Accountable Care Regulations: “Shared Savings” Means Return to Risk

Today’s Managing Health Care Costs Indicator is 429

The Affordable Care Act (ACA) specifies that Medicare will contract with accountable care organizations (ACOs) – groups of primary care and specialty physicians and hospitals that voluntarily coalesce and agree to take financial and clinical responsibility for all care of a population.  The ACA also states that these groups will be paid fee for service, but be able to “share savings” to the extent they are able to deliver care to Medicare beneficiaries for a lower price than expected. I’ll look at these proposed regulations through the lens of behavioral economics (see chart at the bottom of this post).

The regulations are 429 pages long.

“Shared savings” is a conundrum.  It’s hard to get providers to agree to “symmetrical risk,” where they would gain a profit if they deliver care below budget, but they would lose income if they spent more than the budget on a population of patients.  Frankly, we all really hate the potential for loss.  The “stick” of downside risk much more effectively motivates us to act differently because we so hate the possibility of losing. Therefore, downside risk is much more likely to fundamentally alter medical practice and lower resource cost.

In fact, if Medicare merely “shared savings,” this would likely increase costs because by randomness alone some groups would have apparent savings, but those groups with costs above budget due to randomness would be held harmless. Therefore, bonuses would be paid, and would not be offset by penalties.   As noted above, providers who are at risk for losing money are likely to be far more motivated to implement efficiencies in care delivery.

The regulations announced on Friday  use shared savings as a bridge to providers accepting some ‘downside’ risk as well as the potential for upside reward.  The draft regulations envision two shared savings model.

Model One:
Providers would obtain up to 60% of any savings beyond 2% of budget, and would face the potential downside risk if their actual costs exceed the projected expenses by 2%. Providers would have to provide proof they could repay up to 1% of total costs – which I think is the maximum potential provider exposure to loss, although I’m not certain.

Model Two:
Providers would obtain only 50% of savings beyond 2-3.9% of budget depending on size, and would transition to upside and downside risk as of year three. 

Contracts for ACOs require a minimum of 5000 Medicare beneficiaries, and the risk corridor gets smaller as membership increases.   CMS will require ACOs to report on 65 quality measures (domains are patient experience, care coordination, patient safety, preventive health, health of high risk populations).   Shared savings will be predicated on adequate quality performance, and might scale upward with better quality metrics.  Use of electronic records would be mandatory for at least half of the physicians. 

Some quality advocates will be disappointed that the proposed quality measures do not include outcome measures, such as mortality or complications.  However, outcome measures require sophisticated risk adjustment, and often require very large volumes to be statistically reliable. Many providers feel they have less control over outcomes than over processes.  Further, if providers effectively implement processes shown to improve outcomes, the better outcomes are likely to follow.  One problem with Pay for Perfomance was that with only a small number of goals providers were able to “teach to the test” and create workarounds rather than actually improving overall quality. The sheer number of measures makes this less likely in the CMS ACO proposed regulations.

I think that the ACO regulations have dodged a bullet in the initial legislation, which specified that savings would be shared. Instituting a corridor to account for randomness and requiring all groups to have downside risk by year three in exchange for participation will help be sure that groups are not getting a windfall just for being lucky, and will minimize taxpayer cost. Of course, elements that make this a better deal for taxpayers might decrease ACO uptake in the provider community.  I expect an avalanche of provider comment opposing downside risk.

One troublesome area where the ACO regulations stuck close to the legislative script is regarding patient attribution - which ACO will bear responsibility for each patient.  Patients will be retrospectively assigned to an ACO based on having the majority of their primary care at ACO-participating PCPs, who will not be able to join multiple ACOs.  This means that there is no limit to the free choice that traditional Medicare has offered, which is just what patient advocates want. 

However, provider groups won’t be certain for which patients they actually bear responsibility.  This will lower physician confidence that their own performance will have a large impact on the ultimate costs, which could also dampen provider enthusiasm.   Retrospective assignment is also bad news for ACOs that hoped to do aggressive management of the sickest of their ACO members.  They won’t be certain these members are their responsibility until after the end of the year!

My analysis of these critical ACO decisions through a behavioral economics lens:

Impact on Provider Acceptability of Proposed Regs
Likelihood Providers will Improve Their Practices

Require providers to take downside risk by year three
Supercharges provider motivation to change, but gives providers some time to develop infrastructure.  Providers really hate downside risk, though
Risk corridor so that neither first dollar losses or gains are transmitted to the provider
Shields providers from financial losses due to randomness alone, while preventing most unearned windfalls. 
Maximum upside for providers is specified
Decreases the potential for a big win;  this is probably a good idea, but could dampen provider enthusiasm for ACO
Upside is dependent on meeting quality goals
Might be necessary to be sure providers don’t provide too little care, or their quality.  Providers will complain of the expense of reporting on so many measures. Decreases certainty of “winning,” so could dampen provider enthusiasm for ACO.
Process – not outcome quality goals
Providers are likely to feel more in control of whether they achieve incentive, and are thus more likely to improve their processes
Retrospective attribution
Providers will feel like result is less in their control

Three other references of note:

·        WSJ had a nice piece on Atrius Health, the nonprofit parent of Harvard Vanguard, and its ACO efforts 

·        Don Berwick’s introduction of the ACO rules is in the NEJM 

·        Ezra Klein published his rules on addressing health care costs yesterday   I highly recommend them.

 By the way, if you're interested in an ongoing set of links to articles that have interested me (not all of which I blog about), go to this Tumblr site.