Getting More Savings from ACOs: Can the Pace be Pushed?

The Medicare Shared Savings Program (MSSP) has produced modest savings for Medicare and generally been popular among providers, with 561 participating ACOs covering nearly a third of the fee-for-service population in 2018.¹ Despite its early and growing promise, the pace of savings has not been commensurate with the fiscal challenges faced by Medicare, prompting calls for program reform. In December 2018, the Centers for Medicare & Medicaid Services (CMS) finalized “Pathways to Success,” an overhaul of the MSSP that intends to push the pace.

The Challenge of Accelerating Savings in a Voluntary Program

In general, strengthening incentives for ACOs to lower spending is challenging because the program is voluntary. Some approaches—e.g., requiring ACOs to bear “downside” risk for excess spending—may make the program unattractive to some providers, potentially lowering program-wide savings by reducing participation even if enhancing savings among participants. Participation by providers with high spending not explained by patient risk factors is particularly critical to MSSP success because the care their patients receive (including from other providers) presents the greatest opportunities for savings. So far, ACOs with high risk-adjusted spending have indeed lowered spending more than ACOs with low spending.¹

Pushing the pace in the voluntary MSSP is additionally challenging because CMS must specify rules in advance for setting realistic spending targets (or “benchmarks”) for hundreds of organizations, whereas commercial insurers can more flexibly individualize and revise providers’ spending targets through negotiation. Even if CMS sets a pace of benchmark growth that is achievable for ACOs on average, many will fail to meet it and may be discouraged from participating further because, ultimately, ACOs must earn shared-savings bonuses to recoup the costs of their efforts. Lagging ACOs that face downside risk for spending above benchmarks would exit at an expectedly higher rate. These participation considerations would argue for permitting a slow pace.

A faster pace can be set in the MSSP if the alternative is made less attractive. Under the current principal alternative to the MSSP and other new payment models, provider groups miss out on a bonus equal to 5% of professional fees through 2024, receive 0.5% slower annual physician fee updates starting in 2026, and are exposed to the Merit-based Incentive Payment System (MIPS). Moreover, they miss out on the opportunity, at least in the long run, to earn profits from generating efficiencies even as growth in fees (or benchmarks) is slowed. Participation in the MSSP also may help providers gain experience with payment models that may expand in commercial markets or evolve in the future. On the other hand, the MIPS imposes a much lighter quality reporting burden on provider groups than the MSSP, and performance-based incentives in the MIPS have been watered-down thus far² in response to a widespread backlash calling for its repeal. In addition, because the rules governing ACO models have changed so frequently, without a clear direction emerging yet, providers cannot be faulted for weighing short-term losses as ACOs more heavily than the potential long-term gains.

Although the MSSP has grown with new entrants, dropout has been substantial, suggesting that the alternative may not be sufficiently unattractive and the pace set by CMS may already be too fast. In addition to pegging annual benchmark updates to national Medicare spending growth for much of the MSSP’s history, CMS has also “rebased” benchmarks every three years to ACOs’ most recent spending levels, thereby requiring ACOs to achieve progressively lower rates of spending growth to increase their annual savings bonuses. By 2018, 44.5% of ACOs entering the MSSP in 2012–2014 had exited the program. Of those dropping out, only 15.2% had received a shared-savings bonus more than once (vs. 46.3% of ACOs remaining in the program).

The departures have been costly to Medicare. For example, we compared previously published estimates of ACO savings in 2015¹ between ACOs that dropped out by 2018 and those that remained in the MSSP. Although spending reductions achieved by ACOs that eventually dropped out were 20% smaller, they were significant, not offset by shared-savings bonuses, and consequently accounted for much of the net savings generated by the MSSP in 2015.³ This ostensibly surprising result stems from the fact that ACO benchmarks do not represent spending that would be expected in the absence of exposure to the MSSP.³ In fact, benchmarks need not reflect expected spending to achieve their primary purpose—establishing incentives for ACOs to save.

“Pathways to Success”: Potential Participation Effects and Implications

At the heart of the complex “Pathways to Success” are three fundamental shifts. First, ACOs are required to assume downside risk much sooner. Previous rules permitted ACOs to stay in “one-sided” contracts without downside risk for six years, and the vast majority of ACOs opted for this arrangement. Second, “Pathways” will adjust ACO benchmarks for regional spending immediately for all ACOs and give increasing weight to regional spending over time according to a pre-specified progression. Previously, an ACO’s benchmark was based on its own historical spending for the first three years of participation before a progressive blend with regional spending began in year four. The accelerated regional blending will favor providers with low spending for their region and put those with high spending for their region at an immediate disadvantage. Third, CMS promises to get tough on low performers by terminating contracts with ACOs with multiple years of poor financial performance. Of particular concern are providers that may be abusing regulatory relief granted by the MSSP to engage in anticompetitive practices that increase spending.⁴

In isolation, each change may appear to be a conceptually sound refinement that strengthens incentives for ACOs to lower spending. Yet the combination of accelerated downside risk and regionally blended benchmarks means that providers with high spending for their region will face a high risk of potentially substantial losses with little prospect of earning shared savings in the near term. Thus, they may find the alternative more attractive than they have so far. ACOs with low spending for their region, on the other hand, will find the reformed MSSP very attractive, as their participation will effectively be subsidized. If ACOs with high risk-adjusted spending for their region exit and providers with low spending remain and enter, MSSP performance based on the benchmarks would appear improved, but the actual effect would be to increase Medicare spending.

There is already early evidence suggestive of such selective participation in response to the 2016 announcement that ACO benchmarks would be blended with regional spending beginning in the fourth year of participation. Among ACOs entering the MSSP in 2014 and 2015—the first to be affected by the change in 2017 and 2018, respectively—the announcement was followed by a 2017 increase in exit by ACOs with high spending for their region (Table 1). After these departures, CMS reported that 81% of 2014 entrants remaining in the MSSP in 2017 had risk-adjusted spending below their region and thus received upward (beneficial) adjustments to their benchmarks.³ An even stronger pattern of selective participation should be expected from coupling downside risk with regional benchmark adjustments earlier under “Pathways.”

Table 1.

ACO drop out and spending relative to regional average among ACOs entering the MSSP in 2014 or 2015

	Performance year
	2016 (Blending of benchmarks with regional average spending announced)	2017 (Regional blending implemented for 2014 entry cohort)	2018 (Regional blending implemented for 2015 entry cohort)
Cumulative drop-out rate among ACOs participating in 2015, %	8.0	22.9	29.4
ACOs dropping out
No.	16	30	13
Initial spending deviation from regional average in 2014–2015, $/beneficiary*	−10	1,137†	358
ACOs continuing in the program
No.	185	155	142
Initial spending deviation from regional average in 2014–2015, $/beneficiary*	428†	291	284

^*ACOs’ initial spending deviations from their regional average in 2014–2015 are presented as a measure of where ACOs stood at the outset of their contracts relative to their region. The evolution of these initial spending deviations over performance years 2016–2018 then reflects only participation effects; that is, ACO effects on spending during performance years do not affect these initial spending deviations. To estimate the initial spending deviations, we fit a model for total Medicare spending in the first two performance years (2014–2015 for the 2014 entry cohort and 2015–2016 for the 2015 entry cohort) as a function of ACO fixed effects, county fixed effects, and year fixed effects, using Medicare claims for a random 20% sample of beneficiaries. ACO assignment was determined from the CMS beneficiary-level ACO research identifiable files, and the sample of beneficiaries not assigned to an ACO in a given year was limited to those with at least one qualifying service for ACO assignment. Beneficiaries assigned to ACOs that entered in 2012 or 2013 were excluded from the analysis.

^†Spending deviation statistically different from zero at P≤0.05. The statistical significance of drop-out and participation rates was not tested because the policy change encouraged continuation participation of ACOs with low spending and discouraged participation of ACOs with high spending, with ambiguous implications for net participation effects. After adjustment for age, sex, Medicaid enrollment, disability, end-stage renal disease, and hierarchical condition categories risk score, the spending deviation for ACOs dropping out in 2017 was -$685 (95% CI: -$19, $1388).

The provision to terminate agreements with poor-performing ACOs will compound the selection problem by hastening exit of ACOs with high spending for their region. While rooting out providers that abuse the program or fail to respond might be a worthy goal, such ACOs cannot be identified based on their financial performance because the benchmarks do not represent ACOs’ expected spending under the status quo.³ In general, it is very difficult to quantify the response of a single ACO to program incentives — akin to divining the treatment response for a single patient in a clinical trial. The distinction between benchmarks and expected spending will grow more obvious as benchmarks increasingly incorporate a regional component; one would not expect ACOs’ spending to converge to the regional average under the status quo. As benchmarks become increasingly disconnected from expected spending, performance-based terminations would oust increasingly more ACOs that are actually generating savings.

Relying on Incentives to Encourage Savings and Participation

It remains to be seen whether the MSSP unravels into a large subsidy for providers with already low spending, but the evidence to date raises important concerns about the strategies adopted by “Pathways” to speed savings by specifying an acceptable pace and holding ACOs to it. An important principle for improving the MSSP is to build on, rather than jeopardize, its early success. Instead of setting a faster pace for ACOs with high spending, CMS could focus on setting stronger incentives to elicit a faster, albeit uncertain and varied, pace in ways that would not compromise participation. In particular, benchmarks could be based on ACOs’ historical spending and grown at a desirable rate, as in the original rules, but without the periodic “rebasing” to newly achieved levels that has left ACOs with little incentive to lower spending thus far.⁵ Without rebasing or aggressive regionalization of benchmarks, downside risk could be incorporated gradually. If sufficient convergence in spending can be fostered among ACOs in the same area, regional benchmarks could then be considered. In addition, ACOs with low spending for their region could be given higher shared-savings rates to support their more challenging task of cutting waste when there is less to cut, instead of subsidies in the form of regionally adjusted benchmarks.

There is no question that incentives in the MSSP have been weak. Calls for greater savings and the instinct to lay a pathway are understandable. Relying entirely on incentives to work, without holding providers to an expected pace of savings, may be less satisfying from a regulatory perspective because it requires a leap of faith. But we need to know how ACOs would respond to stronger incentives, and the voluntary nature of the MSSP constrains options for reform. If we give ACOs a pace to beat instead of a reason to set their own, we may never know what could have been.