By KIP SULLIVAN, JD
There is no meaningful difference between the performance of Medicare ACOs that accept only upside risk (the chance to make money) and ACOs that accept both up- and downside risk (the risk of losing money). But CMS’s administrator, Seema Verma, thinks otherwise. According to her, one-sided ACOs are raising Medicare’s costs while two-sided ACOs are saving “significant” amounts of money. She is so sure of this that she is altering the rules of the Medicare Shared Savings Program (MSSP). Currently only 18 percent of MSSP ACOs accept two-sided risk. That will change next year. According to a proposed rule CMS published on August 9, ACOs will have at most two years to participate in the MSSP exposed to upside risk only, and after that they must accept two-sided risk.
That same day, Verma published an essay on the Health Affairs blog in which she revealed, presumably unwittingly, how little evidence she has to support her decision. The data Verma published in that essay revealed that one-sided ACOs are raising Medicare’s costs by six-one-hundredths of a percent while two-sided ACOs are cutting Medicare’s costs by seven-tenths of a percent.  Because these figures do not consider the expenses ACOs incur, and because the algorithms CMS uses to assign patients to ACOs and to calculate ACO expenditure targets and actual performance are so complex, this microscopic difference is meaningless.
“Two beellion dawlers”
Even if the difference is not meaningless – even if two-sided ACOs actually save a few tenths of a percent for Medicare – the impact on Medicare spending will be barely noticeable. Verma assures us, without a hint of embarrassment, that her new rule will cut Medicare spending by $2.2 billion over ten years. “The projected impact of the proposal would be savings to Medicare of $2.2 billion over ten years,” she declares in her blog comment.
I feel like we’re in a scene from the Austin Powers movie where Dr. Evil announces he will hold the world ransom for “one meellion dawlers.” Dr. Evil’s sidekick, Number Two, has to advise him that a million dollars is peanuts. Verma’s estimate of 2.2 “beellion dawlers” is essentially zero percent of the trillions of dollars CMS will spend on Medicare in the next decade.
As pathetic as these figures are, they fail to take into account ACO start-up and operating costs. CMS doesn’t know or care what those costs are. The only relevant information we have are some undocumented statements by the staff of the Medicare Payment Advisory Commission (MedPAC) to the effect that ACO overhead is about 2 percent of their benchmarks (their predicted spending).  I suspect 2 percent is low, but let’s take it at face value and do the math. If, as Verma’s data indicates, two-sided ACOs save Medicare seven-tenths of a percent net (that is, considering both CMS’s shared-savings payments to some ACOs and penalties other ACOs that lose money pay to CMS), but these ACOs spend 2 percent doing whatever it is ACOs do, that means the average two-sided ACO is losing one percent.
Is it any wonder Clif Gaus, director of the ACO trade group NAACOS, said Verma’s proposed rule “defies logic” and will “pull the rug out from under” the MSSP? In a press release, Gaus cited a survey indicating 70 percent of ACOs will bail out of the MSSP program if they are forced into two-sided risk contracts in 2019. That is much higher than the 20 percent dropout rate predicted by CMS.
The good news is that Verma may have hastened the demise of a program that isn’t working. Whether Congress ultimately pulls the plug on the ACO project will depend on whether ACO advocates will concede at some point that the ACO fad was based on faith, not evidence, and has failed to work. I predict they will refuse to admit failure and will instead peddle another equally ineffective solution, for example, overpaying ACOs (as the Medicare Advantage insurers and their predecessors have been for the last half-century). I base my prediction on the behavior of ACO advocates. The history of the ACO movement indicates ACO proponents don’t make decisions based on evidence. Let me review that history for you.
Levitating a Lead Balloon
From the moment the ACO concept was born, evidence has not mattered to ACO proponents. The ACO concept was conceived during a conversation between Elliott Fisher and members of MedPAC at their November 9, 2006 meeting. Fisher presented a flabby definition of the ACO to the commission, he failed to describe a single mechanism by which it would cut costs or improve care, and he offered no evidence that this vaguely defined thing would work. And MedPAC swallowed it.
MedPAC has been an enthusiastic proponent of ACOs ever since. The failure of CMS’s pilot test of the ACO notion, the Physician Group Practice (PGP) Demonstration (it ran from 2005 to 2010), did not diminish MedPAC’s faith in the ACO. The final evaluation of the PGP demo, published in 2012, found that the ten PGPs (which were exposed to a version of two-sided risk) saved Medicare a grand total of three-tenths of a percent over the 2005-2010 period. 
MedPAC avoided discussing the possibility that its cherished ACO might be a lead balloon until the fall of 2016, that is to say, ten years after they concocted the ACO, seven years after congressional Democrats wrote ACOs into the legislation that would become the Affordable Care Act, and four years after CMS initiated the Pioneer and MSSP ACO programs authorized by the Affordable Care Act. At its October 6, 2016 meeting, MedPAC staffer David Glass reported that the Pioneer program (which ran from 2012 to 2016) and the MSSP had together raised Medicare’s costs in 2015 by three-tenths of a percent net (see pp. 6-7 of the transcript of the morning session.) 
The commission was disturbed by this report. Several commissioners asked the staff why some ACOs were saving money and others weren’t. The staff expressed doubt that they would ever be able to answer that question. The most helpful comment the staff could come up with was this one by Jeff Stensland in response to a question from commission member Jack Hoadley about why some ACOs dropped out of the Pioneer and MSSP programs. “If you look who drops out, it’s they didn’t make money. If you made money, even if you don’t know why, you stay in,” said Stensland (p. 41 of the transcript in the morning session of the October 6, 2016 meeting). Should we be surprised by the uselessness of that tautological answer? If Fisher and MedPAC never defined the mechanisms by which ACOs were supposed to work their magic, how are the poor staff supposed to explain why some ACOs fail and others succeed?
But over the next year, MedPAC staff came up with what they deemed to be a satisfactory answer – exposing ACOs to more risk will make them save money. Why? Who knows? They didn’t say. Commissioner David Nerenz at one point challenged the staff to explain the mechanism by which two-sided risk might cause ACOs to improve their performance, but they did not respond. At the January 12, 2018 meeting, staff presented slides that showed miniscule differences between one- and two-sided-risk ACOs (see slides 9 and 12 here) and they implied there was some causal connection. Elliott Fisher reinforced this new folklore in an article in the New England Journal of Medicine published in November 2017. “[P]roviders know how to save money,” he and his co-authors intoned, “but they need financial motivation to change their behavior.” Ergo, CMS should switch all ACOs to two-sided risk. As is the norm among ACO proponents, Fisher et al. provided no evidence for this claim.
Despite the tiny differences in the performance of the two types of ACOs, and despite the complete absence of evidence on the mechanism that might have caused these tiny differences, MedPAC bought the advice of their staff, Fisher and other ACO proponents, and decided to celebrate two-sided ACOs in their next report to Congress. “[T]wo-sided ACO models appear to save more than one-sided ACO models,” said the commission in its June 2018 report to Congress (p. 211). They went on to suggest that spending more money on two-sided ACOs would induce more ACOs to accept two-sided risk, and this in turn might save money despite the higher subsidies to ACOs.
By the spring of 2018, the evidence-free claim that exposure to greater risk would improve the performance of MSSP ACOs was hardening into conventional wisdom. That spring Trump administration officials made announcements indicating they too were accepting the new folklore. CMS’s proposed new rule, posted on August 9, and accompanying statements by Verma, made it official. Here is how CMS articulated the new folklore in their proposed rule:
ACOs in two-sided models have shown significant savings to [sic] the Medicare program…. Even more concerning is the finding that one-sided model ACOs, which are not accountable for sharing in losses, have actually increased Medicare spending relative to their benchmarks. Further, the presence of an “upside-only” track may be encouraging consolidation in the marketplace, reducing competition…. [pp. 6-7]
The only accurate statement in that paragraph is the statement that one-sided ACOs have increased Medicare spending. The claim that one-sided ACOs may be encouraging consolidation is accurate, but CMS’s failure to say the same about two-sided ACOs is irrational.
Facing the Evidence
Evidence that the ACO project is failing is piling up. All three of CMS’s two-sided ACO programs – the PGP demo, the Pioneer demo, and the Next Generation program – saved only a few tenths of a percent, while CMS’s mostly two-sided program, the MSSP, raised costs by a smidgeon. All four programs have raised costs if we take into account the ACOs’ start-up and operating costs and CMS’s cost of administering these complex programs.  Evidence indicting the other major “value-based payment” fads – medical homes, bundled payments , and pay-for-performance schemes– is also piling up. The simultaneous failure of all these fads to cut costs spells trouble ahead for the Affordable Care Act (because it relies on “value-based payment reforms” for cost containment), MACRA (because it also relies on “value-based payment” theology), and our entire health care system (because the big insurance companies and the major hospital-clinic chains are spending more money on “value-based payment” fads than those fads are saving, and because these 1,000-pound gorillas are using the establishment’s endorsement of ACOs, medical homes etc. as an excuse to become 2,000-pound gorillas).
The root cause of our nation’s chronic inability to adopt effective cost-containment policies is the chronic inability of the American health policy establishment to make decisions based on evidence, not groupthink. Seema Verma’s decision to bet the farm on two-sided-risk ACOs is the latest example of this problem.
 I derived these percentages from the dollar figures Verma presented in two exhibits in her comment on the Health Affairs blog. Exhibit 1 indicates “all upside-only ACOs” raised Medicare spending by a net of $0.049 billion in 2016 compared with $76.718 billion in expected (benchmark) spending for those ACOs in 2016. The net increase, therefore, amounts to 6 one-hundredths of 1 percent. Similarly, the data Verma presents in Exhibit 4 indicates “two-sided ACOs” saved $0.033 billion in 2016 compared with expected spending of $4.659 billion, or seven-tenths of a percent.
Readers can find nearly identical data in Chapter 8 of MedPAC’s June 2018 report to Congress. Unlike Ms. Verma’s data in her Health Affairs essay, which is broken out by one-sided versus two-sided ACOs, the MedPAC figures are broken out by program – Pioneer, MSSP, and the recently started Next Generation. MedPAC’s tables show that the MSSP program raised costs by one-to-three-tenths of a percent over the period 2013-2016, while the small two-sided ACO programs performed almost as miserably. The two-sided Pioneer program averaged savings of only a few tenths of a percent over the 2012-2016 period, and its successor, the two-sided NextGen program, achieved a savings of two-tenths of a percent in its first year (2016). The table below summarizes these results.
Table 1: Net savings/losses as a percent of total Medicare spending by ACOs according to CMS
 MedPAC staff have testified at several MedPAC meetings that ACO overhead is 1 to 2 percent of ACO Medicare revenues. MedPAC stated in its June 2018 report to Congress that the per enrollee overhead of ACOs is about $200 (p. 236). That equals a little less than 2 percent of the annual cost of insuring a Medicare beneficiary.
 The gross savings and the cost to CMS of the PGP demo are reported at page 64 of the final evaluation.
The physician group practices that participated in the Physician Group Practice demo were exposed to a form of two-sided risk that was slightly different from the version used in the Pioneer ACO demonstration and which is being proposed by Verma for the MSSP. In the PGP demo, the PGPs were not penalized in any given year if they went over their target (their benchmark), but if they saved money in the following year the amount of their shared savings would be reduced by whatever amount they went over budget the previous year.
 See footnote 1 of my comment here for an explanation of how I derived the three-tenths-of-a-percent figure based on the dollar figures Glass reported.
 There are some who argue that three or four articles in the peer-reviewed literature demonstrate that ACOs are saving a percent or two more than CMS’s data shows. MedPAC is among them; Verma obviously is not. MedPAC summarized the evidence for this argument in Chapter 8 of their June 2018 report to Congress. Their argument was not convincing. Only two studies MedPAC cited reported net savings (both were studies for which Michael McWilliams was the author or one of the authors). Those studies reported a net savings of a mere seven-tenths of a percent for the MSSP and three-tenths of a percent for the Pioneer program.
 The exception to the statement that bundled payments are failing appears to be bundled payment for joint and hip replacement. But the reason appears to be reduced prices of artificial joints achieved by the exercise of excessive market power by hospitals, not reduced utilization of services. If we want to reduce prices of medical goods and services, we should do it uniformly via negotiations between CMS and providers, and not let the race to gigantism determine who benefits from bundled payment.
Kip Sullivan chairs the Policy Advisory Committee of Health Care For All MN, and is a member of the Minnesota chapter of Physicians for a National Health Program.