Inside CMS: How Medicare Innovation Models Are Working to Lower Costs & Scale Digital Health

Friends, it's another very special episode of A Health Podyssey. And today, it's a real privilege to welcome Abe Sutton, Deputy Administrator for the Centers for Medicare and Medicaid Services and Director of the Center for Medicare and Medicaid Innovation, the Innovation Center. And as regular listeners know, the Innovation Center was created under the Affordable Care Act to test novel healthcare payment and delivery models. Now fifteen years later, it is the beating heart of the federal government's efforts to spark transformation in healthcare.

Abe Sutton, welcome to our Humble podcast. Thank you for having me on. It's a privilege to be here. And, we asked Abe to come and talk about one model in particular, the access model, and we'll dig into that in just a moment.

But before we do, I thought maybe we could take a step back and just talk briefly about the Innovation Center's work writ large. So if you're ready, we can dive right in. Let's go for it. Maybe just start.

Our focus is really on increasing the affordability of healthcare in this country. If I think about the statutory mandate, it's about improving quality and lowering costs. We have a world class healthcare system. We have new innovations that are available.

We have advancements and different strategies for giving care, but too many people, cannot engage with that system. Seek care is really inaccessible and unaffordable, in America today. I'm not only talking about the Medicare, Medicaid programs who refocus, but in the employer market as well. I think of our purpose in the innovation center as changing the incentive structures, payment and delivery reform so that care can be delivered more affordably in America.

So we change the way that people think and approach the market and care really becomes more accessible. To break that down into a few buckets, if I may, we have 10 new models that we've announced so far in this term. Four focus on the affordability of drugs in this country and continuing to incentive as advancement on pharmaceuticals. Three are about outcome aligned payments.

ACO models, the access model that I know we'll touch on, and really about alignment of incentives so that you're focused on delivering the right outcome. And then the third bucket is really around prevention. It's around getting upstream. It's around, the basic research on what different interventions and wellness interventions can make a difference upstream.

And so I think that through line across all of them is about making care more affordable and more accessible. So if we look back over the past fifteen years of the center, it's obviously tested dozens and dozens of different models. You just said, you know, there are 10 more in this past year. But so far, most of the models that have sort of worked their way through this system have not met the criteria for nationwide expansion after evaluation.

Why is that so hard? Think of our job as akin to a venture capitalist. It's not about how many of your models work versus how many do not. It's about what is the cumulative impact of the models that work and do they justify the investment that has been made?

And so a classic VC model, maybe you do 20 investments and one of them works and returns all the costs of what you put into the 20 investments and then some. And that's the classic model there. It encourages bold bets. It's not about how many work, it's about the cumulative impact on the healthcare system.

And so if I think about a model that really means that people are getting better care, that really means we're going upstream and prevent it and works, it could pay for running 10 models that it turns out were miscalibrated in some way or didn't work because we were taking bold bets. I'm okay with that. And I think that was the intent of having an innovation center. We have learned from practically every model that has happened in the center's history.

Some of them we've learned, okay, if you change the incentive in this way, it won't move the needle enough because you won't have a large enough mindshare. And others we've learned, oh, you just need to recalibrate it this way to account for that externality and have a quality measure on that. And then you could retest it in a different way and we could iterate. We've also seen things where maybe we didn't certify the model, but we were able to draw out an approach and actually embed it within the existing statutory authority of how Medicare works.

And so there are many different ways to approach payment in healthcare, and we've pushed forward learnings through these different iterative approaches. I'll also call out like a more focused portfolio as a result of this iteration. So we did a ton of episodes. Bundles and episodes, I think, were all the rage at the start of the Innovation Center's history, as your listeners will know well, just given this audience.

And episodes theoretically make a ton of sense. So we have episodes in our portfolio in what we do, but we can't get episodes right always. It's hard to calibrate the exact time length. Should this be thirty days, forty five, sixty, ninety?

What code should be in the episode versus what code should be outside of the episode? Oh, did we miss one where then there's a surge of utilization on that code apart from this? Is this the right setting? Should we include the post acute setting in how we're thinking about this?

And I say all this to demonstrate is that it's really hard to calibrate and do an episode well. We're trying to shift on more on ACOs and more on some of the other areas where we could push forward innovation and open source the design of episodes going forward, which is what our CMS administered risk arrangements are all about. It's about letting us collect data as ACOs and specialists enter their own arrangements on that. I look at the history of the center and I actually think it's pretty reasonable for what we set out to do.

And I also think that the return looks better in the future, both on quality and on costs due to the way that models have now been calibrated and the learnings we've had about how to design them. Fair enough. Well, perhaps we can look at the Model as sort of a case study for some of those challenges that you're navigating. As you know, obviously, Access Model aims to increase access to technology supported care services for people with traditional Medicare.

And in the past, Medicare hasn't had a straightforward mechanism to pay for this kind of care. We're talking about apps and online patient portals and various technology supported services. And so under this model, these companies will now begin to be reimbursed under Medicare Part B. There's an outcomes aligned payment approach to it where organizations receive recurring payments for managing qualifying conditions with full payment tied to measurable health outcomes.

So in that context, some have raised concerns, including author Maurice Shah and colleagues in recent pages of Health Affairs Forefront, that because digital health companies have usually a pretty sophisticated patient acquisition capabilities, this model might be particularly prone to cherry picking, to sort of very targeted patient identification, which could then sort of lead toward potential gaming around benchmarks. Is there a risk of this happening, and do you share that concern?

I don't share that concern. And part of that business was intensely debated and thought through in the innovation center before the model went out. Part of that is due to the design of the access model and the access models quality measures. So let's talk about what it takes to drive improvement, in this model.

One, you need to have the patient population engaged. If the patient's engaged, you're getting the monthly payment, and if not, you're not. Two, you need to show improvement and, you need to show improvement on all of the quality measures associated with the track for at least 50% of the population you engage. So you don't need to be perfect, but you have to for 50% drive improvement on all five metrics.

And so just to bring this to life a little for, the cardio, kidney, metabolic health tracks, we're talking about things like A1C levels or BMI. And what we do is we have you take a measure at the beginning where we need actual data. And then a year later, we're taking a measure of where they are and you have to drive improvement, that surpasses the targeted improvement amount for fifty percent of the patients that you're engaging. What I like about this design is that you're selecting people and you're actually improving how they measure on these metrics.

Even if you select people who you think your strategy is likely to drive improvement, okay, I don't consider that gaming. You're literally improving people's health outcomes. That seems like a good outcome. And there might be other people who are better positioned to drive improvement with a slightly different population.

So they'll have a different go to market and engage that other population. The nice thing about our payment design coupled with this quality measures and quality incentive structure is that we're paying relatively little. So our calculation is that if you could take a set of patients who are at one end of the extreme of our measures here and just drive improvement in the required amount for that patient population, it is worth it for us to pay the amounts that we are paying here.

The other note I have is that the improvement is relative to where the patient was. So it's not like it's an arbitrary cutoff of you have a BMI of thirty five and you have to get to thirty to show improvement, but somebody who is thirty six also has to get to thirty such that it's worth getting. It's about the improvement relative to where the patient was. And so I think there will be different strategies and different interventions that are more effective at people with higher BMIs or people with lower BMIs who are still too high for where we want to be, and different companies will adopt different strategies.

Great. So another recent Health Affairs article, this one by Andrew Rundle and colleagues, has pointed to the fact that the Medicare population is quite different from the population that up until now has typically used these platforms. There are higher rates of limited English proficiency, lower rates of health tech literacy, and that that might make it harder for providers and these services to achieve key outcomes, at least compared to how they've been doing so far. What's your office thinking about this challenge and how do you suggest they navigate it?

I think they're 100% right here. I think it is going to be harder to be successful here. I think that is going to be something that the participants in the model are going to need to grapple with and in signing up for this, are taking on the risk of, they do that? And I think that's what it should be.

This is a model and a test, and we want to see people come in and innovate for different ways to do so, to the betterment of the Medicare beneficiaries and for the benefit of the taxpayer that funds this. I think there's pretty good direct translation technology out there nowadays that getting easily from one language to another should be relatively straightforward, but the likelihood of somebody to engage and trust the system and rely on it easily is going to be different. I think engagement broadly is going to be a challenge here that these companies are going to need to take on and invest in if they're going to be the solution that wins out and can scale and serve millions of people in the original Medicare program.

That said, we have a pretty big incentive and reward for doing so. The payments are low per patient, but if you scale to original Medicare and there's no gatekeeper in the sense that a Medicare Advantage or a commercial plan would have a gatekeeper contracting that's in a select one of the successful pilots to then offer nationwide to everyone, we're letting you go direct to consumer. We're letting you go to physicians or hospitals and say, this is right for your patients without a gatekeeper review, as long as you meet our criteria.

So if you prove out that you are the best solution and we publish those results on our website of here's how all the participants are doing, that's a pretty big market opportunity for you. And so there's a real incentive to invest in trying to improve and reach the patient population that access serves. And I think that is important because I don't want to live in a country where the only people who could benefit from care accessible via digital therapeutics between office visits or those who could afford to pay out of pocket, or those who have a very, very good commercial plan.

I want people in Medicare, in the original Medicare program, in Medicaid to have access to these supports as well, is I think that's where we can potentially see the largest benefit of some of these tracking, of some of these prompts. And so I want that to be accessible to the Americans who benefit and are on the programs that the Innovation Center has accountability for. I want to ask a little bit about the role of primary care physicians in this model. And my understanding is that in addition to paying the health technology companies, there's also a piece where physicians receive a small per member per month payment for reviewing digital health plans and coordinating care.

Some have said that that payment is too small, especially if these same providers are bearing potential liability related to providing care to these patients through these platforms. How are you expecting primary care physicians to respond to these rates, and how are you thinking about that in the context of the broader model? It was really important to me when setting this up to impose an obligation on participants to share information back to the providers that people engage with.

And so the participants have an obligation, not just to share something with health information, exchange, but to actually take information and push it to, the designated provider that a patient works with. That is important. It's the first time that we've really imposed a push to a specific provider as opposed to sharing in a network and then letting the provider pull. And I thought that was important just for connectivity and continuity of care.

So let me just say that. We also wanted to then give something to make it economical to ingest that data and review it as you're sitting with a patient and giving them guidance for their care as a physician or as a clinician. And so that's where the co management piece of this emerged. And it's really about having your systems configured to ingest that and then incorporating that in the care plan for the patient.

So there's that connectivity in place. And so we wanted to reward and incentivize this. And what we're going to do is closely monitor both how that plays out and uptake broadly in the model, and then learn and iterate over time. And so it's helpful to have feedback from folks.

We encourage folks to share their perspective with us. We'll also look closely at the data on how this is playing out, what the uptake is, and if this is proving effective and improving the quality of health in this country and lowering the cost of care in this country, because those are the ultimate metrics that Congress has given us to focus on as a center. So is there a world where down the road you might increase the physician payment rate? Anything's possible, right?

We'd have to see the data. We'd have to look at what happens as the model plays out and then react based off of that. I would just keep us anchored on making sure that care is affordable for the American people. I truly want to see us meet the moment of the promising innovation of agentic solutions, augmented solutions with a deflationary payment approach.

That means that the benefit of this flows out to our health system broadly and the American consumer when they're engaging with the healthcare system. Just thinking back to Meaningful Use, we basically saw the digitization of health and healthcare records, put in, in a way that didn't really seem to change the cost trajectory, certainly not in a cost saving direction. Let's think back to how we have bundles existing in our healthcare system, but then as new technology advancements came online, we had to have new technology add on payments outside of the bundle because of the distorted incentive structure that this created.

I'd like to see access and I'd like to see our broader governmental approach to paying for AI in healthcare, not be something where we end up leading to higher payments going out and more cost, because ultimately that flows through to making care more unaffordable, not only for Medicare and Medicaid beneficiaries and the programs that pay for them, but in how it's referenced to the commercial market as well. And so getting this right and ensuring that our payments are appropriate and encourage uptake, but do not inflate costs is important.

One more question about the access model and we're speaking here at Health Affairs where sort of the rigor of evaluation and analysis is sort of at the heart of our mission. And I know when we sort of think about how these models might be sort of changing for a different population, that in the past, these companies have sort of pitched their platforms to corporate HR and benefits managers for whom the standards of data and evidence might not always be sort of super rigorous compared to, for example, physicians and physician societies.

And so I guess I'm thinking about sort of the new potential customers that the model brings in, physicians who might be looking for more peer reviewed studies as opposed to PowerPoint slide decks? And big picture question here is about sort of the maturation and the development of the evidence of these technologies and the market. Do you expect there to be a higher standard of evidence among physicians who are expecting to refer patients to these programs? And how does that potentially affect participation?

I think the first thing to start with here is our evaluation strategy. And to inform that and draw real world evidence and real world conclusions, we are using a randomized approach where we're going to be taking one out of every 10 beneficiaries, that sign up and we're going to put them in a control group. And the reason we could do this is that this is not an existing benefit. So this is an expansion of a benefit to a new population.

And so we're able to say, thank you for showing you were qualified and expressing interest in signing up with this participant. You're in the control group. And, the benefit from that is obviously we'll see how the people in the control group compare to the people who are in the model itself. Now, if participation is off the charts and we have millions of people signing up, we obviously will need a lower number of people in the control, lower percentage, that's just what is necessary to draw a valid conclusion.

If the participation is lower, we could increase that. It's a model. We have the ability to make changes and flex over time. But I think we'll generate real world evidence on how the Medicare population behaves, from the people who have opted to sign up, who we put in to the model versus put into the control.

And we're going to put that for every participant out on our website for everyone to access, and we'll show how different participants compare with one another. And so this will all be accessible to not only every doctor and every hospital and every accountable care organizations, but frankly to every beneficiary who wants to just go and look it up and see. And I think that will impact uptake and direct it towards those that are most effective at both engaging people and improving their health outcomes.

And so will just say randomization, I think is an incredibly powerful tool that we should consider more broadly in our models. We can only do it when creating a new benefit category because we can't restrict access for a beneficiary to something already covered in Medicare, but I am very confident and very excited in the design of how we're going to evaluate this model. In fact, I might be more confident and excited about the evaluation approach to the model than I am about the model is transformative as I think bringing digital therapeutics to the Medicare population and original Medicare is going to be.

The evaluation part is one of the most exciting parts of this model. Cool. Well, you're speaking our language here at Health Affairs and our audience as well, so I know I'm excited to read that evaluation, and we'll look forward to seeing what comes out of it. And I appreciate that you guys have made that priority.

Well, before we wrap up, thank you for really digging into the weeds on the AXIS model. I really appreciate it. I did want to ask you about another set of innovation center models, which is something you alluded to in your introduction, which is those related to drug prices and various versions of what what we're broadly calling most favored nation drug pricing. You recently announced the postponement of the balanced model, And then in addition to that, there's the generous model, which is focused on states negotiating most favored nation rates.

There's the GLOBE model and the GUARD model. Obviously, Medicaid has long had a low, you know, lowest price program. And I guess I'm curious if you're worried that this sort of wall of sound approach may be sending confusing signals or multiple signals to the markets, and curious how, the Innovation Center is coordinating all of these different programs where there's a good amount of overlap and what you expect the markets to sort of take away from that. I called out drugs as one of the interest areas because drugs, you know, biotech and AI, large language models are probably the two areas where we have the potential for leveraging advancements in a way that is truly deflationary in the healthcare system.

Things that do not involve one to one labor input to delivery, but things that are capitalizable. And so drug advancements could potentially avoid hospitalizations, could avoid a lot of treatments. Same thing with getting people engaged in a preventive health activity, getting people to have support between office visits from AI. And so I'm excited about both of these areas.

On the drug space, I came into a team that was pretty small relative to the potential in the innovation center. We had 19 FTEs dedicated to working on drug models. We've now grown that to 41, and so around double the team. We had one drug model, that had been designed, negotiated with manufacturers, the cell and gene therapy model for sickle cell, and coming in, there was some uncertainty what would happen.

And we quickly were like, oh, this is a really clever design. This makes sense. It's a win for Medicaid programs. It's a win for the manufacturers and it will improve quality for these beneficiaries in line with our statutory mandate.

And we leaned in pretty hard to try and recruit states to say, you should sign up for this model. And so we had a whole team, the Medicaid team, the innovation center team calling up states, and then where necessary, we had the Medicaid, director, the director of CMCS. And then, I also got involved in calling different states, were helpful to try and get them to agree to come in on the model. We looked at that foundation from the cell and gene therapy model for sickle cell and said, we should grow that out.

And so the generous model, which is about most favored nation for drugs in the Medicaid program, actually takes a similar approach. Manufacturers voluntarily coming in to say, let's give this different pricing to Medicaid, let's agree to standardized formulary placement prior authorization regimens, and then make that price available to states. And states have the opportunity in generous drug by drug to select, is this a drug that I should select under generous or is the rebate structure and the supplemental rebate that I negotiated already, whether on my own or through a collaborative, a better deal for me.

And because it's drug by drug, as there's variation between what states have been able to negotiate, there is a win for states. If your price is lower than the most favored nation price, because supplemental rebates don't count in that calculation, great, just keep that. But if your price is higher, switch to the MFN price. And so that's just a win for state budgets and state Medicaid programs that I hope will improve the sustainability of the Medicaid program and bring most favored nation pricing to our Medicaid programs and the most vulnerable beneficiaries on an insurance program in the country.

Then we have the approach for Medicare where we've laid out two mandatory models, one on Part B and one on Part D to also utilize most favored nation pricing. Those are ones that we have proposals out but have not finalized at this time. And so we are going through the comments and we appreciate the public's engagement with them. Finally, we have the balanced model, which as you noted is a GLP-one coverage model.

And while we did take a step to delay its initiation in Medicare, we are moving forward with its initiation for Medicaid coverage for GLP-one drugs to make an affordable price available to Medicaid programs to the extent that they choose to opt in to that program for one of the transformative advancements in the biopharmaceutical space, that we have at this time. Within the GLP-one space, we are also running a four zero two demonstration, which is a different type of demonstration project, called the bridge demonstration, which will run through the end of twenty seven and give access to Medicare beneficiaries who have qualifying conditions at the $50 price the president negotiated for these products.

All of that is what we've come out with so far, but I truly believe that there is more room that we could build on in the drug space as the innovation center. I think there's a lot more to be done to make drugs more affordable and encourage investment in advancements in these spaces. So the cell and gene therapy model, we should look at expanding it to other indications. There are other areas where there are not aligned incentives to prove out the uses for different things, and we could potentially fix that using the power to reshape our reimbursement flows.

And so I'm really excited about this space. I think there's more to do. We just need to recognize that the signal is going to particular set market segments or particular payers as we look at, you know, Medicare, Medicaid, B versus D, you know, generics, brand name, cell and gene therapies versus other types of advancements. And so we just need to be aware of the market structure and what we're targeting in anything that we come out with.

Say a little more about that awareness. How are you kind of monitoring, or what are you hearing from manufacturers and payers? Are they happy to sit down at the table with you? I mean, you are the behemoth in this space, so I imagine they're not gonna walk away.

But how how are all these different programs kind of being received by folks on that side of the table? In general, we have more positive receptions to voluntary models, than to mandatory models on the part of those who will be mandated into them. So I'll just acknowledge that. I also think mandatory models are an incredibly valuable tool for lowering costs in our healthcare system and ensuring there's not selection bias in how a model is tested.

And so I think it's an important part of our toolkit. There's a reason why we proposed three mandatory models last year, Guard, Globe and our ambulatory specialty model, it's because I think based off the history of the innovation centers test, there's some data suggest that they are more likely to be certifiable than, the voluntary models. And so that's part of our calculus in fulfilling our statutory mandate. But I do think in the drug pricing space, there is room for win wins as the cell and gene therapy model shows where we can align incentives and have voluntary participation as well.

There are many people who have come to us with proposals on biosimilar pricing of, can you do this differently? It really disincentivizes us making investments here. Or, you know, there is a real pipeline of cell and gene therapies that the FDA is reviewing. What does reimbursement look like on the other side?

Is there a smart approach you could design that works for us and works for these programs? And so people are actually coming to us with ideas, which is helpful. And we're asking questions, we're vetting. We also have a number of ideas of our own.

And so that dialogue and input is helpful as we're advancing our thinking. Well, Abe Sutton, thank you so much for taking the time to chat with us here today. I really appreciate it. Thanks for joining us here on A Health Podyssey.

It's a privilege to get to join. If I could just put one note in at the end. Yeah. I would love to encourage any of your listeners who are thinking through the next move in their careers or next move for graduate students of theirs to consider sending resumes our way at the Innovation Center.

We are actively hiring and we are looking for folks with the expertise and knowledge on the drive to reshape our approach to healthcare and make care more affordable. Thank you for having me on. It's a privilege to get to join and speak with you and this audience. Absolutely, and to our listeners, thanks for tuning in.

If you enjoyed this episode, leave a review, recommend it to a friend, and of course tune in next week. Thanks, everyone. Thanks for listening. If you enjoyed today's episode, I hope you'll tell a friend about A Health Podyssey.