The Medicare Shared Savings Program In 2020: Positive Movement (And Uncertainty) During A Pandemic

On August 25, the Centers for Medicare and Medicaid Services (CMS) released performance results for the eighth performance year (2020) of the Medicare Shared Savings Program (MSSP). The results for this year need to be interpreted carefully given that the majority of the performance year occurred during the global COVID-19 pandemic. As a result of the pandemic, health care utilization patterns changed markedly, especially early in the public health emergency, with increases in acute care for COVID-19-related concerns, substantial decreases in elective surgeries and procedures, and decreases in primary care and preventive services. Early analyses suggest that COVID-19-related use did not increase as much as was feared, and that much other health care recovered from a substantial dip and began to move back toward baseline levels by the end of 2020. However, as noted on August 31, 2021, in the Medicare Trustees’ report, “[t]here is an unusually large degree of uncertainty with … COVID-related impacts and … future projections could change significantly as more information becomes available.”

To help accountable care organizations (ACOs) during the public health emergency, CMS introduced multiple flexibilities to the MSSP in an interim final rule effective May 2020 (and later a final rule effective January 2021 that will affect MSSP’s 2021 performance year). The flexibilities include changes to: shared savings and losses calculations (such as carving out certain COVID-19 episodes of care); quality reporting; benchmarks; and beneficiary assignment, including an expanded definition of primary care services to include telehealth codes for telephone communications, e-visits, and virtual check-ins.

Recognizing the multiple changes during the 2020 performance year, including the full implementation of the “Pathways to Success” redesign, this post outlines key findings from the performance data and Milliman Torch Insight data, including the following:

• Overall, 513 different ACOs participated in MSSP in 2020, a 5 percent decrease from last year, but the number of overall covered lives increased. The average ACO was slightly larger in size compared to 2019.
• The program saw savings in 2020, exceeding $1.86 billion in net savings to CMS compared to benchmarks.
• ACOs generated approximately $190 net savings per attributed beneficiary, an increase over the $85–$88 net program savings per beneficiary from 2019.
• Sixty-seven percent of ACOs received the shared savings bonus in 2020, compared to 50–57 percent of ACOs in 2019.
• Similar to 2019’s results, all types and sizes of ACOs achieved net savings per capita.
• Thirty-seven percent of ACOs adopted two-sided risk models (up from 33 percent in 2019), with 88 percent of ACOs in downside risk tracks receiving shared savings bonuses compared to 55 percent in upside-only risk.
• While ACO quality is difficult to assess for 2020, our initial analysis suggests the feared quality drops did not occur. ACOs may also have been better than other providers at continuing primary care and preventive services during the pandemic.
• ACOs with a larger number of contracts achieved savings at higher rates suggesting that having a critical mass of value-based payment contracts matters.
• While savings are challenging to assess, ACOs may have been in a better position than other providers to continue delivering needed care during a public health emergency and to continue positive shifts in health care delivery, such as appropriate use of telehealth.

We share more details on these key findings below and articulate policy implications for the MSSP, highlighting how the program may evolve after the public health emergency.

A Descriptive Look At MSSP ACOs In 2020

Exhibit 1 summarizes the characteristics of MSSP ACOs for 2020. The 2020 group of MSSP ACOs differed from prior years in that ACOs had slightly more experience in the program (4.4 years in 2020 versus 3.9 years in 2019 and 3.4 years in 2018). There were also fewer ACOs in 2020 than 2019, with about 5 percent fewer overall. Slightly more ACOs in upside-only arrangements dropped out of the program than those in downside risk tracks (similar to prior years). However, the average size of ACOs has slightly increased to approximately 20,700 attributed beneficiaries, so that the MSSP ACO program included approximately 10.6 million attributed beneficiaries in 2020—around 28 percent of traditional Medicare beneficiaries.

Exhibit 1 also shows the number of ACOs by risk track, leadership type, and revenue status, all of which are fairly similar to last year’s results.

Exhibit 1: ACO characteristics in the Medicare Shared Savings Program, 2020

Source: Authors’ analyses of public CMS MSSP performance data and proprietary Milliman Torch Insight data.

Overall Cost And Quality Results For 2020

As shown in exhibit 2, ACOs continued to achieve net savings compared to their benchmarks for the fourth year in a row. Net program savings compared to benchmarks exceeded $1.86 billion, an increase of approximately $670 million from 2019 (not shown). Sixty-seven percent of ACOs received a shared savings bonus in 2020, an increase over the 50–57 percent of ACOs that did so in 2019. ACOs generated approximately $190 net savings per attributed beneficiary (compared to their benchmark), an increase over the $85 to $88 net program savings per beneficiary from 2019. There continues to be significant variability among ACOs, ranging from net losses of approximately $4,000 per beneficiary to net savings of $3,800 per beneficiary (not shown).

Exhibit 2: ACO participation and savings over time in MSSP

Source: Authors’ analyses of public CMS MSSP performance data.

The trends of increasing net program savings per capita and percentage of ACOs achieving shared savings from the prior years have continued—a good sign for the program when coupled with growing covered lives and increasing adoption of downside risk. However, the net savings numbers must be interpreted with caution. Given that we calculated net savings compared to program benchmarks, it is increasingly important to understand the current process for calculating ACO benchmarks, which depends on a combination of regional and historical components. Before each ACO enters the program, CMS determines the organization’s benchmark based on its historical use; that benchmark is updated each performance year based on a blend of national and regional health care cost growth trends with limited adjustments for changes in case-mix and severity. The effect of comparing ACOs’ 2020 performance against these two components is not yet known—regional comparisons will likely be minimally affected by the pandemic, while COVID-19 usage may be different from historical patterns. A further nuance for savings calculations is that CMS implemented additional program flexibilities that affected savings calculations during 2020, such as excluding COVID-19-related care costs.

Another consideration is that comparisons against program benchmarks have historically underestimated the likely savings that would have occurred if there was no ACO program. Such an estimate would require the use of research methods to establish a plausible counterfactual (as researcher did in the studies here, here, here, and here), and the currently available CMS data do not allow for that level of analysis. However, the data do provide an early sense of how ACOs generally fared during the public health emergency, with ACOs tending to have positive financial performance even during difficult conditions.

One concern was that the public health emergency might substantially reduce overall quality. However, that does not appear to have happened, and ACOs were able to maintain quality while juggling unexpected challenges. It is somewhat difficult to exactly assess quality for 2020, given that CMS provided multiple flexibilities on reporting quality measures for the public health emergency. This included waiving requirements to report some measures, changing some measures from pay for performance to pay for reporting, and allowing ACOs to receive either their quality score or the national average due to the Extreme and Uncontrollable Circumstances Exception.

One marker of quality is an ACO’s ability to continue primary care visits during the public health emergency. For ACOs participating in MSSP for 2019 and 2020, we found the rates of primary care visits (evaluation and management [E&M] visits per 1,000 person-years) declined by 7.6 percent. However, this was less of a decline than for Medicare overall, in which we saw primary care visits per person nationally dropped by 13.3 percent (drawing on Medicare Part B utilization data and examining E&M visits defined by the Berenson-Eggers Type of Service [BETOS] system [M1A-M6]). Careful follow-on health care claims analyses (some of which are already starting) can validate these early results and provide more context about the reasons for the change, such as primary care use beyond E&M visits. However, the high-level takeaway is that ACOs seemed to be able to shift more rapidly during the pandemic.

Trends In Shared Savings For Different Types Of ACOs

As with prior years’ releases, we examined the MSSP results to identify how different types of ACOs fared, based on their ownership structure, size, age, and other characteristics. In general, the trends remained the same from prior years.

Exhibit 3 explores the portion of ACOs that received bonuses and shared savings by ACO type. Physician-led ACOs were most likely to receive a bonus and generate savings compared to hospital-led and integrated ACOs. These findings are consistent with findings in previous years in which physician-led and smaller ACOs were more likely to generate shared savings. Physician-led ACOs received bonuses and generated savings at rates of 70 percent and 85 percent, compared to hospital-led ACOs at 66 percent and 78 percent, respectively, and integrated ACOs at 63 percent and 85 percent, respectively.

Exhibit 3: Percentage of ACOs receiving bonus and achieving shared savings, by ACO governance structure type, 2020

Source: Authors’ analyses of public CMS MSSP performance data and proprietary Milliman Torch Insight data.

In the past, we had noted that the longer a program had been in MSSP, the more likely it was to achieve shared savings and earn bonuses. As shown in exhibit 4, ACOs that were more experienced did seem more likely to reach savings, but the most noticeable effect was the sudden increase in the share of ACOs achieving savings after the third year of being in the MSSP—with 52 percent achieving savings in their third year and 83 percent doing so in their fourth. This is in line with our previous work noting that ACOs have an initial learning curve and need a period of one to three years to get acclimated to the MSSP. However, the age trend was less pronounced in 2020 than in prior years, with much more variation among years.

Exhibit 4: Percentage of ACOs achieving shared savings, by ACO’s time in the MSSP, 2020

Source: Authors’ analyses of public CMS MSSP performance data.

Reaching VBP Critical Mass

An ongoing question for evaluating value-based payment (VBP) models, including ACOs, is the extent to which organizations need a critical mass of their payments in such arrangements before it is sustainable to redesign care fundamentally. We examined these questions by looking at whether organizations with more ACO contracts were more likely than those with fewer contracts to receive bonuses or generate savings compared to their benchmark. In exhibit 5, we see that ACOs with a greater number of contracts were more likely to receive bonuses and generate savings (64 percent of ACOs with one contract received bonus payments compared to 86 percent of those with four or more contracts; 80 percent of ACOs with one contract also generated savings for the Medicare program compared to 95 percent of ACOs with four or more).

Exhibit 5: Percentage of ACOs receiving bonus and achieving shared savings by number of contracts, 2020

Source: Authors’ analyses of public CMS MSSP performance data and proprietary Milliman Torch Insight data.

When organizations participated in ACO programs beyond Medicare, the ACOs were more likely to receive bonuses and generate savings. When organizations only participated in Medicare ACO programs, only 65 percent of ACOs received bonus payments, and 81 percent generated savings. If organizations participated in Medicare and Medicaid ACOs, the percentage of ACOs receiving a bonus increased to 80 percent. If they participated in commercial ACO programs in addition to Medicare, 73 percent of ACOs received bonuses, and 88 percent achieved savings. This effect of payer participation was further increased if the organizations participated in Medicare, Medicaid, and commercial ACO programs; 92 percent of these ACOs were able to receive a bonus payment, while 100 percent generated savings.

Exhibit 6: Percentage of ACOs receiving bonus and achieving shared savings, by payer participation, 2020

Source: Authors’ analyses of public CMS MSSP performance data and proprietary Milliman Torch Insight data.

Increased payer participation and number of contracts (exhibit 5) both led to greater success in generating savings for ACOs. This may indicate that there are potential spillover effects from other contracts and payers that lead ACOs with greater VBP involvement to be better positioned to achieve savings.

Shared Savings By Risk Level

Thirty-seven percent of MSSP ACOs were in two-sided risk models in 2020, marking a continued increase in this trend. This year, ACOs in two-sided risk models were more likely than those in one-sided models to receive bonuses and generate savings (exhibit 7). Fifty-five percent and 88 percent of ACOs in one-sided and two-sided risk models received bonuses, respectively, and 75 percent and 97 percent of those in one-sided and two-sided risk models generated savings, respectively. Previous years have also revealed the same results, indicating that ACOs that are able to take on greater risk may have also improved organizationally to succeed in the model.

Exhibit 7: Percentage of ACOs receiving bonus and achieving shared savings by risk level, 2020

Source: Authors’ analyses of public CMS MSSP performance data and proprietary Milliman Torch Insight data.

Even though it seems that ACOs with greater risk were able to achieve higher levels of success, in light of the pandemic, CMS chose to pause automatic advancement along its glide path for increasing levels of risk for performance year 2022.

Discussion And Policy Implications

COVID-19 has strained the health care system overall, revealing its limited resilience. In general, organizations involved in value-based payment arrangements, including ACOs, managed the public health emergency better than others. VBP gave health care organizations more financial flexibility to redirect resources to where they were needed, such as rapidly ramping up telehealth or redeploying care coordinators to serve as contact tracers. Furthermore, organizations participating in VBP models were more likely than other providers to have built organizational competencies, such as new care models or infrastructure, they could leverage during the public health emergency. Therefore, we should expect that organizations participating in the MSSP, even in upside-only tracks, would have better resilience than organizations not participating in VBP models at all. Accordingly, CMS policies should encourage more participation in the MSSP, potentially with some modifications to performance measures and benchmarks based on learnings from the public health emergency.

This year’s ACO results reinforce the above theme. MSSP performance data from 2020 showed important positive trends that must be interpreted carefully given the extraordinary impact of the pandemic, with COVID-19 changing health care use patterns in ways that become clearer with time. As noted, the drivers of savings compared to benchmarks are a complicated story. The regional component of benchmarks should not be heavily affected by COVID-19 as the overall impact of the pandemic should largely be netted out. The pandemic’s impact on the historical component is less clear. When compared to historical trends, early results suggest that COVID-19-related use did not increase as much as was feared, while use of other care dropped dramatically then recovered toward baseline levels.

Moreover, ACOs may have been in a better place than others to continue needed chronic care management during the pandemic due to population health tracking systems, care coordinators, and greater ability to shift to virtual and at-home care options. In fact, this seems to have happened; across Medicare, costs were 7 percent lower in 2020 relative to 2019, while 2020 ACOs were collectively only 3.7 percent below their historical levels. This situation will need to be watched carefully in 2021, as ACO financial performance could be more challenging if there is substantial postponed care (although it seems like that may not have been the case).

One concerning issue for the future is that CMS postponed new entrants to the MSSP for 2021. This lowers the number of opportunities for organizations to join VBP models, which reduces readiness for the next emergency. CMS data suggest that only 477 ACOs exist for 2021 compared to 513 for 2020. When coupled with the pause on applications for the 2022 cohort of the Global and Professional Direct Contracting Model, this decrease means organizations have fewer options to join accountable care models through CMS, and smaller organizations may have even fewer effective options given the level of financial risk required for some models. Suspending ACO models rather than finding ways to build on them is a missed opportunity to advance better care.

VBP suspensions are a further challenge given that CMS has identified accountable care models as a key strategic priority, and our recent research has found that ACO growth across all payers, including Medicare, has plateaued since 2018. CMS should un-pause participation in ACO-like models and could consider continuing an upfront-savings option like the recently ended ACO Investment Model, which successfully increased MSSP expansion to rural areas, underserved areas, and new areas with low ACO penetration—while saving Medicare money and decreasing unnecessary use. There are also opportunities to take steps to support advanced ACOs in multipayer contexts, which the Health Care Payment Learning and Action Network (LAN) aims to encourage. Such multipayer efforts will also create a better environment when the Center for Medicare and Medicaid Innovation (the Innovation Center) rolls out its next round of models.

There may also be opportunities to strengthen the ACO model to help with the end of the pandemic and to improve resilience overall. CMS should consider longer performance periods to allow for utilization and population health measures to be averaged over several years. It should also consider updated performance measures and pandemic-related “quality boosts” to reward ACOs that did well in managing COVID-19, as well as consider quality measures that encourage getting beneficiaries vaccinated or getting back to accountability for population health measures that were deferred early in the pandemic. Furthermore, there may be ways to build on the lessons learned about virtual care models, as ACOs develop ways to provide greater access to telehealth coupled with a focus on appropriate use and accountability for total cost of care and outcomes. For models administered by the Innovation Center, this may be through greater clarity and technical guidance on telehealth waivers along with common waivers across all Innovation Center-sponsored models.

In 2020, CMS implemented multiple flexibilities to help the health system, but they do make it difficult to compare this year’s results (especially quality results) to prior years and to interpret in general. Those flexibilities were successful in helping organizations continue in the program and manage the unprecedented strain. There are some technical fixes that could be made for the future, such as refining the benchmark calculations so that they reflect non-COVID-19 performance more accurately (or else 2020 and 2021 will have to be excluded from future benchmark calculations). Technical revisions may be needed to the “Extreme and Uncontrollable Circumstances Exception,” as it was designed for regional and not national disasters.

Another key priority for the Biden administration is to advance equity, especially given the disparities revealed through the pandemic. There is growing movement in VBP models to address social needs, which can help advance equity. In theory, VBP models should be useful for improving equity, as the models give greater financial flexibility for health care organizations to meet people’s health-related social needs and to proactively reach out to populations, as opposed to waiting for people to come for a clinic visit. However, payment model design factors can limit how well VBP models encourage equity, such as performance measure specifications, risk adjustment, and data collection. These challenges can be especially challenging for safety-net organizations, although multiple safety-net organizations have shown substantial success under VBP arrangements. For the ACO program, a first step may be CMS improving race and ethnicity data, which could be used for stratifying performance measures and enabling ACOs to conduct proactive care outreach to reduce disparities.

In summary, the Medicare Shared Savings Program continues showing positive trends for 2020. MSSP and other VBP programs have helped health care organizations be resilient during COVID-19 by providing organizations with care delivery and financial flexibility. CMS and other payers should expand opportunities for health care organizations to participate in accountable care programs while adjusting the program to advance its strategic equity goals.

Authors’ Note

Robert Saunders has a consulting agreement with Yale-New Haven Health System for the development of measures and development of quality measurement strategies for the Center for Medicare and Medicaid Innovation alternative payment models under the Centers for Medicare and Medicaid Services Contract No. 75FCMC18D0042/Task Order No.75FCMC19F0003, “Quality Measure Development and Analytic Support,” Option Year 1. David Muhlestein is employed by Leavitt Partners, which consults about ACOs and works with providers and ACOs. Dr. Bleser has previously received consulting fees from Merck for research for vaccine litigation unrelated to this work, consulting fees from Gerson Lehrman Group, Inc., on health policy subject matter expertise unrelated to this work, from BioMedicalInsights, Inc., for subject matter expertise on value-based cardiovascular research unrelated to this work, from StollenWerks LLC, for consulting on Medicaid managed care expertise unrelated to this work, and serves as board vice president (uncompensated) for Shepherd’s Clinic, a clinic providing free health care to the uninsured in Baltimore, Maryland. Mark B. McClellan is an independent board member on the boards of Johnson & Johnson, Cigna, Alignment Healthcare, and Seer; co-chairs the CEO Forum for the Health Care Payment Learning and Action Network; and receives fees for serving as an adviser for Blackstone Lifesciences, Cota, and MITRE. The Duke-Margolis Center for Health Policy values academic freedom and research independence, and its policies on research independence and conflict of interest are available at: https://healthpolicy.duke.edu/research-independence-and-conflict-interest.