How Trailblazer Aneesh Chopra Makes Meaningful Change

Melanie Marcus: I am so pleased to welcome Aneesh Chopra to our podcast as our ninth and final guest this year. And who better to cap off our second season than the nation's first chief technology officer, who served under President Barack Obama?  

There was a lot that led up to that, though. It was 1995. Aneesh was at his first job at Morgan Stanley working in investment banking, and he had a front row seat when Netscape went public. It was here that Aneesh saw firsthand how technology—the Internet, in this case—could make huge changes in the world.  

Jump forward a decade to 2006, and Aneesh was serving as the fourth secretary of technology for the state of Virginia, where he led technological innovation in the state government. It was a kind of training ground for what was to come as the nation's first CTO. And he's still innovating today as president of health IT company Care Journey.  

Aneesh, tell us a little bit about yourself.  

Aneesh Chopra: I grew up as the son of Indian immigrants in New Jersey. Dad went to Villanova and then went home to marry Mom. He shared with me stories of this famous college classmate of his, Sam Pitroda, who was from rural parts of India but came to this country for graduate school. Sam was an entrepreneur who was an expert in the field of telecommunications, and the country was going through a transition from analog phones to digital infrastructure for phones. And so he caught that wave and was able to be a pioneering entrepreneur for innovation in telecommunications.  

He became a very successful entrepreneur and patent holder but decided in his thirties to go back to India for a penny-a-year salary and effectively became the chief technology officer for India. And in that particular time, the goal was to get access to telephony to all 100,000+ rural villages. I think India had 3,000 phone lines for a nation of 300 million people.  

And you can say, “We need a multibillion-dollar subsidy to bring all this Western phone technology to the rural parts of the country.” Or another political party would say, “Well, we can't afford it. Lower taxes, let the market solve it.”  

And the engineer said, “Hey, there's a better way. We can actually invent a domestic Indian approach that was able to—for $30 million—design a native, digital-first, modern telecommunications pipe.” And that allowed the private sector to work off of the open standard to actually extend that telephony. So by the end of the decade, without a huge amount of political-subsidy arguments, they were able to extend phone service to every remote village.  

And that story I share with you because it was in the back of my mind while I was a policy geek and interested in government. I always had the lens of an engineer's mind.  

Marcus: Let's dig in on your role serving with President Obama as the first chief technology officer for the nation. How did that role come about?  

Chopra: President Obama was campaigning and said that part of what he wanted to do was tap the technical capabilities of the country to solve problems. If you remember, President Obama was a bottom-up change agent, not a top-down change agent. And he had seen firsthand from the campaign how technology allowed for everyday Americans to get involved. I don't mean this to be a political statement, but his philosophy was to bring that mindset to governing.  

Now, at the time, there were very few state governments that were similarly inspired to put technology in the Cabinet rank. Virginia happened to be one of them. I was Virginia’s secretary of technology. We were just awarded best-managed state, best state for business, best state to raise a family. And the dot that connected was that my governor, Governor Tim Kaine, was a finalist to be vice president. So as the Obama team was doing diligence on Governor Kaine, they saw his success managing a world-class government. And it was bipartisan, by the way. We have a tradition in Virginia that governing be a priority regardless of party. And that resulted in me serving on the transition team, helping to advise on what it is that a technology secretary does in government.  

Marcus: How did your focus on healthcare happen?  

Chopra: We were in the middle of an economic recovery and the president, to his credit, said, “I'm not going to just throw cash in terms of stimulus”—which was needed at the time. About $100 billion of the Recovery Act was dedicated to these seed investments for national priorities. So, we put $10 billion into the electrical grid. We put $30–40 billion into broadband expansion. We did work, obviously, with the HITECH Act for electronic health records deployment.  

So that portfolio happened before I was even nominated or Senate-confirmed. So, there was already a healthcare IT agenda that had people in charge of execution of the plan. We had ONC and they were doing a great job. But specifically, the president wanted an adviser in the White House that could cut across all of the business units or the functioning parts of government and the public/private-sector relationships to make sure we were maximizing the value of technology, data and innovation. 

A big gap in our health IT strategy was that it was the last vestige of on-prem, locally constructed software. We were entering the period of Internet-based, cloud-based electronic health record systems, the chance to move to modern Internet standards for data sharing.  

What is now the FHIR API ecosystem began as a modest $15 million R&D grant that ONC administered—and that I had a big role in championing. And that, of course, was the beginning of what now is a regulated Internet standard for the entire industry, which I'm very excited about.  

Marcus: Absolutely. Take us through that story of what happened from 2010 or so to today.  

Chopra: We wanted to build a more population-health-based care delivery system. That was obviously what was anticipated with the later Affordable Care Act. But for reasons of timing and economic reality, we frontloaded the IT investment. And as your audience might know, you don't put tech in front of care redesign. In a dream world, you want care redesign to drive the tech.  

But it's okay. We were given the hand we were given, so we needed to land a pre-demand-signal Meaningful Use construct that was really geared towards value-based care.  

We thought we could do that with the subsidies to say, “Look, I know you don't want to record blood pressure in a discrete place, or smoking status.” It was seen as data-entry clerk-type work and frustrating. But it was critical if you're going to look longitudinally at the population to say who should be engaged to reduce their risk of heart attacks.  

The next, triple-axel piece was that if we could gently nudge the industry over a decade, eventually we would see demand for change come less from government regulations and more from the concept of value-based care. The Affordable Care Act would take hold. CIOs would tell the EHR systems, “This is the way we want to prioritize our work anyway,” and we'd emerge in this beautiful, orchestrated ten-year roadmap.  

We're laughing because obviously we did not land the triple axel. So, we had a rough patch. Did people do the government compliance with gusto and love, or was it the minimum necessary to get through the regulatory loopholes and get right back to the fee-for-service features that customers really wanted to buy?  

And so, we had this frustrating underinvestment in stuff that worked and overinvestment in stuff that wasn't really aligned with the vision of where we're heading. We were increasing pressure points to say, “You've got to move to these more detailed standards.”  

It's been a bit of a challenge for the first part of the decade, but we come to you with good news. With the Cures Act and where we are today, now I believe there is a more balanced demand signal for these value-based features and the market is building up that natural request. And so, with the implementation of the FHIR APIs in the Cures Act and the roadmap on where we're heading to make more progress in more areas, it's going to feel a lot better this next decade than the spotty, transactional history we had in the first.  

Marcus: Without the first, you wouldn't have the second. There's a lot of progress that's been made. We've got an electronic health record adopted. We have the concept of value-based care, and we keep saying, “When is it going to tilt the scales?” And it hasn’t quite yet.  

Chopra: Maybe we could have ripped the Band-Aid off sooner and gone to Internet-based APIs in Stage Two, and not waited for Stage Three and the Cures Act. Would that have put us in a better place? Who knows?  

That’s the kind of academic discussion it’s enjoyable to have a podcast to discuss. But remember, Surescripts was a huge part of the hockey stick moving up and to the right. Nothing made us happier than to see the Surescripts e-prescribing transactions going up, and it was an affirmation that we can move to digital and it can have an impact. I wish we saw the same trend for the rest of the transactions.  

Marcus: That is progress that needs to continue to happen. Let's go to interoperability, because that's where we're headed. You just talked about the Cures Act. And interoperability—the ability to exchange health information for the right person, the right time when they need it—is still evolving.  

Chopra: I'm more bullish on where we are now than I think the perception is. So that's why I was keen to have this discussion.  

Marcus: Oh! So tell me why you're bullish. 

Chopra: Okay. Let me do a process reason and then a substantive reason.  

The process reason is we flipped it—I think correctly—to engineer interoperability to the patient as the first use case for the new technology standard. And the reason why that's critical is that there doesn't need to be a trust framework and a governance and a national TEFCA and a whole litany of things. All of that falls by the wayside when a consumer invokes her right of access.  

And if we could supply that right of access with the technical approach that allows anyone to put their health information on their phone—today, every American can put their medical record, their common clinical data set now called USCDI, on their phone for free if they have an iPhone or an Android phone, without any middleman, vendor, fee structure... It works. Because everyone knows how to point data to the consumer. And so we spend more time getting the technical standards right in terms of the data model and less time arguing over the “trust framework” on how you can request data of me, and I can determine whether it's appropriate, and then respond with the minimum data necessary and then look up in the contract: are we allowed to share this, and do we want to share this, and what are the business terms of sharing it?  

Generation One of interoperability, the original four NHIN networks, two of them were in Virginia, so I had a front-row seat. When you start with B2B and then you want to get technical consensus on data model and technical consensus on transmission method, you're putting in a lot of burden. You can barely get out of the room years and hours later with just consensus on what's the minimum data necessary when you're asking for data in the context of a care transition. So that's why we've been stuck in treatment-only, CCD, minimum-data-necessary plans in B2B interop for the last decade.  

But now, because we first focused on data elements to the consumer, we can reapply that same technology stack to the thornier business model challenges. And the key transition technology from the consumer-facing world to the B2B world is the advent of bulk FHIR—that piece of technology that has now shipped across 270+ EHR-certified products as of 12/31/22. We now have the technical ability to implement data sharing, facilitating it point to point, without, let's call it, special effort, as the language of Meaningful Use called for and the Cures Act called for.  

That's a gift. Now, what's missing are the contracts and the negotiations to unlock it. But maybe this is an assignment where your colleagues at Surescripts might say, “We're best positioned to facilitate point-to-point data sharing through bulk FHIR to bring that future interoperability world that seems like it's another half a decade away to something that can be demonstrated this calendar year.” 

Marcus: Let's go back to data being deployed to the patient or the consumer. Having tried some of those things, it does take it outside of HIPAA protection, and most consumers don't know what that means.  

Chopra: Yeah, so this is an important part of my—call it a volunteer job, with Governor Leavitt, with David Blumenthal, with David Brailer. We co-chaired the creation of the CARIN Alliance as a kind of multi-stakeholder collaborative. One of the first deliverables we shipped was a voluntary but enforceable code of conduct, and in the voluntary but enforceable code of conduct—believe it or not—we exceeded HIPAA requirements in terms of trust.  

As an example, today, there's literally no disclosure when my hospital takes my data, deidentifies it and sells it to third parties. I get no notice, I get no agency to say that I'd rather not do that. But because it's outside of HIPAA, it happens every day, and it's a multibillion-dollar market where people are buying and selling my deidentified data.  

In the CARIN Alliance code of conduct, we required any app that agreed to take on this code to publicize what their policy is with respect to deidentification and to make sure that policy was carried all the way through its supply chain of third-party vendors.  

Now, are there people that say in their disclosure, “We have the right to sell this data and you don't have a choice”? I'm sure, because that's how the healthcare system operates. But at least there's disclosure. My doctor doesn't tell me that disclosure. My hospital doesn't tell me that disclosure. And so, I thought that was progress.  

But let's go backwards. Before we go to problems, let me describe why that was an opportunity for goodness. As you remember, we made the business decision in our policy that you can keep proprietary data models in healthcare. We regulated the conversion of some of those data elements into what is now the USCDI and made them a FHIR-based data model. So that data element now sits in a format that has no intellectual-property constraint. Anyone, you or me, can read a FHIR data model. I don't have to pay anybody for a code set to learn what this FHIR language means in the real world.  

That's not how the rest of healthcare operates. I have to pay to interpret CPT codes and this and that and the other. Everyone's got a hand in the cookie jar. But in the FHIR data model, it's completely free and open source, which means everybody who's subject to regulation is moving their proprietary data sets into an open data model.  

This is critical because it then doesn't matter whether the destination is the consumer who's requesting it or whether I use that to share information for a value-based care contract. That data model is open. So because we built to the consumer first, we got everyone to build Version One of an open data model. And the transmission method was to a consumer-designated app triggered by their username and password. With bulk FHIR, we've now got a second transmission method where you can aggregate patients into a cohort or a registry, and then you can transfer all of the records in that registry to a third-party application that doesn't need additional patient consent or to involve the consumer in any kind of administrative friction, if they follow existing legal frameworks for data sharing.  

Marcus: Let's go to artificial intelligence. We’ve got this amazing tool that is going like wildfire. It's going at exponential pace across the country, across the world.  

Chopra: I’m more giddy about this moment than when we saw the Internet blossom. This feels like I'm a kid in a candy store with excitement about ways to solve big, meaningful problems in a relatively short period of time and at low marginal cost. So let's go.  

Marcus: Where does your mind go immediately around artificial intelligence? Let's start there.  

Chopra: So, I've been speaking of a dream for a digital health advisor or what I refer to as a health information fiduciary service, whose role is to basically run decision support in my best interest, based on the information that it has about me.  

In so much of our healthcare system, the data is inert until it's too late, and then we show up in the emergency room and then the systems start running and then everybody starts asking questions and, oh my goodness, putting you on dialysis—even though if we knew you had kidney disease even just two years earlier, we could have put you on the new meds that could actually eliminate the need to be on dialysis. Are you kidding me? How could we have missed that as an opportunity?  

So, part of this is, at scale, if one has access to longitudinal information and is trusted, then we can run these models to predict the next best course of action that might involve me seeing a new doctor, getting a new test, adopting a new care plan... These are all known if we can understand and read all of PubMed and all of the recommendations: people like me have had an interest in this, and you should track for that; I'm of South Asian descent, so I've got a genetic predisposition to heart disease and I should be monitoring and making sure I'm on a statin a little bit earlier, even if the guidelines say this or that. So how do we create a marketplace that will reward health information fiduciary investments?  

The challenge I see is that the technology is often put to its highest and best economic use, which may be different from the highest and best patient use. Highest and best economic use is, hey, I'm in a Medicare Advantage (MA) plan, they need to get revenue by looking up my health conditions. There, the highest and best use of AI might actually be risk-adjustment coding—which is about getting paid more for what I have and the conditions that I have.  

And I say that in a slightly negative way—I don't mean to be negative, I love everybody, I love MA and everything. But our economic incentives today don't reward organizations to invest in consumer decision support. We've got care-gap analytics to the doctor. We've got some kind of patient matching. I want to get an appointment with the first available orthopedic surgeon, and we might have scheduling systems and texting and chat options. But fundamentally, should I be going to that doctor for this reason, or should I be going to see a physiatrist first to do some more conservative treatment of my back pain?  

Those kinds of decision-support systems about not accessing the system and not using the thing that I'm looking to get into in a fee-for-service way—there’s no financial incentive to build those. And so that's an area where I see great potential in the use of AI in healthcare. I'm not 100% sure that the current economic incentives will drive us towards the highest and best use of this capability.  

Marcus: That's really interesting, because where you end up is the combination of value-based care and super-consumerism. That's what puts this on fire, so to speak.  

Chopra: And this is why I'm so bullish on policies like voluntary alignment. Now, if I'm a Medicare patient, I can see any doctor I want, I might go to a primary care doctor and paternalistically, that doctor chose to be in an ACO, so I just follow suit. But you know what? They haven't done anything for me. Maybe in theory they sent me a letter and said they're going to coordinate my care, but I'm not so sure I know what that means at the moment.  

But with voluntary alignment, there could be a competing clinic down the street that says, “You know what? Check us out. We've got a South Asian heart clinic and we might want to track you. That's going to be monitoring-based and data-based. And if you want, we can give you some specific guidance based on your health needs.” And maybe if I switched to that doctor—I'm not switching my health plan, I'm switching my doctor—I might be enrolled in their value-based care arrangement through voluntary alignment. That's a really powerful change.  

Marcus: Oh, sign me up! 

Chopra: So CMS is launching that right now. In 2022, it really took off in the ACO REACH Model. Only a couple of million people are enrolled in that model, but 10% of them signed a form that said, “I want to voluntarily align because you are offering me something that's going to feel like I'm getting more coordinated care.”  

That's a different model—direct-to-consumer value-based—than this sort of paternalistic, you just do whatever you do and behind the scenes we're going to true up all the math and the books and everything else. I love consumer agency in voluntary alignment, and so let's—fingers crossed—think that the highest and best use of AI will be in organizations that embrace voluntary alignment, that take total cost-of-care responsibility and commit to quality measurement and the like. And that they're going to do all of that in my best interest. They're the ones that are able to see those investments pay out in shared savings.  

Marcus: That's an amazing vision. As I said, sign me up!  

Chopra: My mom and dad are both on Medicare—touch wood, they're reasonably healthy—but they've got chronic conditions, and they go through this process every year where there's an insurance broker. They live in Florida, so they get spammed every minute for MA plans. They want freedom of choice, they're going to stick with med supp. But here's the interesting thing. They spend hours with their trusted broker: Do I do MA this year or not? Do I switch my Part D plan or not? Do I choose this med supp option or not?  

Ten hours of conversation. Did they ever once get asked by that broker, “Would you like to be in an accountable care arrangement and would you like to voluntarily align? So it's not just which plan, but which doctor is going to build the services to give you 24/7 access, coordinate your care, be responsible?” They live literally in the epicenter of value-based and they’re not offered anything. They don't know if they're in an ACO.  

Man, we've made this so paternalistic. We don't have to be! Let's be practical. In the dream world, we want AI to be helping consumers make better decisions. In the right-now world, it's going to be used to deal with denials and call center volume and operational needs.  

So in an ideal world you'll have a 2x2 matrix: what gets me productivity in the enterprise and what gives me improved clinical outcomes. And so you really want to have a portfolio approach of AI projects. And perhaps in reality we're going to lean a little bit more on the behind-the-scenes administrative-burden reduction use cases and be a little bit more careful about when to introduce it. We don't want hallucinations. It's too early. We have not yet trained models on longitudinal clean health data to give really good information about what to do next.  

I have a path that I see to move us in that direction. You see a clue of this in the announcement with Epic and with Microsoft, which is that we're going to train the AI models to write really good queries, to ask questions of the sensitive data—teaching it to ask smart questions but not actually exposing the PHI in the process.  

Marcus: So I've spent some time with our technology team on AI, just understanding it, and just cleaning the data itself is most of the battle.  

Chopra: Yes. So let's separate out these concepts. I think in threes. Today's foundational model, let's call it, it reads the Internet and everything that's been written in the Internet and is fed that through a machine, and these parameters are getting bigger and bigger. So GPT-4 is like a thousand times the size of GPT-3, and who knows where GPT-5 will be, but you can start to feel like it's going to learn everything, whether we actively and explicitly taught it healthcare-specific things or not. It's sort of a natural and inevitable training exercise.  

Not a lot of organizations have the capital to fund the kind of computing necessary to do all of that work. So today, that's in the hands of the few, the training. But then we have this opportunity to take that training and organize it for our specific environment. We can tune it.  

So, say I'm a health plan. I can take all of my criteria for prior authorization—which maybe it doesn't know because it hasn't read those materials—and I can add that additional information into the mix. And so there's going to be a layer of analytic work to design the fine-tuning step. You're not going to do a lot of data manipulation in that step. You're mostly going to feed it, like one of those incinerators—like "Back to the Future” style, where he threw every bit of food scraps into the machine that converted it to energy. Whatever data sets you have will go into this fine-tuning step.  

And then the third layer is the interaction layer, and that's where I was describing the business strategy of, say, teaching it how to fish in a data set that we otherwise don't want it to get access to. And it could learn how to investigate that data set. So unlocking your sensitive PHI through an AI analyst may be the near-term opportunity to unlock the value—which doesn’t require what you're describing, which is pre-generative-AI. When you were just doing predictive AI, oh my goodness, you had to label everything. And so 80–90% of the budget was data labeling. Now that we're in a world of generative AI, it does create a new option to engage with these tools without having to do that.  

Marcus: So now you're with Care Journey, where you've been the president for eight years. You took this role over shortly after being the U.S. chief technology officer. So tell me about the mission and what you're doing there.  

Chopra: So to rewind the tape, I actually ran for lieutenant governor of Virginia after I left U.S. CTO. And my dream was that we would improve the public/private handshake. There would be opportunities to problem solve. And I was hoping to bring state governments into the future by saying—for health, for energy, for education—assume we have the laws that we have, we now want to execute those laws in a way that helps people. We can do a lot more together if we could improve what I call handshakes and handoffs. The government's largely agreed to open up a lot of the digital infrastructure, and if we could hand off that information to partners that are trusted to solve problems, we could make a difference. 

In the Affordable Care Act, one of those examples was that CMS agreed to release the longitudinal Medicare claims data for purposes of provider performance measurement. I can't give my mom and dad recommendations on who the best doctors are because the Yelp reviews are all about things like, is the room clean? And it's not exactly data driven. So Care Journey was born after my campaign loss because my board chair and a big supporter of mine, Sanju Bansal—who co-founded MicroStrategy, had a long history in technology and data sets—said, “Aneesh, if you want to be on the government side encouraging the private sector to do stuff, why don't you just step over the line, stand on the private side, take a lot of that information that's been public and help make it useful?”  

So we tried half a dozen ideas, from Blue Button to a whole range of ideas. We put ideas on the wall and said, “Which of these are going to have an impact on healthcare?” And the one that's taken off is the idea that we can use the Medicare data for purposes of provider performance measurement. And it's still early days, but I'm grateful that we're able to mine the largest linked longitudinal data set with openly available measures of quality and outcomes, and to bring that to life through our membership program. We're serving over 130 organizations, mostly ACOs, but also health plans and entities like U.S. News, who's now incorporated some of our data into their provider profile pages. I love it because I feel like I'm continuing the work I was doing in government, just on this side of the public/private handshake.  

Marcus: So we're going to wrap up here with a couple of questions I always ask. What inspires you?  

Chopra: So Sam Pitroda was my inspiration that you can actually make meaningful change to a billion people by bringing technology, data and innovation to solve problems. What a great high on problem solving, if I'm not arguing over a tweak to an old law, but I'm actually engineering a better way to do what's already in the political domain as consensus. So, boy, I wake up every day thinking about how we do more of that in energy and health and education and banking. 

Marcus: And what excites you most about healthcare today?  

Chopra: It is the creation of the health information fiduciary. I see the earliest of indicators that value-based care organizations in total cost of care with a voluntary alignment and these new technologies—they're going to usher in a completely better care experience without any new law or any new action and budget. We can make the current systems work better.