Performance Results Of The Medicare Shared Savings Program In 2021: Continued Uncertainty With Positive Movement

On August 30, the Centers for Medicare and Medicaid Services (CMS) released performance results for the ninth performance year (2021) of the Medicare Shared Savings Program (MSSP). The 2021 data represent the first full year of performance since the start of the COVID-19 pandemic, which continues to impact Medicare spending and cause a high degree of uncertainty in the Medicare program as noted in the 2021 Medicare Trustee’s Report. To that end, CMS maintained programmatic flexibilities, as detailed in our previous article detailing 2020 performance results, throughout the 2021 performance year to help accountable care organizations (ACOs) cope with the unprecedented situation. These flexibilities include removing COVID-19-episode costs from the benchmark and expenditure calculations and allowing ACOs to receive the higher of either their quality score or the national average. As such, this year’s results must continue to be interpreted with caution.

In addition to MSSP performance results, in August the Center for Medicare and Medicaid Innovation (the Innovation Center) released the final summary performance data for the Global and Professional Direct Contracting (GPDC) model, which is transitioning to the ACO Realizing Equity, Access, and Community Health (REACH) model. While the Innovation Center did not release ACO-level results for the GPDC, the cumulative year-to-date figures from April 2021 through December 2021 offer an opportunity to access the GPDC relative to the MSSP at a high level. In total, 53 organizations participated in the GPDC in 2021; however, more than 100 organizations are potentially participating in the ACO REACH model beginning in 2023. Compared to the GPDC, the REACH model uses a benchmark composite measure that incorporates a combination of the Area Deprivation Index and Dual Medicaid Status. At the start of 2023, REACH ACOs will be required to implement health equity plans to identify and mitigate health disparities. Additionally, the portion of the ACO’s performance year benchmark held “at-risk,” contingent upon the ACO’s quality score in REACH decreased to 2 percent from 5 percent in the GPDC, among other changes.

Key findings from this year’s MSSP performance data analysis include:

• Overall, 475 different organizations participated in the MSSP, which is a 7 percent decrease compared to last year’s 513 and a 12 percent decrease from 2019. This is partially explained by CMS’s COVID-19 flexibilities.
• The total number of assigned beneficiaries decreased from 10,614,589 in 2020 to 10,124,325 in 2021.
• The program saw more than $1.6 billion savings in 2021, with similar percentages of ACOs generating net savings (81 percent) and receiving shared savings (58 percent) as 2020.
• ACOs generated approximately $190 net savings per attributed beneficiary, approximately the same as last year.
• Physician group-led ACOs were slightly more likely than hospital-led or jointly led ACOs to realize savings relative to their benchmark and to receive bonus payments.
• Of the 12 ACOs with the majority of their beneficiaries in dual enrollment, all but one of these ACOs produced shared savings.
• More than 41 percent of ACOs were in two-sided risk, up from 37 percent last year and 33 percent in 2019, and two-sided ACOs tended to have higher shared savings rates and slightly better mean quality scores.

We share more details on these key findings below, noting how these changes may factor into the policies that CMS may finalize in its proposed Physician Fee Schedule rule and other potential programmatic changes to ACO REACH. For additional details on how our analysis was done, see the appendix. We find that despite the financial strains of the COVID-19 pandemic, ACOs still performed well and adapted toward urgent patient needs. This suggests that beyond generating cost savings, ACOs can help to address structural challenges in health care such as workforce shortages, fragmentation across payers and providers, and other operational inefficiencies. Increasing the MSSP participation is important to addressing these macro challenges.

A Descriptive Look At MSSP ACOs In 2021

Exhibit 1 provides a descriptive overview of the MSSP ACOs in 2021, including distributions by risk track, governance type, geographic region, and revenue status. In total, there were 475 ACOs in the 2021 MSSP program, a 7 percent decrease from 2020 and 12 percent decrease from 2019. CMS had paused entry into the MSSP program for 2021 during the public health emergency, which limited new ACO entrants. Although total participation in the MSSP decreased, the amount and proportion of ACOs taking on downside risk increased, and those that dropped out of the program tended to be upside-only models.

Exhibit 1: ACO characteristics in the Medicare Shared Savings Program, 2021

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

Exhibit 2 displays the longitudinal performance of the MSSP ACOs over the past decade. Perhaps most importantly, ACOs continued to achieve program savings relative to their benchmark, generating approximately $1.66 billion in total net savings with an average net savings of $190 per capita following the five-year trend of positive net savings. Again, this trend is potentially driven by the delay of new entrants. Compared to the 2020, the percentage of ACOs generating savings relative to their benchmark stayed approximately the same, but the percentage of ACOs receiving performance bonus payments decreased considerably compared to 2020 (but remained similar to 2019). The decline in shared savings rate is likely due to a combination of factors including COVID-19. In 2020, there were substantial declines in health care use as people cancelled elective visits and procedures, while in 2021, use returned to similar levels as before the public health emergency. Selection bias, in which potentially a high number of ACOs concerned about performance terminated participation may also play a role in the decline.

We also compared shared savings rates of all ACOs in 2021 relative to the initial entry date into the program. We found that the oldest ACOs initiated before 2015 were more likely to receive shared savings than younger ones. Low-revenue ACOs, or ACOs with less than 35 percent of total Medicare fee-for-service expenditures from assigned beneficiaries, tended to have joined in recent years. As part of CMS’s efforts to support ACOs with limited resources, low-revenue ACOs have additional time before progressing to downside risk under Pathways to Success rules, and CMS has proposed to provide advance payments to new ACOs that are low revenue.

Exhibit 2: ACO participation and savings over time in MSSP

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

Overall, the trends of maintaining net program savings per capita show the MSSP continues to be successful despite fewer beneficiaries and organizations participating in the program. Important to remember, these decreases are likely driven by COVID-19 and programmatic flexibilities CMS implemented, including pausing new ACO entrants for 2021. The current analysis provides a timely check on the program’s effects, given that full counterfactual comparisons (or comparing to what would have occurred if there were no ACO program) will require additional data and analysis. Past counterfactual analyses and other thought pieces (for example, here, here, here, and here) have typically confirmed benchmark trends and estimate even higher savings from the program. Furthermore, while COVID-19 stressed the health system in many respects and further exposed known weaknesses, the current analysis comports with prior work showing that providers in value-based payment arrangements fared better during a public health emergency than others.

Trends In Shared Savings And Participation For Different Types Of ACOs

Exploration Of Results By ACO Type

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. Exhibit 3 compares the shared savings rates of different types of ACO governance. All types of ACOS, including physician-led, hospital-led, or jointly led, were highly likely to generate savings compared to their benchmarks, and more than half of each type received a bonus payment. Physician-led ACOs were the most likely to generate shared savings bonuses and savings relative to the benchmark, followed by jointly led ACOs, and hospital-led ACOs were the least likely to earn bonus payments and savings relative to the benchmark. These results are consistent with previous years’ findings that suggest physician-led ACOs tend to be more successful at generating shared savings. However, health care usage patterns were also heavily impacted by COVID-19, which may have affected the performance of hospital versus physician-led ACOs by shifting care toward non-acute sites.

Exhibit 3: Percentage of ACOs generating and receiving savings, stratified by ACO governance structure type

Source: Authors’ analyses of propriety Milliman Torch Insight data.

Exploration Of Results By Geography

Exhibit 4 explores how ACOs performed collectively in their states, with only two states—South Dakota and Kansas—having shared losses greater than savings. Regionally, states in the South and West appear to perform slightly better on average than those in the Midwest and Northeast, especially in New England. These results largely comport with our previous analysis on geographic variation and MSSP performance, showing nationwide improvements. Notably, Texas was one of the leading states in savings performance. Texas-inclusive ACOs generated more than $306 million in net savings, accounting for 18 percent of total net savings across all states, and produced an average net savings of $227 per capita (versus $186 per capita for non-Texas ACOs). Among Texas ACOs, those that operated only in Texas outperformed those that operated in Texas and multiple other states: Texas-only ACOs saved $285 per capita on average, while multistate ACOs with Texas saved on average $174 per capita. Additional analyses that leverage other data sets, including hospital price transparency and Health Care Cost Institute commercial data, are necessary to understand ACO penetration and performance in state and regional markets, especially relative to this year’s decline in participation in the MSSP.

Exhibit 4: Mean savings for ACOs by state

Source: Authors’ analyses of public CMS MSSP performance data and propriety Leavitt Partners Torch Insight data.

Shared Savings By Risk Level

In 2021, 41 percent of ACOs were in two-sided risk, representing both an absolute and relative rise in downside risk adoption compared to previous years (see exhibit 5). Consistent with past results, ACOs in two-sided risk models were more likely to achieve savings relative to the benchmark and receive shared savings bonuses than upside-only models. Among two-sided models, 74 percent of ACOs received shared savings bonuses, while 46 percent of ACOs in one-sided risk models did the same.

Exhibit 5: Percentage of ACOs generating and achieving shared savings by risk level

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

Shared Savings Based On Population Characteristics And Safety-Net Providers

To analyze implications for health equity, we compared MSSP performance by Medicare-Medicaid dual eligibility, race and ethnicity, and federally qualified health center (FQHC) involvement. Ideally, a comprehensive analysis would complement spending analyses with analyses of quality performance, but we limited our use of the composite quality score because several factors have complicated the measure’s reliability, such as COVID-19’s impact on assessing care delivery, CMS’s reporting flexibilities during the pandemic, and changing coding patterns. Exhibit 6 shows the shared savings generated for ACOs with varying levels of beneficiaries dually enrolled in Medicare and Medicaid. In general, ACOs with low dual enrollment rates were more likely to receive shared savings bonuses, although ACOs with the highest quintile of dual enrollment produced the greatest net savings.

Honing in further, there were 12 ACOs with the majority of their beneficiaries in dual enrollment. All but one of these ACOs produced shared savings, and their average net savings was $449 per capita. These results indicate that ACOs in the lower quintiles of dual enrollment tended to have greater consistency in generating shared savings, but ACOs with the highest rates of dual enrollment achieved the greatest savings overall. This indicates a significant opportunity for ACOs that serve traditionally underresourced populations to receive large bonus payments when they have the flexibility in revenue to design programs to provide comprehensive, patient-centered care. However, additional analyses are needed to understand if these ACOs are perhaps expending more resources on programmatic and organizational improvements for their assigned beneficiaries than other ACOs, which may reduce the additional savings, or if COVID-19 flexibilities somehow benefited these providers to a greater extent.

Exhibit 6: ACO savings stratified by proportion of dual beneficiaries

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

Exhibit 7 breaks down the shared savings results of ACOs by their proportion of beneficiaries by race and ethnicity, according to CMS data categories. In general, ACOs with higher proportions of patients of color were more likely to realize savings than those with fewer patients of color. These results are important in light of the strong focus on equity by the Innovation Center and CMS more broadly, as they show an association between ACO success in achieving shared savings and great beneficiary racial and ethnic diversity. There were small but noted differences with regards to quality performance when we looked at ACOs in these quintiles. ACOs in the highest quintiles had slightly lower-quality scores than those in the lowest quintiles, suggesting that more resources for quality care should be targeted to safety-net providers that care for racially and ethnically diverse patients.

Exhibit 7: ACO savings stratified by proportion of beneficiaries identifying as Black, Asian, Native American, Hispanic, or other race

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

We also explored ACOs that have greater relationships with safety-net providers. We examined safety-net relationships in two ways: We evaluated net savings per capita for ACOs based on the amount of primary care services provided at FQHCs and rural health clinics (RHCs), and we evaluated ACOs based on how many FQHCs were partners in the organization. When stratified into equal quartiles (exhibit 8), we found ACOs performed similarly well regardless of the proportion of primary care visits at FQHC and RHCs, although there were slight improvements in shared savings for those ACOs with limited FQHC/RHC visits.

We also analyzed ACOs based on the number of FQHCs participating in the ACO (exhibit 9). When analyzed this way, there did appear to be a positive correlation between ACO performance and a higher number of FQHCs for those accountable providers with a concentrated group of safety-net providers. Specifically, while a majority of ACOs have limited to no engagement with FQHCs, ACOs with more than 16 FQHCs (40 in total) were able to generate approximately $120 more in net savings per capita than those ACOs with limited or no FQHC engagement. Additional analyses are necessary, however, especially those that analyze ACO performance from panel data, to understand why this trend may have occurred. Of note, when we examined the mean of the Hierarchical Condition Category model for risk adjustment in the MSSP across ACOs in each FQHC cohort, we found that ACOs with more FQHCs in network tend to have noticeably lower scores than those with fewer FQHCs. This comports with anecdotal evidence that FQHCs do not have the same culture of coding as other providers, which could understate their true savings, given how coding ties into risk adjustment and benchmarks.

Exhibit 8: ACO savings stratified by FQHC/RHC primary care visits

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

Exhibit 9: ACO grouped based on number of FQHCs in their network

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

Experience With Value-Based Payments And Advanced Models

A critical consideration in evaluating ACOs is whether participation in multiple types of value-based payments (VBPs) creates a critical mass that not just sustains transformational efforts but results in a snowballing effect of efficiencies. Meaning, if providers have a greater percentage of their revenue tied to accountable contracts, do they generate additional savings? To understand this question, we evaluated how the number of other contracts an ACO has impacts its mean MSSP savings bonus payment (exhibit 10). While most MSSP ACOs only participate in the Medicare program, performance differed from those with multiple contracts. Although this may correlate with overall ACO size, as quality and mean percentage benchmark scores did not appear to have a relationship with the number of contracts per ACO, our previous analysis shows that when organizations participated in ACO programs beyond Medicare, the ACOs were more likely to receive bonuses and generate savings. For this 2021 analysis, we found that size was not strongly associated with ACO savings, but the largest ACOs had slightly higher shared savings rates. While the specific causes of these trends are unknown and may be confounded by governance structure, larger ACOs with more contracts may have been better equipped to deal with the changing care patterns during COVID-19.

Exhibit 10: Amount of ACO savings received by number of contracts, 2021

Source: Authors’ analyses of public CMS MSSP performance data and propriety Leavitt Partners Torch Insight data.

With that in mind, we also examined how providers in Medicare direct contracting arrangements—“advanced” payment reforms with larger shifts away from fee-for-service to person-level payment—compared to results from this year’s MSSP performance. MSSP ACOs were able to achieve a higher degree of savings/reduction in spending compared to their benchmark than Global and Professional Direct Contracting Model (GPDC) ACOs. However, the differences may be explained by COVID-19 and gaining experience in a new program, as well as programmatic differences.

Discussion And Policy Implications

This year’s results again demonstrate MSSP’s continued success in generating Medicare savings during continued public health threats. While results must be interpreted carefully given the impact of COVID-19, there are several important findings and policy implications, especially relative to major programmatic changes that CMS has proposed. Indeed, CMS appears committed to using the MSSP as a major vehicle for value-based care transformation, including a multitude of proposed reforms in the calendar year to increase participation, especially among providers that care for traditionally underserved populations. These changes include giving some ACOs longer ramp time before entering into risk, making participation in the ENHANCED track (a two-sided risk track with the greatest degree of risk and reward)optional, improving benchmarking by incorporating a prospectively projected administrative growth factor, and providing a one-time fixed payment of $250,000 and quarterly payments for the first two years of the five-year agreement period for new, low-revenue ACOs.

One concern about VBP has been whether it can improve equity. This year’s analysis found that a number of ACOs have patient populations with significant numbers of historically marginalized populations, and that those ACOs were able to be successful. Furthermore, we found that ACOs with unusually large percentages of their beneficiaries dually eligible for Medicaid (>50 percent) achieved substantial shared savings per capita, although this included only 12 ACOs. The ACO program and other VBP programs can be useful tools for improving equity if well designed, given that they have more flexibility to address social drivers of health services, invest in population health management, identify and act on disparities in health care quality and outcomes, build partnerships with community-based organizations, and expand their workforce to involve community health workers, who understand the local community and can help people address their health needs. The MSSP program can further encourage ACOs to increase their enrollment of marginalized populations and reduce health disparities, such as through reforms in CMS’s risk-adjustment and benchmarking, and incorporating additional VBP design elements specifically centered around equity. CMS must monitor these innovations for unintended consequences such as shifting resources away from other potential high-need beneficiaries. This year’s results also showed that nearly a quarter of MSSP ACOs include safety-net providers, and that these ACOs can succeed. In combination, this indicates there is a significant opportunity for ACOs that serve historically marginalized and low-income populations to receive large bonus payments when they have the flexibility in revenue to design programs to provide comprehensive, patient-centered care. More research is needed to understand how safety-net providers can be most successful under VBP models and whether model designs need to be modified to ensure more safety-net providers can successfully participate in ACOs.

Similar to the previous year’s analysis showing that the breadth of an ACO’s participation beyond Medicare may have a relationship to receiving bonus and achieving shared savings, this year’s analysis shows ACOs with multiple contracts may perform better than ACOs in the MSSP alone. As such, finding a path forward on directional alignment between value-based programs in government and commercial contracts is important for the MSSP’s continued success. To that end, CMS should be a leading voice in multipayer alignment and continue to build on state collaborative efforts, such as those in Maryland. The Health Care Payment Learning and Action Network offers CMS a platform to coordinate with payer stakeholders to develop aligned foundational elements for new payment models. It is likely that commercial payers and employers also experience similar multicontract savings benefits.

Like last year, we did not include much analysis of quality because the COVID-19 pandemic affected care patterns that made quality more challenging to assess, and CMS implemented several flexibilities for the model that made the quality data less accurate. A high-level analysis suggests that ACOs tended to perform well on quality, meeting the program benchmarks. It will be important to gauge all spending analysis above based on the impact on quality as well, as ACOs are designed to improve quality while reducing costs.

Overall, this year’s results confirmed last year’s in that VBP models continue to be useful, particularly during a public health emergency. The ACO model offered providers greater revenue flexibility during COVID-19, potentially allowing them to quickly adapt their care delivery and build up critical capabilities to cope with the pandemic better than other providers. Care delivery today remains challenging for many providers—marked by continued COVID-19 burdens and significantly worse population health outcomes, as well as workforce and strains in adapting to new expectations and models for care delivery. Providers received substantial federal financial relief to address fee-for-service revenue shortfalls during the pandemic, which enabled them to weather the disruptions but did not necessarily help them take steps to make adaptations. As the pandemic emergency assistance winds down, an updated and enhanced set of Medicare ACO programs are likely to be even more important in providing the financial and operational flexibility for these organizations to adapt, without adding substantial health care costs and with more accountability for achieving better outcomes.

With a decade of experience in the MSSP model, we continue to see positive movement among ACOs in improving overall quality and reducing costs, as well as promising signs in improving equity, especially for people served by the safety net. Given the decline in participation, it is important that MSSP opportunities continue to be expanded and reach patients who would benefit from the accountability, care coordination, and patient-centered tools that VBP provides.

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 2. He also has been an external reviewer for The John A. Hartford Foundation, and he is a co-chair for the Health Evolution Summit Roundtable on Value-Based Care for Specialized Populations. David Muhlestein is employed by Health Management Associates, 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.