Subscribe
Share
Share
Embed
Where we share our weekly news debriefs and discussions with industry experts. These are lo-fi recordings aimed at giving our readers more opportunities to engage with our analysis and a view into some of the conversations that shape it.
Martin: So the LA Times recently
published an op-ed from Hashem
on regulating clinical AI.
Oura just announced a recent partnership
with Counsel Health, utilizing
their clinical AI for their members.
We're really excited to have him chat
with us about those things and more.
Good morning.
Uh, I believe you're out in LA,
so I think it's still morning.
Hashem: I am out in LA, and it is morning-
Yeah ⦠but I've lost my voice a little
bit because I'm a lifelong Knicks fan.
So that is- Yeah.
Martin: There's a lot of that going
around the HTN office, believe it or not.
Oh, good.
Right.
Um, so we, we absolutely, uh, empathize.
I wanted to start a little
bit with your LA Times piece.
And give us, like, what is your
sort of current view on how we're,
as a country and, you know, 50
states and the District of Columbia,
approaching regulating clinical AI?
And then as a sort of follow-up to
that, what, what should we be doing?
What's your prescription?
Hashem: Yeah.
Um, it's great to see you guys,
and thanks for having me on.
Uh, what is currently happening is
difficult to sum up in the time that
we have- ⦠because it's currently a
pretty patchwork regulatory framework,
where it is, I think over the last
year, every single state has offered
some legislation around AI, and
some states have offered multiple.
And what that legislation looks like
ranges vastly from state to state.
You have bills that have been introduced
in places like New York and Colorado
that would really limit the ability
of AI to be face, patient-facing in
a lot of ways, and you have other
states, uh, like the sandbox experiment
that's going on in Utah, where AI is
fully practicing medicine right now.
Uh, and I think that the federal
government has stepped back from
regulating in a lot of ways currently to
try to see what the states are doing and
what the states have the potential to do,
and they've created a pretty innovative
environment where you have, uh, you know,
a laboratories of democracy approach is
the phrase that I use in that LA Times
piece, where they're really observing
what is happening in each of these states.
Um, I think that we, we believe
the federal government sh- can
set a floor for what the practice
of AI clinical care looks like.
You know, this is the minimum
that every state needs to meet
in order to participate in this.
And then on a state-by-state basis,
that experimentation can and should
continue to happen, because that's
definitional to kind of US law is
the state practice of medicine.
And so in a really big picture sense,
I think it starts with a floor set
by the federal government, and then
states innovating- Either well on
top of that or meeting that floor
of here's what needs to happen
Kevin: Hashem, I'd be curious.
I, uh, that, that, that thought
resonates with me very much.
I, I look at the state of Utah and the
pilot they have ongoing, and I look at
that pilot and I'm like, "I don't even
know if that seems like the floor to
me of where it should be," because like
I, I, I was looking through the, the
state provided an update on how that
pilot's going and how many people are
going through it, and all of it still
seems to be reviewed by a human today.
Like, it's not yet AI doing the,
the refills of prescriptions.
They are seemingly thoughtfully testing
this as they're rolling it out, which
strikes me as a very thoughtful way to
do this and address some of the issues
that are, are coming up in the state,
uh, as I'm sure we've all seen here.
But even like, that seems like one step
in the right direction, but not yet
at the floor for where I'd imagine the
federal government moves over time.
I'd be curious, like, how, how do
you think about how we should be
thinking about what the floor is
and what the conversation is around
how we collectively define that?
Hashem: Great question.
So I agree.
I don't think that in Utah right now we're
at the floor, and I think that in two to
three years we're gonna look back on this
as not being close to the eventual outcome
of where AI can already be helping doctors
and be helping us expand access to care.
Um, I think that there's a common
perception that the federal government
getting involved in any capacity, whether
it's through the FDA, whether it's
through CMS, whether it's through other
entities within the federal government,
would slow down some of this innovation.
And I think actually putting a
stamp to say, "We are going to
allow this when you meet XYZ,"
would allow this to speed up faster.
And I think once the federal government
is able to help set that floor so that
people who participate in Utah, U- U- you
know, for example, this isn't illegal.
I'm not going to go against the FDA
coming out in three months saying,
"No, we're not allowing this."
Once you set that, then I think you can
start to have significantly faster and
iterative intervention for something
like you pointed out, wh- maybe it's
prescription renewals and taking
doctors out of the loop for that.
Or maybe it's something more aggressive
like autonomously managing urinary
tract infections, which essentially
happens in other countries.
Um, once that is set, once you're able to
have some kind of framework for, okay, I
now know the field that I'm playing on,
I know the rules of this game, then I
think that you can look to, uh, use the
Utah framework that they've set out, which
I think is a good one, to study this.
I mean, the idea of progressing from
a doctor in the loop to one click
care, to autonomous care, using data
to do so, having that data be publicly
available, and producing good science
and research off of this, I think that
Utah path is actually a very good path
Martin: It feels to me like you haveâ¦
You're in such a unique position.
You're a doctor.
Hashem: I am.
Martin: You're also the medical
director, uh, for clinical research
and policy at Counsel Health.
We're seeing some emerging fault
lines, I think, between providers
and the bots, if I'm, if I'm being
a little cute, on, on what should be
allowed, what shouldn't be allowed.
I'm curious if you're feeling any sort of
ambivalence or, or dissonance about this.
And more broadly, like w- what should
we think about as like a fundamentally
doctor job and a fundamentally robot
job, and what's your view on that?
Hashem: Yeah, I mean, that is
the, the ultimate question, right,
that, that a lot of people are
Martin: asking right now.
Just small, you know, just
small morning questions for you
on this, uh, Monday morning.
Hashem: Yeah.
I mean, one thing thatâ¦
I obviously get asked that question
a lot, and one thing that I think
is important to frame that within
is what currently happens, right?
I think a lot of times there's a
comparison against an ideal version, and
then there's a comparison to reality.
So by that same framing,
like I work clinically.
I work clinically at UCLA.
Is it a doctor job to get
a patient a cup of water?
I would vehemently say yes to that.
But is it a doctor job to sit in front
of an EMR constantly going through
medication renewals for people who
weren't able to access care and otherwise
and end up in the emergency department?
Um, I think at Counsel we are
very, you know, physician-driven
and physician-aligned.
Our CEO is a doctor.
We have four or five practicing
clinicians that, that work on the
platform, or all of our doctors that
work on the platform are obviously
practicing clinicians, but they're
in leadership positions at Council.
And we are very much making an
effort to align with the physician
community to make sure that anything
that we're doing is in the service
of making doctors' jobs better.
And I think that a lot of this AI
that's happening can really allow
doctors to, you know, to be cute,
practice at the top of their license
to, when I'm working clinically in the
emergency department, not be spending
my time renewing medications, to really
fundamentally try to combat the supply
and demand mismatch that we're seeing,
that I know a lot of primary care
doctors face as well, really redefining
what that relationship can look like.
And so I personally, and I, we as a
company, don't see much of a split
between, you know, what a job of a bot is
and what a job of a doctor is, because as
a company, we see those as so entwined.
Kevin: One of the things that we've been
paying attention to recently is, um,
uh, the ACCESS model, as part of the
ACCESS model, consumer wearables, and
everything that's going on in that space.
Obviously, Oura is one of the leaders
in that market, and they just announced
a partnership with y'all as part of
their latest product release where,
um, as I understand it, they're
y- they're gonna have access to
council providers in the Oura app.
Could you explain a little bit about,
um, that partnership, what's going on
behind the scenes, how you guys are,
are working with Oura, and how the
experience has come together across both?
Hashem: Yeah, I mean, a lot of work
has gone into the experience and, you
know, to integrate this and to have this
within Oura's amazing patient-facing
app already has taken a ton of work
from a lot of stakeholders on our team.
Um, I think that this is kind of a
revolutionary experiment in healthcare.
I think for a long time we've
been talking about the ability of
wearable data to influence clinical
decision-making, and if it could do so.
And this, as far as I know, is the
first time that we're testing that
hypothesis to see what we can do with it.
Um, I think right now, uh, this
week for the first time, we will be
rolling out to s- w- speak with Oura
patients and we are going to be able
to take the biometric data that they
provide us, ranging from heart rate
variability to looking at, you know,
uh, women's health, looking at menstrual
cycles, looking at, uh, pregnancy or
fertility tracking, and integrate that
into our clinical decision-making.
And I think this is something
that patients have really felt
the need for for a long time.
You know, you hear stories about patients
going to their doctor saying, "I have th-
I have X finding from my sleep last night.
What can we do about that clinically?"
And doctors really not being
able to take action on that.
And on the flip end, you have this,
uh, you know, now what problem of
patients plugging that data into Claude,
plugging that data into Frontier AM- AI
models, and not being able to have any
clinical action that, that follows that.
And I think it's gonna be pretty
exciting to study rigorously
what this can look like.
Martin: I'm curious, so I, I, I've
heard from some folks, we interviewed
Zeke Emanuel last week, and he wrote
a book called Eat Your Ice Cream.
He's sort of famously kinda anti-wearable.
The concern we hear from some folks
is that- Uh, you know, ChatGPT,
Claude, wearables, these are actually
gonna increase demand for healthcare
services, that we're not gonna sort
of achieve any savings from this.
Instead, we're just gonna have a lot
more utilization, perhaps more people
showing up at the emergency room.
I'm curious what your
perspective on that is.
Hashem: Yeah, I mean, I've spent
a lot of time thinking about this
from two different, uh, angles.
Number one, I'm aâ¦
I'm personally a big runner, and I run
without any wearables, without a watch.
I go, I run, you know, an hour
a day without any information.
So I relate to somebody who does not
want to see that information, but I
think that it's becoming a rare breed.
On the other end, I'm a researcher, right?
I'm a health services researcher,
and what I study is unnecessary
emergency department utilization.
So I'm very interested in making
sure that that's not the outcome of
what this looks like, and I thinkâ¦
I think about it in two ways.
Uh, one, new technology allows
us to completely rebrand what the
supply and demand curve looks like
in a way that nothing else has.
So the idea that we now
have patients' data, ifâ¦
In an old world, I agree that
the ability not having any access
with that data does present, you
know, what are you going to do?
You're gonna go to an
emergency department, you're
going to go to urgent care.
You're gonna, you know, hammer your
primary care doctor over MyChart.
But in a world like, you know, where
you have counsel, what you're able
to do is communicate with a doctor
at any point asynchronously in order
to take action on that information.
And so once you're addressing this at
that point of access and you're not just
creating more information, but you're
also creating more points of access, then
I think that allows us to really study
what this can look like and to potentially
impact it because you're, you're changing
what those access points look like.
The second part that I think about is
that in an old world where you did not
have, you know, any kind of AI informing
what the interpretation of that data
looked like, any kind of AI that allowed
a doctor to spend more time face-to-face
with the patient interpreting the
data that they're bringing in, that
data is potentially non-actionable.
But I think we have the ability
to create an entirely new realm
of science . Like, what doesâ¦
I don't think you can just say off
the bat, "Oh, heart rate variability,
um, looking at doesn't mean anything."
We're, we're just gonna say
no to taking in this new data
be- before we even have it.
I think you need to take that data
in, and then using AI, try to create
an entirely new world of science to
say what does interpreting this look
like with an entirely new framework
Kevin: I'd be curious, uh, we've heard
from CMS leadership the, the goal of
implementing AI in a deflationary way.
That's part of the thesis behind ACCESS.
Uh, Martin articulated some of the,
the, the concerns about AI being
used in inflationary ways and, and
driving up costs in the system.
If you were bending the ear of CMI
leadership, CMS leadership on how they
can ensure that AI is, is used differently
in healthcare, it has a deflationary
effect, what would you be thinking about?
How would you be prioritizing the
conversation in, in current environment?
Hashem: I think ACCESS
is set up very well.
I think that they have done a really
strong job setting up ACCESS, and I
think the pushback around prices is
a signal of that . I mean, if you are
getting pushback around prices, then
that probably means that you are doing
a very good job promoting companies that
are trying to do this differently, which
I think is the, the goal of, of ACCESS.
Um, I think that the way that
we think about this internally
is aligning with a lot of the
things that ACCESS already does.
I think taking AI and making sure
that payments that are coming
out of it are outcomes-based
payments, which is obviously part
of the, um, framework of ACCESS.
Another way to think about that
is thinking about it in terms of
time-based payments, so almost
like a subscription model to an AI
service, because the concern is in
a fee-for-service world, you canâ¦
This is endless utilization, right?
And once you take the cost of care
down to the cost of compute, that
could-- You-- Well, any incentives
will align to find somebody, uh,
a way to spend that endlessly.
So another way that we've been thinking
about it internally is more along a
subscription-based model, um, that
aligns with those outcome payments.
So if I were speaking to CMI
leaders, I would applaud them
for what they are currently doing
with ACCESS because I think that
that is, uh, a very strong start.
And then trying to think of ways, uh,
together with companies that are really on
the cutting edge of doing this, like, of
seeing how we can also make it time-based.
Martin: This was super helpful.
I'm really excited to see how all
of this plays out and any more
of your writing that you have.
Really enjoyed the piece in the LA Times.
Uh, if folks are interested in learning
more, where can they follow you for
more writing, more thoughts on, on this
developing space, um, and all of that?
Hashem: Yeah.
Well, our company is Counsel Health,
and that i- we have a website.
And then I am on, uh, LinkedIn, and
I'm trying to, you know, start writing
a little bit more there and, and
posting, and so you can find me there.
Martin: Great.
Well, we will include a link
to that in the show notes.
But thank you so much for your time today.
We really appreciate it.
Hashem: Thank you, guys.
It was great to see you.
Nice chatting with you.
Nice seeing you.
Yep, see you.
Bye.