Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:01
You're listening to There's a Better Way,
0:04
smart talk on healthcare and technology.
0:07
If you're up for energizing and story-driven
0:10
conversations with national healthcare
0:12
leaders driving industry innovation across
0:14
the country, then you are in the right
0:17
place.
0:22
I am so pleased to welcome Anish Chopra
0:25
to our podcast. Anish
0:27
is our ninth and final guest this year,
0:30
and who better to cap off our second
0:32
season than the nation's first
0:35
Chief Technology Officer who served
0:38
under President Barack Obama. There
0:41
was a lot that led up to that, though, and I
0:43
can't help but think of the stepping stones.
0:46
It was 1995. Anish
0:49
was at his first job at Morgan
0:51
Stanley, working in investment
0:53
banking, and he had a front-row
0:56
seat to when Netscape went public.
1:00
It was here that Anish saw firsthand
1:02
how technology, the internet in this case,
1:05
could make huge changes in the world.
1:09
Jump forward a decade to 2006, and Anish
1:12
was serving as the fourth Secretary
1:14
of Technology for the state of Virginia,
1:17
where he led technological innovation
1:19
in the state government. It
1:21
was a kind of training ground for
1:23
what was to come as the nation's
1:25
first CTO. And
1:28
he's still innovating today as
1:30
President of health IT company
1:33
CareJourney. We're honored to
1:35
have Anish join us today, so let's get
1:37
right into this episode.
1:42
Welcome to There's a Better Way, Anish. Thanks
1:45
for having me. Your career is quite
1:47
amazing. Among other things, as I noted,
1:50
you were the nation's first Chief Technology
1:52
Officer under former President Obama,
1:55
and today you're the President of CareJourney.
1:58
We're going to cover a lot of things.
1:59
and healthcare and in technology
2:02
in general. But with luminaries and
2:04
health IT like yourself,
2:06
we like to start where you started.
2:09
So tell us a little bit about yourself. Where
2:11
did you go to school, grow up, and that kind
2:13
of thing? I grew up as the son
2:15
of Indian immigrants in New Jersey. Dad
2:17
went to Villanova and then went home to marry
2:20
mom.
2:21
When I grew up, he had
2:23
shared with me the stories of this
2:25
famous college classmate of his by the name
2:28
of Sam Petroda,
2:29
who was from rural parts of India, but
2:31
came to this country for graduate
2:34
school. But Sam was an entrepreneur
2:37
who was an expert in the field of telecommunications
2:40
and the country was going through a transition
2:43
from analog phones to digital
2:45
kind of infrastructure for phones.
2:47
And so he caught that wave and was
2:50
able to pioneering entrepreneur on
2:52
the next turn of the crank
2:54
for innovation
2:55
in telecommunications.
2:57
I became a very successful entrepreneur, patent
2:59
holder, and he sold businesses
3:02
that have been successful,
3:03
but decided in his like 30s,
3:06
after having exited all this, to go
3:08
back to India for a penny a year
3:10
salary
3:11
and effectively became the chief technology
3:14
officer for India.
3:15
And in that particular time,
3:18
the goal was to get access to telephony
3:21
to all 100,000 plus
3:23
rural villages. And I
3:25
grew up learning about the
3:27
idea that we can use technology and
3:29
innovation to solve public problems
3:32
in a way that was different than an argument
3:34
of,
3:35
hey, we should subsidize
3:37
existing telephone lines. I
3:39
think India had 300,000 phone lines for a nation
3:42
of 300 million people.
3:44
You can say we need a multi-billion dollar subsidy
3:46
to bring all this Western phone technology
3:49
to the rural parts of the country,
3:51
or another political party would say, well, we
3:53
can't afford it, low taxes, let
3:55
the market solve it. And the engineer
3:57
said, hey, there's a better way.
3:59
we can actually invent a domestic
4:02
Indian approach that was able
4:04
to, for $30 million, design a native
4:08
digital first modern
4:10
telecommunications pipe.
4:12
And that allowed the private sector
4:14
to work off of the open standard
4:16
to actually extend that telephony.
4:18
So by the end of the decade,
4:20
without a huge amount of political subsidy
4:23
arguments, they were able to extend phone
4:25
service to every remote village.
4:28
And that story I share
4:30
with you, because it was in the back
4:32
of my mind while I was a policy
4:34
geek and interested in government, I
4:37
always had the lens of an engineer's mind.
4:39
So well, let's dig in on your role, serving
4:42
with President Obama as the first chief
4:45
technology officer for the nation. How
4:47
did that role come about? President
4:50
Obama was campaigning and said
4:52
that part of what he wanted to do
4:54
was tap the technical capabilities
4:56
of the country to solve problems. If you remember,
4:59
President Obama was a bottom up change
5:01
agent, not a top down change agent.
5:03
And he had seen firsthand from the campaign
5:07
how technology allowed for everyday
5:09
Americans to get involved.
5:11
I don't mean this to be a political statement, but
5:13
just his philosophy was to bring
5:15
that mindset to governing.
5:17
Now, at the time, there were very few
5:20
state governments that were similarly
5:22
inspired to put technology in the cabinet
5:25
rank.
5:26
Virginia happened to be one of them. I was
5:28
Virginia's Secretary of Technology. We
5:30
were just awarded best managed state,
5:32
best state for business, best state to raise a
5:34
family. And
5:35
the dot that connected
5:37
was that my governor, Governor Tim Kaine, was
5:40
a finalist to be vice president.
5:42
So as the Obama team was
5:44
doing diligence on Governor
5:47
Kaine, they saw
5:48
his success managing a world
5:50
class government.
5:51
And it was bipartisan, by the way.
5:53
We have a tradition in Virginia to have governing
5:56
be a priority, regardless of party.
5:59
And that was resulted in me serving
6:01
on the transition team, helping to
6:03
advise what it is that a technology
6:05
secretary does in government.
6:08
Wow. With a focus on healthcare, what we
6:10
haven't covered there is how did that focus in
6:12
healthcare happen? So what happened was,
6:14
if you recall, we're in the middle of economic
6:16
recovery, and the president,
6:18
to his credit, said, I'm not going to just throw
6:21
cash in terms of stimulus, which was
6:23
needed at the time. About $100 billion
6:26
of the Recovery Act was dedicated to these
6:29
seed investments for national priorities.
6:32
So we put $10 billion into the electrical grid.
6:34
We put $30 billion, $40 billion into broadband
6:37
expansion. We did work, obviously,
6:39
with the High Tech Act for electronic health records
6:41
deployment.
6:42
So that portfolio happened
6:44
before I was even nominated for
6:46
Senate confer. So there was already
6:49
a healthcare IT agenda that
6:52
had people in charge of execution
6:55
of the plan. We had ONC, and they were doing
6:57
a great job. But specifically,
7:00
the president wanted an advisor in the
7:02
White House that can cut across all
7:04
of the business units or the functioning parts of government
7:07
and the public-private sector relationships
7:09
to make sure we were maximizing
7:12
the value of technology, data,
7:14
and innovation.
7:15
And to segue into the next part of our discussion,
7:18
a big gap in our health IT strategy
7:20
was it was the last vestige
7:23
of on-prem, locally
7:25
constructed software. We
7:27
were entering the period of internet-based,
7:30
cloud-based electronic health record systems,
7:32
the chance to move to modern
7:34
internet standards for data sharing.
7:36
And so that whole what is now the Fire
7:38
API ecosystem
7:41
began as a
7:42
modest $15 million R&D
7:45
grant that ONC administered that I had
7:47
a big role in championing. And
7:49
that, of course, was the beginnings of what now
7:52
is a regulated internet standard for the
7:54
entire industry, which I'm very excited about.
7:56
Absolutely. Take us through that story
7:58
of what happened from 2010 or so to today.
8:03
We wanted to build
8:05
a more population health-based
8:08
care delivery system.
8:11
That was
8:12
obviously what was anticipated with the
8:14
later Affordable Care Act,
8:16
but we had for reasons of timing
8:19
and economic reality,
8:20
we frontloaded the IT investment. And
8:23
as your audience might know, you
8:25
don't put tech
8:26
in front of care redesign
8:28
in a dream world, you want care redesign
8:31
to drive the tech.
8:32
But that's okay, we were given the hand we were given.
8:34
So we needed to land a pre-demand
8:38
signal, meaningful use
8:40
construct that was really geared
8:43
towards value-based.
8:45
We thought we could do that with the subsidies
8:47
to say, look, I know you don't want to record blood
8:49
pressure at a discrete place or smoking
8:52
status. It was seen as
8:53
data entry clerk type work and frustrating,
8:56
but it was critical if you're going to look longitudinally
8:58
at the population to say who should be
9:01
engaged to reduce their risks of heart
9:03
attacks.
9:04
So then the next triple-axle piece was
9:06
that
9:07
if we could gently
9:09
nudge the industry over
9:11
a decade,
9:12
we could ideally
9:14
transition the demand signal from
9:16
government regulation as the requirement
9:19
to the
9:20
concept of value-based care, the
9:22
Affordable Care Act would take hold,
9:24
while CIOs would tell the EHR systems,
9:26
this is the way we want to prioritize our work anyway
9:29
and we'd emerge in
9:31
this beautiful orchestrated 10-year
9:33
roadmap.
9:35
We're laughing because obviously we
9:37
did not land the triple-axle.
9:40
So we had
9:43
a period of rough patch in
9:45
terms of, did people do
9:47
the government compliance with gusto
9:50
and love or was it like minimum
9:52
necessary to get through the regulatory
9:55
loopholes to get right back at the fee-for-service
9:57
features that
9:58
customers really wanted to buy?
10:00
And so we had this frustrating underinvestment
10:04
in stuff that worked
10:05
and overinvestment and stuff that wasn't
10:08
really aligned with the vision of where we're heading
10:10
and increasing pressure
10:12
points to say you got to move to these more detailed
10:14
standards.
10:15
It's been a bit of a challenge for
10:17
the first part of the decade, but
10:20
we come to you with good news. We don't come to you with bad news. The
10:22
Cures Act, bipartisan, and where
10:25
we are today. Now I believe there is
10:27
a more balanced demand signal
10:29
for
10:30
these value-based features.
10:32
And the market is building up that
10:34
natural request.
10:36
And so the implementation
10:39
of the Fire APIs in the Cures Act,
10:42
the roadmap on where we're heading to
10:44
solve more progress in more areas,
10:46
it's going to feel a lot better
10:48
this next decade than maybe the spotty
10:51
transactional history we had in the first.
10:53
Without the first, you wouldn't have the second. There's
10:56
a lot of progress that's been made. We've
10:58
got an electronic health record
11:00
adopted. We have the concept
11:03
of value-based care, and we keep
11:05
saying, what is it going to tilt the scales? And it
11:07
hasn't quite yet.
11:09
If we could dream the dream, maybe
11:11
we would have ripped the Band-Aid sooner and
11:13
said, go to internet-based APIs
11:16
in stage two
11:18
and not wait for stage three and Cures
11:20
Act.
11:21
So there's a dream where we kind
11:23
of pulled forward some of the requirements, and
11:25
would that have put us in a better place? Who knows? But
11:28
that's the academic discussion. It's enjoyable to
11:31
have a podcast to discuss. But remember,
11:33
Sure Scripts was a huge part of the moving
11:35
of the hockey stick up and to the
11:36
right.
11:37
Nothing made us happier than to have the Sure Script
11:40
slides of E-Prescribing going up and
11:42
to the right.
11:43
And it was like affirmation after affirmation
11:45
that we can move to digital
11:47
and it can have an impact. And so to some degree,
11:49
I
11:50
wish we had the same up and to the right
11:52
for
11:52
the rest of the transaction staff.
11:54
That is progress that needs to continue
11:56
to happen. Well, so let's talk about that. I
11:58
mean, there's so much to just talk about. about in the history,
12:00
but let's just move on a little bit.
12:02
So let's go to interoperability, because that's
12:04
where we're headed. You just talked about
12:06
the Cures Act and all,
12:08
and interoperability, the ability
12:11
to exchange health information for the right
12:13
person at the right time, when they need it,
12:15
is still evolving.
12:16
That's the
12:18
best way to put it. I'm more bullish on
12:20
where we are now than I think the perception is.
12:22
So that's why I was keen to have a chance. Oh, so tell me why you're
12:24
bullish.
12:26
Okay, let me do a process
12:28
reason for bullish,
12:29
and then a substantive reason.
12:31
The process reason
12:33
is we flipped it,
12:35
I think correctly,
12:38
to engineer interoperability
12:40
to the patient
12:42
as the first use case for
12:44
the new technology standard.
12:47
And the reason why that's critical
12:49
is that there doesn't need to be a trust
12:51
framework and a governance and
12:54
a national TEFCA and a whole
12:56
litany of things, the minimum data necessary,
12:59
B2B,
13:00
business friction and technical challenge.
13:03
All of that falls
13:05
by the wayside when a consumer invokes
13:08
her right of access. And
13:10
if we could supply that right
13:13
of access with the technical approach that
13:15
allows anyone to put their health information
13:17
on their phone, today,
13:19
every American can put their medical record,
13:22
their common clinical data set now called USCDI,
13:25
on their phone
13:26
for free. If they have an iPhone
13:29
or an Android phone for free,
13:32
without any middleman vendor fee
13:34
structure, it
13:35
works because everyone knows
13:38
how to point data to the consumer.
13:40
And so we spend more time getting the technical
13:43
standards right
13:44
in terms of the data model, and less
13:46
time arguing over that quote trust framework
13:49
on how you can request data
13:51
of me and I can determine whether it's appropriate
13:54
and then respond with a minimum data necessary
13:56
and then look up to the contract and are we allowed
13:59
to share this? We want to share this and
14:01
what are the business terms of sharing
14:03
it. Generation one of
14:05
interoperability, the original four
14:08
NHIN networks, two
14:10
of them were in Virginia, so I had a front row seat.
14:13
When you start with B2B and
14:15
then you want to get technical consensus on
14:17
data model and technical consensus on
14:19
transmission method,
14:21
you're putting a lot on a burden.
14:23
You can barely get out of the room years
14:26
and hours later with just consensus
14:28
on what's minimum data necessary when you're
14:30
asking for data in the context of a
14:32
care transition.
14:34
So that's why we've been stuck in treatment-only
14:36
CCD minimum data necessary
14:39
land
14:40
in B2B Interop for the last decade.
14:43
We're now moving in a slightly different
14:46
lens because the process was all
14:48
data elements to the consumer
14:50
and now we can reapply
14:52
that same technology stack
14:55
to the more thorny business
14:57
model challenges.
14:59
And the key transition technology
15:01
from the consumer facing to the B2B
15:04
world
15:05
is the advent of bulkfire.
15:08
That piece of technology
15:10
that
15:10
has now shipped across 270
15:12
plus
15:12
EHR certified products as
15:17
of 1231, 22 when they all had to meet the deadline.
15:21
We now have the technical
15:23
ability to
15:24
implement data sharing,
15:27
facilitating it point to point
15:30
without let's call it special effort
15:33
as the language of meaningful use called
15:35
for, Cures Act called for.
15:37
That's a gift.
15:39
Now what's missing are the contracts
15:41
and the negotiations to
15:43
unlock it.
15:44
But maybe this is an assignment your colleagues at
15:47
Sure Scripts might say, we're best positioned
15:49
to facilitate
15:51
point to point data sharing through bulkfire
15:54
to bring that future interoperability world
15:56
that seems like it's another half a decade away
15:58
to something
15:59
that's not a big
15:59
that can be demonstrated this calendar
16:02
year.
16:03
Let's go back to the patient and
16:05
to data being deployed to
16:07
the patient or the consumer. So having
16:09
tried some of those things, it does
16:12
take it outside of HIPAA protection. And
16:14
most consumers don't know what that means.
16:17
Yeah, so this is an important part
16:19
of my, call it, volunteer job.
16:22
With Governor Levitt, with David Blumenthal,
16:24
with David Brailler, we co-chaired the
16:26
creation of the Karen Alliance as
16:28
a kind of a multi-stakeholder collaborative.
16:31
One of the first deliverables we shipped
16:34
was a voluntary but enforceable
16:36
code of conduct. And in the
16:38
voluntary but enforceable code of conduct,
16:40
believe it or not, we exceeded
16:42
HIPAA requirements
16:45
in terms of trust.
16:47
As an example,
16:48
today, there's literally no
16:50
disclosure when my hospital
16:53
takes my data, de-identifies
16:56
it and sells it to third
16:58
parties.
17:00
I get no notice.
17:02
I get no
17:03
agency to
17:04
say that I'd rather not do that.
17:07
But
17:07
because it's outside of HIPAA,
17:10
it happens every day and it's a multi-billion
17:12
dollar market where people are buying
17:15
and selling my de-identified
17:17
data. In
17:20
the Karen Alliance code of conduct, we
17:22
required any app that agreed to take on this code
17:27
to publicize what their policy
17:29
is
17:30
with respect to de-identification. And
17:32
to make sure
17:33
that whatever that policy was
17:36
carried
17:37
all the way through its supply chain of third-party vendors. Now,
17:42
are there people that say
17:44
in their disclosure, we have the right to sell this data and you don't have
17:47
a choice? I'm sure.
17:48
Because that's how the healthcare system operates. But
17:51
at least there's disclosure.
17:53
My doctor doesn't tell me that disclosure. My
17:55
hospital doesn't tell me that's disclosure.
17:57
And so I thought that was progress.
17:59
But let's go backwards before we go to problems.
18:02
Let me describe why that was an opportunity for goodness.
18:05
As you remember,
18:07
we made the business decision in
18:09
our policy that
18:11
you can keep proprietary
18:13
data models in healthcare.
18:16
We regulated the conversion
18:18
of some of those data elements into
18:21
what is now the USCDI
18:23
and made them a fire-based
18:25
data model so that data element
18:28
now sits
18:30
in a format that has no
18:32
intellectual property
18:34
constraint.
18:35
Anyone, you or me, can read
18:38
a fire data model. I don't have
18:40
to pay anybody for
18:41
a code set to learn
18:44
what this fire language means in
18:46
the real world.
18:48
That's not how the rest of healthcare operates. I have
18:50
to pay to interpret CPT
18:52
codes and this and that and the other. Everyone's
18:55
got a hand in the cookie jar. But
18:57
in the fire data model,
18:58
it's completely free and open
19:01
source, which means
19:03
everybody is moving their proprietary
19:05
data sets
19:06
who's subject to regulation
19:08
into an open data model.
19:10
This is critical because
19:12
it then doesn't matter whether the destination
19:14
is the consumer who's requesting it
19:17
or I use that to share
19:19
information for a value-based care contract.
19:23
That data model is open.
19:25
So because we built to the consumer
19:28
first, we got everyone
19:30
to build version one of an open
19:32
data model. And the transmission
19:34
method was to a consumer-designated app
19:37
triggered by their username and password.
19:39
With bulkfire, we've now got
19:42
a second transmission method where
19:44
you can aggregate patients
19:46
into a cohort
19:48
or a registry.
19:50
And then you can transfer all
19:52
of the records in that registry
19:54
to a third-party application
19:56
that doesn't need additional patient
19:58
consent or...
19:59
or to involve the consumer
20:02
in any kind of administrative friction
20:04
if they follow existing legal
20:06
frameworks for data sharing.
20:09
OK, all right. Let's go to
20:12
artificial intelligence. Yes. As
20:15
we're talking about interoperability, it's data
20:17
everywhere. And now we've got
20:19
this amazing tool
20:22
that is going like wildfire. It's
20:24
going at exponential pace across the
20:26
country,
20:27
across the world. Lately so. All
20:29
right, right. I'm more giddy about
20:31
this moment than it was when we saw the internet
20:33
blossom. So this feels like
20:36
I'm a kid in a candy store with excitement about ways
20:38
to solve big meaningful problems in a
20:40
relatively short period of time and at low marginal
20:43
cost. So let's go. So
20:45
let's
20:45
talk about it. Where does your mind go
20:47
immediately around artificial
20:50
intelligence? Let's start there. Well,
20:52
yeah. So I've been speaking of a
20:54
dream
20:55
for a digital
20:56
health adviser
20:58
or what I refer to as a health information
21:01
fiduciary service
21:03
whose role is to
21:04
basically run decision support
21:07
in my best interests based on the information
21:09
that it has about me.
21:11
So much of our health care system is that
21:13
the data is inert
21:15
until it's too late.
21:17
And then we show up in the emergency room, and then
21:19
the systems start running, and then everybody starts
21:21
asking questions. And oh my goodness,
21:24
putting you on dialysis, even though if we knew
21:26
you had kidney disease even just two years
21:28
earlier, we could have put you on the new meds
21:30
that can actually eliminate the need to be on dialysis.
21:33
My goodness, avoidable dialysis. Are
21:35
you kidding me?
21:37
How could we have not missed that as an opportunity?
21:39
So part of this is at
21:42
scale, if one has access
21:45
to longitudinal information and is trusted,
21:47
has access and
21:50
is trusted,
21:51
then we can run these models
21:54
to predict
21:55
the next best course of action
21:58
that
21:58
either might involve. involve me seeing
22:00
a new doctor,
22:02
getting a new test,
22:04
adopting a new care plan.
22:07
These are all known
22:09
if we understood
22:12
and could read all of PubMed
22:14
and all of the recommendations. People like
22:16
me have had an interest of this and you should track
22:18
for that.
22:19
I'm South Asian descent, so I've got a genetic
22:21
predisposition to heart disease and I should be monitoring
22:24
and making sure I'm on a statin a little bit
22:26
earlier, even if the guidelines say this or that.
22:30
So how do we create a
22:32
marketplace that will reward
22:36
health information fiduciary investments?
22:39
The challenge I see is
22:41
that the technology is often put
22:43
to its highest and best economic use,
22:46
which may be different from the highest and best patient
22:49
use. Highest
22:51
and best economic use is, hey,
22:53
I'm in a Medicare Advantage plan, they
22:55
need to get
22:56
revenue by looking up my
22:59
health conditions. The highest and
23:01
best use of AI might actually be risk
23:03
adjustment code, which
23:05
is about getting paid more for what I have
23:07
and the conditions that I have. And
23:10
I say that in a slightly negative way. I don't
23:12
mean to be negative. I love everybody. I love MA.
23:15
I love everything. But our economic incentives
23:17
today
23:19
don't reward organizations to
23:21
invest
23:22
in consumer
23:24
decision support.
23:26
We've got CareGap analytics to
23:28
the doctor.
23:29
We've got
23:30
some kind of patient matching. I
23:32
want to get an appointment with the first available orthopedic
23:35
surgeon. We might have scheduling systems
23:37
and access
23:38
texting and chat options.
23:40
But
23:41
fundamentally, should I be going to that
23:43
doctor for this reason or should I be
23:45
going to see a physiatrist first to do some
23:47
more conservative treatment on my back pain?
23:49
Those kinds of decision support systems about
23:52
not accessing the system and
23:54
not using the thing that I'm looking to get into in
23:56
a fee-per-service way. There's
23:57
no financial incentive to build those.
24:00
And so that's an area where I see
24:02
great potential in the use of AI
24:04
and healthcare. I'm not 100% sure
24:07
that the current
24:08
economic incentives
24:10
will drive us towards the highest and best
24:12
use of this capability. That's really
24:14
interesting because it's really the combination
24:17
where you end up as the combination
24:20
of value-based care and super
24:22
consumerism. That's what puts
24:25
this on fire, so to speak. And
24:27
this is why I'm so bullish on policies
24:29
like voluntary alignment,
24:31
where I'm a Medicare patient, I can see any
24:33
doctor I want. I might go to a primary care
24:35
doctor and paternalistically, that
24:38
doctor chose to be in an ACO, so
24:40
I just follow suit. But you know what? They
24:43
haven't done anything for me. Maybe in theory,
24:45
they sent me a letter and said they're going to coordinate
24:47
my care, but
24:48
I'm not so sure I know what that means at the moment.
24:50
But
24:51
with voluntary alignment, there could be a
24:53
competing clinic down the street and say, you know
24:55
what?
24:56
Check us out. We've got a South Asian
24:58
heart clinic
24:59
and we might want to track you. One
25:01
of that's going to be monitoring-based and data-based.
25:04
And if you want, we can give you some specific
25:06
guidance based on your health needs. And
25:09
maybe if I switch to that doctor, I'm not switching
25:11
my health plan,
25:12
I'm switching my doctor, I might
25:14
be enrolled in their value-based care arrangement
25:17
through voluntary alignment. That's
25:19
a really powerful change to this. Oh, sign
25:21
me up. So if CMS is launching
25:23
that right now, calendar year 22,
25:26
it really took off in the ACO REACH
25:28
model.
25:29
Only a couple million people are enrolled in
25:31
that model, but 10% of them
25:33
signed a form that said, I want to voluntarily
25:36
align because you are offering
25:38
me something
25:39
that's going to feel like I'm getting more coordinated care.
25:42
That's a different model, direct to consumer
25:44
value-based,
25:46
than this sort of paternalistic, you
25:48
just do whatever you do and behind the scenes, we're
25:50
going to true up all the math and the books
25:52
and everything else.
25:53
I love consumer agency
25:55
in
25:56
voluntary alignment. And so let's
25:58
fingers crossed think.
27:59
ideal world, you'll have a two by two matrix.
28:02
What gets me productivity in
28:04
the enterprise and what gives me
28:07
improved clinical outcomes?
28:10
And so you really want to have a portfolio
28:12
approach of AI projects.
28:14
And perhaps in reality,
28:16
we're going to lean a little bit more on the
28:19
behind the scenes,
28:21
administrative burden reduction, use
28:24
cases, and be a little
28:26
bit more careful about when
28:28
to introduce. We don't want hallucinations. It's
28:31
too early. We have
28:32
not yet trained models
28:35
on longitudinal clean
28:37
health data to give
28:39
really good information about
28:42
what to do next. I have a path that
28:44
I see
28:45
to move us in that direction. You see
28:47
a clue of this in the announcement with Epic
28:49
and with Microsoft,
28:51
which is that we're going to train the AI models
28:54
to write really good queries to
28:56
ask questions of the sensitive data
28:59
if we're not going to give it itself the sensitive
29:01
data. So then it's like a derivative
29:04
works of AI, where you're
29:05
teaching it to ask smart questions,
29:08
but not actually exposing the PHI
29:10
in the process.
29:12
So I've spent some time with our technology
29:14
team on AI, just understanding
29:16
it and just cleaning the data
29:19
itself is most of the battle.
29:22
Yes. So let me separate out these concepts,
29:26
I think in threes. Today's
29:29
foundational model,
29:30
let's call it, it reads the internet
29:33
and everything that's been written in the internet and
29:36
is fed that through a machine and
29:38
these parameters are getting bigger
29:40
and bigger. So GPT-4 is like a thousand
29:43
X the size of GPT-3 and who
29:45
knows where GPT-5 will be, but
29:47
you can start to feel like
29:49
it's going to learn everything, whether we actively
29:51
and explicitly taught it healthcare specific
29:54
things or not. It's sort of a natural and
29:56
inevitable training exercise.
29:58
Not a lot of organizations have the cap. capital
30:00
to fund the kind of compute necessary
30:02
to do all of that work. So today, that's
30:05
in the hands of the few, the training.
30:08
But then we have this opportunity to take
30:10
that training
30:12
and to organize it
30:13
for our specific
30:15
environment. We can tune it.
30:17
So say I'm a health plan, I can
30:19
take all of my criteria
30:21
for prior authorization,
30:24
which maybe it doesn't know because it hasn't
30:26
read those materials. And I can add that
30:29
additional information into the mix.
30:31
And so there's going to be a layer of analytic work
30:34
to design the fine
30:37
tuning step. You're not going to do a lot
30:39
of data manipulation in that
30:41
step. You're mostly going to
30:43
feed it like
30:44
one of those incinerators. You can throw
30:46
back to the future style, you threw every bit of
30:49
food scraps into the machine that converted to
30:51
energy. I think it's going to be whatever data
30:53
sets you have
30:54
without it being fancily curated,
30:57
will go into this fine tuning step.
30:59
And then the third layer is the interaction
31:01
layer. And that's where I was describing the
31:03
business strategy of say, teaching
31:06
it
31:07
how to fish in
31:08
a data set that we otherwise don't want it to get
31:10
access to.
31:11
And it could learn how to investigate
31:13
that data set. So that
31:15
unlocking your sensitive PHI
31:18
through an
31:19
AI analyst
31:21
may be the near term
31:23
opportunity to
31:24
unlock the value. It doesn't
31:27
require what you're describing,
31:29
which is pre-generative AI.
31:32
When you were just doing predictive AI,
31:34
oh my goodness, you had to label
31:36
everything. And
31:39
so 80, 90% of the budget was data labeling.
31:41
Now that we're in a world of generative
31:43
AI,
31:44
it does create a new option
31:46
to engage with these tools without having
31:48
to do that data. Great. So
31:50
now you're with Care Journey, where
31:53
you've been the president for eight years.
31:55
You took this rollover shortly after
31:58
being the US chief.
31:59
technology officer. So
32:03
tell me about the mission and what you're doing
32:05
there. So to rewind the tape,
32:07
I actually ran for Lieutenant Governor of
32:09
Virginia
32:10
after I left USCTO. And
32:12
my dream
32:14
was that we would improve the public-private
32:16
handshake. There
32:17
would be opportunities to problem solve.
32:19
And I was hoping to bring state governments into
32:22
the future by saying for health, for energy,
32:24
for education, assume we have
32:26
the laws that we have,
32:28
we now want to execute those laws in
32:30
a way that help people. We can do a lot
32:33
more together if we could improve that
32:35
what I called handshakes and handoffs. The
32:37
government's largely agreed to open up a lot of
32:39
the digital infrastructure. And if we could hand
32:42
off that information to partners that are
32:44
trusted to solve problems, we can make a difference.
32:47
In the Affordable Care Act, one of those
32:49
examples was that CMS
32:52
agreed to release the longitudinal Medicare
32:54
claims data for
32:55
purposes of provider performance
32:57
measurement. I can't give my mom and dad recommendations
32:59
on who the best doctors are because the Yelp
33:02
reviews are all about things like is the room
33:04
clean and it's not exactly data-driven.
33:08
So Care Journey was born after my
33:10
campaign loss because my board chair and
33:12
a big supporter of mine, Sanju Bansal, co-founded
33:15
MicroStrategy, had a long history in technology
33:18
and data sets, said, Anish, if you want
33:20
to be on the government side encouraging the private
33:22
sector to do stuff, why don't you just step over
33:25
the line, stand on the private side, take
33:27
a lot of that information that's been public and help make
33:29
it useful. So we tried half a dozen ideas
33:32
from Blue Button to a whole range of ideas.
33:34
We put ideas at the wall and said, which of these
33:36
are going to have an impact in healthcare? And the
33:38
one that's taken off
33:40
is the idea that we could use the Medicare
33:42
data
33:43
for purposes of provider performance
33:46
measurement.
33:47
And it's still early days, but I'm grateful
33:49
that we're able to mine the largest
33:51
linked longitudinal data set
33:54
with openly available measures of quality
33:57
and outcomes and to bring that to
33:59
life through our own.
33:59
our membership program, we're serving over 130 organizations,
34:03
mostly ACOs, but health plans and
34:05
entities like US News, who's now incorporated
34:08
some of our data into their provider profile
34:10
pages. I love it because I feel like I'm
34:12
continuing the work I was doing in government
34:15
just on this side of the public-private handshake.
34:17
So we're going to wrap up here with a couple of questions
34:20
I always ask. And it is
34:22
perfect right after you talk about the
34:24
ideas that you put on the table around
34:27
how you can bring public and private together
34:29
for a project, what inspires you?
34:32
How do you get inspired? So Sam
34:35
Petrotta was my inspiration that you can actually
34:37
make meaningful change to a
34:39
billion people by bringing technology,
34:42
data, and innovation to solve problems. What a great
34:44
high on problem solving if
34:46
I'm not arguing over a tweak to an old law,
34:49
but I'm actually engineering a better way to
34:51
do what's already
34:52
in the political domain is consensus.
34:55
So boy, I wake up every day thinking
34:57
about how do we do more of that in energy and
34:59
health and education and banking.
35:01
Yeah, that's amazing. Inspired
35:03
from a very young age and still
35:05
today is what inspires you. That's great.
35:07
And what excites you most about healthcare
35:09
today?
35:11
It is the creation of the health information
35:13
fiduciary. I see the
35:15
sort of earliest of indicators that
35:18
value-based care organizations in
35:20
total cost of care with a voluntary
35:22
alignment and these new technologies.
35:25
They're going to usher in a completely
35:27
better care experience without any
35:29
new law or any new action
35:32
and budget. We can make the current
35:34
systems work
35:35
better. And so that
35:37
to see members of Care Journey I love
35:40
work on those problems gives me joy.
35:43
That's fantastic. Thank you so much. It
35:45
was, it's been a great conversation. We've got
35:47
the history of where we've been and how
35:49
we got where we are and the vision for
35:51
where we need to go. It's just been a delight
35:54
talking with you. Thanks so much for joining. There's
35:56
a better way.
35:57
Have a great day. Thanks for
35:58
having me.
36:03
Our conversation today, Anish, was
36:05
a great reminder for why I work
36:07
in healthcare and for why the work
36:10
we're doing here at Sure Scripts is so
36:12
important. I love how you
36:14
spoke about making meaningful change
36:17
by bringing together technology,
36:19
data, and innovation to
36:21
solve problems, and how
36:24
that can be done at scale for
36:26
a billion people, like your
36:28
father's famous classmate, Sam Petrota,
36:31
who helped to modernize telecommunications
36:33
in India. Of course, I'm
36:36
biased, but I love what you
36:38
said about nothing making you happier
36:40
than to see the hockey stick of e-prescribing
36:43
moving up to the right, in
36:46
no small part because of
36:47
Sure Scripts.
36:48
And this affirmed that we can
36:51
make a digital transformation in
36:54
healthcare. And I love
36:56
when President Obama needed a
36:58
White House advisor who could bridge
37:01
public and private partnerships in
37:03
the name of technological innovation.
37:07
Everything you've done in your life, like serving
37:09
as Virginia's fourth Secretary of Technology,
37:12
meant you were ready for
37:15
the role. Thank you for being on our show
37:17
today, Anish, and for inspiring
37:19
both me and our listeners.
37:26
Thank you for listening in today. If
37:28
you've enjoyed this podcast, please rate,
37:31
subscribe, and review. There's
37:33
a Better Way, smart talk on
37:36
healthcare and technology.
37:38
With your help, we'll
37:40
continue to bring great conversations
37:42
to the fore and to the wider
37:44
listening public.
37:46
Thank you.
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More