How ACO Conveners Are Changing the Medicare Shared Savings Program (MSSP) Behind the Scenes

Health Affairs Publishing’s Rob Lott speaks to Adam Markovitz of the University of Michigan about his recent paper exploring the growing role of third-party firms in Medicare ACOs, highlighting how they have contributed to wider participation and more geographically dispersed networks while raising questions about how these structures relate to shared savings outcomes.

Order the June 2026 issue of Health Affairs.

Sign up for our free Health Affairs newsletters to stay up to date on health policy news and analysis.

Rob Lott: Friends, before we get to this week's paper, a quick

programming note. This will be our last episode before we take

a month long summer sabbatical. We'll be off for all of July,

and we'll be back out there with new episodes, new research, new

brilliant authors beginning again in August. So let me say

thanks for a great first half of twenty twenty six, and here's

wishing you all a happy Independence Day and a happy

birthday, Shuli. But first, let's do this one more time.

Let's talk about accountable care organizations, one of the

biggest bets that national policymakers have placed over

the last two decades. The goal, of course, of ACOs is to slow

cost growth and improve care by incentivizing groups of

clinicians and organizations to take collective responsibility

for total cost of care. Organizations earn bonuses when

average spending for their beneficiaries falls below a

financial benchmark. As our guest today describes, ACOs

were, quote, originally conceived as local clinically

integrated networks of physicians and hospitals that

collaboratively care for shared patients, close quote. But a

funny thing has happened over the last fifteen years with the

Medicare shared savings program.

That's Medicare's flagship ACO model. We started seeing ACOs

made up of clinicians and systems from all over the

country. The thing that connects these disparate providers is not

that they're based in Chicago or Billings, Montana or the Blue

Ridge Mountains. Rather, they are being brought together by

third party firms commonly referred to as conveners. Now

what to make of these conveners?

On one hand, they manage the administrative and technological

demands that might otherwise prevent many practices,

especially small independent practices from participating in

an ACO at all. On the other hand, it's not exactly

consistent with policymakers' original vision and intention

for the program. Now change like this, after all, can be good. It

can be bad. But either way, it tends to make people nervous and

uncertain.

But here we are, nerves and all. So for our podcast today, let's

embrace the uncertainty and see if we can learn a little more

about these conveners and the road ahead. I'm here with doctor

Adam A Markovitz, an assistant professor of general internal

medicine and a practicing general internist at the

University of Michigan. Together with colleagues, he has a new

paper in the June issue of Health Affairs, studying and

describing, quote, third party convener firms and the rise of

geographically dispersed, High Earning Medicare ACOs. This is a

really interesting paper on a relatively undercovered topic,

and I really can't wait to dig in.

Doctor. Adam Markovitz, welcome to A Health Podyssey.

Adam Markovitz: Thank you so much for having me. It's really

a pleasure to be here.

Rob Lott: Well, why don't we start with a little bit of

history? When the ACO model and the Medicare Shared Savings

Program in particular were first developed, Do you have a sense,

did policymakers expect the rise of third party conveners? Did

they foresee this? And if not, why not?

Adam Markovitz: I don't I don't think so. I think when ACOs were

were first being developed, I don't think policymakers really

foresaw the rise of these third party convener firms. And, you

know, I think to think about why I I think you really have to,

like, go back to where this idea originally came from. So ACO is

really traced to Elliot Fisher and his colleagues at Dartmouth,

who first published a paper, honestly, health affairs, I

guess, nearly twenty years ago called Creating Accountable Care

Organizations, the extended hospital staff. And, you know,

at that time, I think the vision was really like a local,

clinically integrated network of physicians in hospitals that

were collaboratively caring for the patients that they actually

share.

So, you know, essentially a virtual delivery system where

they could manage the complex needs of their patients. And,

you know, to put that in context, know, the underlying

problem we're all wrestling with is fee for service. We worry

that it pays for volume, not value. And Medicare's first big

answer to this was Medicare Advantage. Hand a fixed

capitated budget to private insurers, and they can basically

figure out the spending.

And ACOs at the same time are really traditional Medicare's

answer to MA. You know, we still want to move off of pure fee for

service, get away from these problematic incentives for

volume. But unlike Medicare Advantage, we wanted those

decisions to sit with doctors, not private insurers. You know,

I think the rationale, thinking is that doctors are closer to

patients, that they have an ethical obligation to do the

right thing so that they can deliver better care at lower

cost. So, you know, the whole premise of ACOs is in some sense

to replace the insurer with the physician.

But this assumes that ACOs can replace insurers with doctors,

which means that we actually have to require that doctors act

like insurers, optimize networks, navigate complex

rules, bear actuarial risk. And those are functions that like

health payers handle centrally all the time, every day and at

scale. But most practices, especially smaller and

independent ones just aren't built for for them. They don't

have the infrastructure, the experience, the financial

reserves. I I'm a primary care physician.

I'm a health policy researcher who studies this, I wouldn't

have any idea how to build and run an ACO on my own. So instead

of market emerged to do just that. And I think that's that's

really where the ECO conveners come in. So, you know, in

hindsight, it should have been obvious that if CMS creates some

super complicated set of rules for how ACOs are run, how

benchmarks are set, that some third party firms would rush to

fill this void and help providers participate. But no, I

I don't think we really anticipated this.

Certainly, this the issues of selection, were certainly

somewhat foreseeable at the time. People warned about how

benchmarks could be gamed. A lot of thought went into historical

benchmarks versus regional benchmarks. How much should we

risk adjust. But I don't think we really foresaw that there'd

be this whole class of third party firms that would become

the vehicle for, addressing this.

And, you know, I think that the tell here is that fifteen years

in, CMS still doesn't have any published formal records of

these convener relationships.

Rob Lott: Okay. So fast forward fifteen years, as you said. Here

we are. You're about to set out on this study that we're gonna

talk about today. Did you have a theory before you conducted the

research about how the presence of third party conveners in this

ecosystem might affect the ecosystem itself?

Adam Markovitz: You know, honestly, like like most

researchers, I I don't I hadn't given conveners a whole lot of

thought going into this work. I was certainly aware of some of

the big ones, Caravan, Aledade. I'd seen their publicity on

social media. I'd seen some talks given, but I I didn't

really have a clear sense of how much involvement they actually

had in the program or what their ACOs would really look like. If

anything, I sort of vaguely assumed they would be doing the

obvious good thing that they advertised, you know, organizing

small, local, independent practices into ACOs.

And then one day, I was going through the the CMS's public use

files as as one does in their free time as one does. And and I

noticed something. Like, there's this column in the public use

file where it states which states ACOs operate in. And, you

know, while a lot of them were just in one state, two state,

maybe three states, some of them were spread across five, ten, 15

states, sometimes in totally different parts of the country.

And that just didn't make any sense to me.

Like, why would an ACO be organized that way? The whole

idea is local doctors courting care for shared patients. I will

say on the other hand, I'd spent a lot of time during my PhD

dissertation evaluating the MSSP, And one of the things I'd

found and published both in health fairs and annals of

internal medicine was evidence of selection bias, with ACOs

appearing to recruit low cost clinicians, and that a lot of

what looks like savings in the program was really that

selection of low cost doctors. So I think putting together

these two findings, you know, I did start to wonder whether or

not these super geographically dispersed ACOs were, you know,

essentially like a fingerprint for some of that selective

recruitment, but just done on a nationwide scale that I would

not have appreciated otherwise. And then the question is, these

conveners partly the ones enabling it?

Because frankly, a lot of what they're hired to do is to

recruit and aggregate clinicians in the ACOs. So I think that was

definitely a hunch. But, know, I did wanna hold on to both

possibilities because I do think there's, you know, there's the

good story that conveners really do enable participation. They

bring this capital, the analytics, the care management,

to small practices that could never do this on their own. And

then there's the bad story that they're essentially engaging,

sort of in arbitrage, aggregating low cost clinicians

across the country to beat these benchmarks.

And so I would probably say my prior was that it's probably

both. You know? There's probably some good and bad. And honestly,

that's probably how I still still feel about it having

performed this study.

Rob Lott: A a little from column a and a little from column b.

And so let's dig into what you found when you conducted the

research. You looked at national data from 2012 to 2021 and

examined trends in conveners involvement over time. You also

looked at how geographically dispersed they were compared to

non convener ACOs, and then you looked at those associations

with shared savings. So tell us what you found.

Adam Markovitz: Yeah. So using those national data, I I think

there were really three central findings that we had, you know,

bearing in mind this is from 2012 through 2021 in the MSSP.

So first, these conveners grew from 11 to at least 23% of ACO

beneficiaries in the program. Second, these networks became

gradually more geographically dispersed over time,

particularly among these convener run ACOs. So among the

convener run ACOs, these dispersed ACOs went from nine to

45% of beneficiaries versus from one to 8% among nonconvener

ACOs.

And meanwhile, these, you know, local sort of single community

ACOs nearly vanished going from 20% down to 4% by 2021. And then

the third finding was was about the money, about the bonuses. So

when you look at the shared savings bonuses, it really was

the dispersed convener ECOs that earned the most by by really a

wide margin, earning about a $171 per beneficiary per year,

while the local convener ACOs earned the least around $95 And

I think the key contrast though is that among the non convener

ACOs, that gap between the dispersed and the locally, the

local ACOs basically disappeared and was not significant. So

really, it it was it wasn't just a convener effect. There is

something specific to the dispersed convener ECOs that

seem particularly good for earning bonuses.

Rob Lott: So there are conveners that are operating more locally

less less dispersed. Is that right? And those struggled as

well. Or or perhaps let let's not say struggled. They didn't

thrive quite as much as the more dispersed ones.

Is that fair?

Adam Markovitz: Correct. Yes. Exactly.

Rob Lott: I wanna ask a little more about the details and

perhaps some of our the possible mechanisms that might be driving

these differences. But first, let's take a quick break. And

we're back. I'm here with doctor Adam Markovitz, talking about

the rise of third party convener firms and, geographically

dispersed high earning Medicare ACOs. So you just told us about

some of the findings and the fact that those most

geographically dispersed third party convened ACOs were doing

really well in terms of earning shared savings.

And I guess one way to look at that is that, you know, they've

found the trick and the other way is to say that they're

really good at what they do. And so I'm wondering, you know, how

you think about the mechanisms that might be explaining why

dispersed convener affiliated ACOs earned the highest shared

savings.

Adam Markovitz: Yeah. It's a great question. You know, and I

wanna be careful here. So this is descriptive study. It's not a

causal one.

And second, the mechanisms I'm about to describe are not

mutually exclusive and some elements of each of them could

be true. So I think the first mechanism is the good story and

it's the story the conveners themselves tell, that they

genuinely improve care, better analytics, better care

management, better quality of care, and that's how they lower

spending. And to at least get at that, we looked at whether these

dispersed conveners actually had the highest quality performance

in the MSSP, but we didn't see that. So the dispersed convener

ACOs scored no better on overall quality metrics than the local

ECOs or the non convener ACOs. So they may have been delivering

better care, but didn't show up in sort of those aggregate

quality measures.

The second is actually reverse causality. So it's just simply

basically the best ACOs that succeed and then they expand

over time. They grow into new markets and then they look

dispersed. And in that story, dispersion is really a marker of

success and not a cause of it. And because ours was a cross

sectional study, that certainly is a real possibility that that

may be driving some of our findings.

And then the third story I think is is the bad one of strategic

selection. So that's where conveners are deliberately

constructing networks to beat spending benchmarks, cherry

picking low cost clinicians from across the country, regardless

of whether or not those clinicians share any patients or

are actually doing anything to subsequently reduce spending.

And, you know, while the convener firms are not

advertising this, there is a huge cottage industry of ACO

consultants, ACO actuaries whose entire job it is to do exactly

this, and they are not subtle about it. So I think the

clearest example is a tool called, Milliman's ACO Builder,

which is essentially a tool for cherry picking low cost doctors.

So they went out, bought Medicare claims data on every

physician nationwide, mapped out who's high cost and who's low

cost, and then they sell you that map so that you can just

grab the low cost ones and build your network, to be, like as

they send their website, structured for maximum

performance and shared savings.

So to test whether or not that was happening, we basically

built our own ACO builder, but in reverse. So we took each

ACO's patients and asked a simple question. How does their

spending compared to other patients in the same county?

Because that's the regional benchmark. And if you've

recruited low cost clinicians, your patients should look

unusually cheap relative to their local peers.

And that's exactly what we found. So the dispersed convener

ACOs had the lowest spending relative to the region of any

group, roughly $400 below the regional benchmark. While the

dispersed non convener ACOs had the highest, only about $100

below. And again, this is cross sectional, so this could also

reflect subsequent savings that the ACOs did, generate. But, you

know, one thing to note is, like, the firm selling tools

like ACO Builder, they're not even in our study.

So we could only identify sort of the formal conveners who

actually run ACOs and list themselves as, the ECO executive

or the public contact for the ACO in the public files. But

there's this entire layer of analytic consultants that's

essentially invisible to us researchers and I think to CMS.

And these consultants are not coy about what they're doing. So

in that sense, I think our paper is to some extent the tip of the

iceberg. But once you start looking for it, it's really

everywhere.

And I think most of us, so I will say as an academic

researcher, I speak with policymakers, I think most

people still picture ECOs as like these local ECOs who are

not sophisticated enough to cherry pick or gain benchmarks

like this. And honestly, I don't think a lot of them are, but you

don't you don't have to be. You don't need your hospital's head

of pop health to know how to do this. You just need them to know

someone who does and hire them. And I do think that's exactly

what they're doing.

Like, you spend ten minutes on LinkedIn, as I do now, it's wall

to wall consultants, actuaries advertising the software,

hosting webinars on you know, there's a new ACO lead program.

And, like, every day, people are posting about, you know, how you

should structure your ACO for maximal savings. So I think

there is this whole industry that is hiding in plain sight.

It's just not the mental model that we as researchers have been

bringing to these programs.

Rob Lott: Remind me to connect with you on LinkedIn after this

conversation.

Adam Markovitz: Would be a So

Rob Lott: your findings highlight this tension, right?

That's what we're talking about here today between sort of the

benefit of increased participation made possible by

third party conveners versus maybe this potential undermining

of the program when bonuses are achieved through network

optimization in theory as opposed to actual coordination

and quality improvements. And I'm curious, you alluded to some

of the entities in this ecosystem being invisible to

CMS. Do you have a sense of to what extent all of this is on

CMS's radar and how the agency has navigated this tension as

it's made changes to various ACO models over the years?

Adam Markovitz: Yeah. And I think that's exactly the right

way to frame it because it it is a real tension. Know? So the

same firms that do lower the barrier to entry for, like,

these small practices and make it easier to really, you know,

improve quality and lower spending, they're also the same

firms that make it easiest to win without doing anything. I do

think this is definitely on CMS' radar.

You can, I think, read it right off of some of the recent ACO

design choices? So, like, for instance, just thinking about

ACO REACH, so that's the the ACO model that LEED is now replacing

in this coming year. So REACH really was CMS' push, sort of

like in the convener sense, it was their push to try to reach

these small and rural practices that MSSP had left behind. It

even opened the door for private firms to directly contract with

CMS for the first time in traditional Medicare. There's

not been much empirical work on REACH yet, but our own

preliminary data show that REACH ACOs are doing some of the same

things we found in MSSP, selecting these low cost

clinicians.

And with the lead, you read some of the design choices, I think,

as a direct response to some of those concerns. So I, again, in

my free time as one does, I've read through large portions of

the RFA, And now they actually ask they have a a question about

who runs your ACO, and there is a checkbox for conveners. I've

not looked through MSSP's RFA, so maybe they did did before,

but to me that was noteworthy. And in terms of how LEED is

designed to curb selection, they are now starting ACOs on a pure

historical benchmark rather than a regional one. So you're

measured against your own past instead of being rewarded simply

by finding cheap clinicians relative to their local peers.

And also like in MSSP but not in REACH, they require the whole

practice to participate. So unlike in REACH, you can't sort

of cherry pick individual low cost doctors. And then they

actually, in their RFA, explicitly ban conveners from

tin swapping, which means where you basically shuffle the same

practice between your ACO. So they say that if a practice is

in an ACO run by a convener and they leave that ACO, they can't

join another convener, another ACO run by that same convener

for three years. So clearly, they are seeing something in

their data in REACH or in MSP that they're concerned about.

And they have all this stuff about curbing risk coding. They

want to use AI infer Abe Sedin talks about AI inferred risk

scores, which is basically built on objective signals like

prescription fills rather than the diagnosis codes entered by

by the doctor. Mhmm. You know, at the same time, I think

there's a bigger point, and it is why I'm skeptical that any of

this will fully or even adequately solve the problem.

You know, some of these are are genuinely, I think, really good

ideas.

But the real big tension here, it's it's not one loophole. This

is a all of these are voluntary programs, and they're all built

on benchmarks and any benchmark can be gamed. And the voluntary

part of that makes the benchmarks even the most

problematic because to get people to join, you have to make

it attractive. And people only join if they think they can make

money. You can never really tighten down the rules all the

way without losing the participation that the program

depends on.

And then the final vicious loop is that every single rule that

you add to try to stop gaming makes the program more complex.

And the more complex the program gets, the more the doctors and

the hospitals need a third party firm to navigate it for them. If

you look at who runs these convener firms, many of them

used to work for CMS. They used to work for CMMI. They are

specifically the people who know the rules the best.

And once you're out of government, of course, you would

go run a consulting firm about the program that you helped

create. And, of course, if you are a small practice struggling

to make ends meet in fee for service and you wanna make money

from value based payment, of course, you would contract with

a consultant. So to some extent, this is everyone just acting

rationally. But I think it's ended us in a really difficult

spot. You know?

So I think that's that's sort of where we've where we've left

off.

Rob Lott: Well, a difficult spot, but perhaps a glimmer of

hope that we might be able to continue learning about these

conveners and continue studying them with researchers like

Doctor. Adam Markovitz and his colleagues who've really done

interesting, important work with this paper. I encourage our

listeners to check it out in the June issue of Health Affairs.

Doctor Adam Markovitz, thanks so much for taking the time to chat

with us today. Really enjoyed it.

Adam Markovitz: Thank you. It's really been a pleasure.

Rob Lott: And to our listeners, thanks for tuning in. If you're

a rational actor, you might recommend this podcast to a

friend, leave a review, and, of course, tune in next week.

Thanks, everyone.

Adam Markovitz: Thanks for listening. If you enjoyed

today's episode, I hope you'll tell a friend about A Health

Podyssey.