:12
what happened to the patient in the room.
10:14
The scribe's purpose is simply to make
10:17
Sarah faster by that, you know,
10:19
unprofessional management business speak.
10:22
That's not the purpose.
10:23
The purpose is to make her fully present
10:25
for the patient.
10:27
Those are very different goals with very
10:29
different outcomes.
10:31
If it's efficiency,
10:32
it's going to lack effectiveness.
10:35
If it's effectiveness,
10:36
it's going to be got effectiveness.
10:39
Management keeps confusing them and the
10:41
patients pay for it.
10:43
So if you look at most large scale
10:46
operating platforms,
10:48
in healthcare.
10:49
They're getting AI scribes to claim they
10:52
have a better workplace for the therapist
10:54
so they have less notes at home,
10:56
claim the therapist can see more patients,
10:58
but none of them are talking about the
11:00
very thing that it's meant to do,
11:02
bring humanity back to healthcare and
11:04
allow them to be fully present and be
11:06
more effective to produce better outcomes.
11:08
That's the real difference.
11:10
You know,
11:11
the first time I took a look at
11:12
an AI scribe a year and a half,
11:14
two years ago, I, of course,
11:16
default to like an advertising campaign.
11:18
And I said,
11:19
this might bring eye contact back.
11:21
And as simple as that is,
11:23
that felt more humane than the twenty
11:25
eight minutes per note saved.
11:28
That felt better.
11:30
You talk about a cockpit analogy instead
11:34
of automation,
11:36
where
11:39
The autopilot does the mundane.
11:41
It maps, it flies,
11:43
it goes up and down,
11:44
but the human comes in at different
11:46
points.
11:47
Why is automation the wrong mental model
11:50
for healthcare?
11:51
Because behind AI,
11:52
automation has got to be the next most
11:54
used word recently.
11:55
yeah it is it's the wrong word again
11:57
because it implies automation implies
11:59
replacement you know on a factory line a
12:02
machine removes the human in a clinic that
12:05
collapses on contact with reality right a
12:09
therapist
12:10
can't hand clinical judgment to a
12:13
probabilistic system.
12:14
It can't automate that out of it.
12:16
No patient ever got better because a model
12:19
was confident, right?
12:22
So the cockpit is the right mental model
12:25
because the roles stay very, very clear.
12:28
The autopilot flies the plane, altitude,
12:32
course, systems.
12:34
The pilot commands it.
12:36
It takes over at takeoff landing.
12:39
The pilot takes over at landing at
12:42
takeoff.
12:43
In the instance,
12:44
something unexpected happens.
12:46
Nobody says the autopilot replaced the
12:49
pilot.
12:51
His or her interventions are fewer,
12:52
better leveraged, and at higher stakes.
12:54
It allows them to be more present for
12:56
when the other problems that can happen.
12:58
And so that's the architecture for
13:00
healthcare AI that I really, really like.
13:03
It allows the therapist to be on
13:05
autopilot.
13:06
The system manages the workflow.
13:08
The clinician manages the decision.
13:10
It manages the human behavior.
13:12
interactions.
13:13
This idea of very much every time we
13:15
have high technology,
13:16
we have to have high touch.
13:18
And so automation allows the therapist to
13:21
be on autopilot to be more hands-on and
13:26
more high touch.
13:28
The minute you call it automation,
13:29
you've already really started to design
13:32
the wrong outcome, in my opinion.
13:35
One phrase that stuck out was,
13:37
and this feels like it could be on
13:38
a bumper sticker or a T-shirt or an
13:39
ad campaign, was AI drafts,
13:42
clinicians decide.
13:44
Why is that line in the sand so
13:46
important for healthcare leaders designing
13:48
or implementing these AI systems?
13:52
Because generating a note is the easy
13:55
ninety percent.
13:56
The question that earns trust is how you
13:59
confirm that it's true.
14:01
You know,
14:01
a clinician who reviews what the model
14:03
produced and signs it isn't a rubber
14:06
stamp.
14:07
She's an accountable professional
14:09
exercising judgment.
14:10
A clinician who signs without reviewing is
14:12
ratifying, not verifying.
14:14
Right.
14:15
And the liability sits with the clinic
14:18
clinician either way from a legal
14:19
standpoint.
14:21
So AI drafts,
14:22
clinicians decide isn't a slogan,
14:24
it's a design principle.
14:26
It forces accountability into the workflow
14:29
rather than bolting it on after the fact.
14:32
You know,
14:33
any system that doesn't build verification
14:35
as the product, not generation, you know,
14:38
as a malpractice machine, you know,
14:40
dressed up with efficiency software,
14:42
right?
14:43
I won't work with one and I won't
14:45
recommend working with one.
14:46
I mean, that's just the key.
14:47
You have to really let clinicians be
14:49
clinicians and not, you know, defer to AI.
14:54
There's a great saying,
14:55
and it's absolutely true,
14:56
and I've said it,
14:57
and that is that AI doesn't have a
14:59
bad day.
15:00
Well, that's exactly right.
15:02
But clinicians do.
15:03
And when clinicians have a bad day,
15:05
their notes aren't going to all of a
15:06
sudden reflect a good day.
15:09
And so you need verification and you need
15:11
the therapist to actively engage them.
15:14
All right,
15:15
if a healthcare CEO and operator called
15:17
you tomorrow and said, hey,
15:18
we're investing heavily in AI,
15:21
what's the one question you'd ask to have
15:23
them reveal whether or not they're
15:25
building something that restores humanity
15:27
or just another productivity tool?
15:32
Yeah, I would say,
15:33
are you measuring your AI scribe and your
15:35
investment in AI by the patient face time
15:39
with the therapist or by how efficient or
15:42
throughput or how many visits per day?
15:45
You know,
15:45
you're not seeing board decks with patient
15:48
face time and outcomes.
15:49
You're seeing board decks with visits per
15:51
day efficiency.
15:52
And oh, yeah,
15:54
we're a better workplace because our
15:55
therapists don't have to do work at home.
15:57
And so therefore we're trying to solve,
16:00
you know, retention and turnover.
16:02
How about changing the narrative a little
16:04
bit and using AI to help therapists
16:06
produce better outcomes and more effective
16:08
time with their patients as opposed to the
16:11
efficiency trap?
16:13
If an operator gives me a productivity
16:16
number on how much AI helped their
16:18
productivity, the patients per day,
16:19
the visits per hour,
16:21
or notes completed faster,
16:23
they're using AI to accelerate the very,
16:26
very problem that was burning people out.
16:30
In fact,
16:31
I wrote even a couple of years ago
16:35
in the Substack about how my initial fear
16:38
around AI was that it was going to
16:40
be used as the tool to see more
16:44
patients per day or be more productive.
16:47
I didn't realize how prescient that was
16:48
because that's exactly what's happened.
16:50
You're not hearing about how more
16:52
effective a therapist is because people
16:54
aren't getting the heart of the burnout
16:56
program.
16:56
They're trying to control the symptoms
16:58
through efficiency.
17:00
The money question here,
17:02
and we've heard this before on pitches or
17:04
in ad campaigns,
17:05
if AI gives every clinician an extra
17:08
thirty minutes a day,
17:10
what's the worst thing a leader could do
17:13
with those thirty minutes that has been
17:15
given back?
17:16
Yeah,
17:16
the worst thing they can do is tell
17:18
the therapist that they have to see more
17:20
patients,
17:21
that they have to do more compliance
17:23
review,
17:23
that they have to do more documentation,
17:25
that they have to get online and pass
17:27
their HIPAA course for the year.
17:30
Allow the therapist some time to reflect,
17:34
to think about their patient care,
17:36
to mentor, to free them up,
17:38
to do clinical work.
17:40
The thing that drives people to become a
17:42
PT was you want to do clinical work.
17:45
Nobody ever became a physical therapist
17:46
because they wanted to be an expert in
17:48
documentation to ever do be a compliance.
17:50
They didn't know the minute rule existed.
17:53
It's all of those kind of things.
17:54
I almost think we ought to make therapists
17:56
sign prenuptial agreements when they get
17:58
into the profession their first day.
18:00
Right.
18:02
You know, and, you know,
18:03
really expose them to the negatives around
18:06
all of these kind of things.
18:07
These distractors are going to put them in
18:10
a.
18:11
in a period where they're going to really
18:12
be there.
18:14
It's almost like that could be used as
18:16
part of a burnout prevention program.
18:17
All right.
18:19
If the conversation about AI usually
18:21
starts with one question,
18:22
what can technology do?
18:24
Allow Larry to help you think it should
18:27
start somewhere else.
18:28
What should the technology give back?
18:30
Because if AI only makes clinicians
18:31
faster,
18:32
I feel like you think we've missed the
18:34
point.
18:34
If it gives them back their attention,
18:36
presence, humanity,
18:37
then maybe technology becomes the thing
18:39
that helps us care more.
18:41
Again,
18:41
if today's conversation challenged the way
18:43
you think and made you a little
18:44
uncomfortable, good.
18:45
That's the point.
18:46
Subscribe to The Operator on Substack.
18:48
And this is part one of a series
18:51
called Call to Care in the Age of
18:52
AI series,
18:53
which will lead into a re-release of some
18:56
literature as well.
18:57
You want to front tease that?
18:59
Yeah, absolutely.
19:00
So Call to Care was a book I
19:01
wrote in two thousand twenty,
19:03
giving humanity health care back to.
19:06
And it's really how we can take positive
19:09
psychology and the social science,
19:10
social neuroscience and understand that
19:14
the soft skills in health care are the
19:16
hard skills.
19:17
And those are the ones that make a
19:19
huge difference on patient outcomes.
19:21
And so you'll see it's expanded with some
19:23
additional chapters on AI, of all things,
19:27
and virtual health,
19:28
which didn't exist even in two thousand
19:30
twenty,
19:30
along with some new psychological
19:32
principles that have stood the test of
19:34
time from an intervention standpoint,
19:36
including implementation intentions,
19:38
some additional things on mindfulness,
19:41
how to be mindful without feeling like you
19:43
always have to meditate.
19:44
And some other areas that I think people
19:46
will chant.
19:47
So at the end of the day,
19:48
the book's got about twenty five percent
19:49
expanded and updated on material.
19:51
So very excited about it.
19:52
All right.
19:53
More to come with that.
19:54
Until next time, keep operating.
19:57
Absolutely.