Background

In late July, the Center for Medicare and Medicaid Innovation (CMMI) announced updates to the ACO Realizing Equity, Access, and Community Health (REACH) model beginning in performance year (PY) 2025.¹ This announcement coincided with the recent publication by CMMI of the Global and Professional Direct Contracting (GPDC) Model PY2022 Evaluation Report, which cites some of the early findings of the GPDC model, REACH’s predecessor. Specifically, the report states that “as of PY2022, gross Medicare spending increased for Standard DCEs” (DCEs are Direct Contracting Entities which are similar to the ACOs in REACH). Using a statistical method known as difference in differences, the authors of the aforementioned report estimated an overall $193.3 million² increase in net Medicare spending for DCE participants in Standard Accountable Care Organizations (ACOs). Perhaps to respond to this report’s citation of the increased costs as well as to improve overall model sustainability, CMS introduced the following changes that will reduce expected financial shared savings in PY2025 and PY2026 for many REACH ACOs.

The table in Figure 2 shows a summary, from the ACO’s perspective, of whether the PY2025 changes are expected to produce gains, losses, or vary with the ACO’s specific circumstances.

Below we describe the changes in model feature updates and estimated impact to REACH ACOs. In some instances, we also include an estimated financial impact.⁵

1. Maintain current benchmark blend of historical/regional expenditures for Standard ACO regional blend rate

Description: The blend of historical baseline expenditures and regional expenditures will remain at 55% and 45% for PY2025 and PY2026, respectively, instead of shifting to a 50%/50% blend as previously scheduled.

Impact: We compared the risk-adjusted aged/disabled (AD) expenditures of the 104 Standard REACH ACOs in PY2024 to their respective regional expenditures, and found that approximately two-thirds are regionally efficient⁶, i.e., risk-adjusted expenditures are lower than the rate book. These efficient ACOs will experience a reduction to their expected PY2025 benchmarks due to the reduced regional weight (unless they hit the regional adjustment ceiling both with and without this change). We estimate the impact of this update on the 104 Standard ACO’s AD benchmark reduction for this cohort of ACOs to be -$27 million, or approximately -0.1% of the their aggregate benchmark. On the other hand, regionally inefficient providers will benefit by an approximately +$7.3 million increase (+0.1% of their aggregate benchmark). When accounting for the 100% and 50% shared savings rate applied to gross savings of Global and Professional ACOs, respectively, this ultimately results in approximately a $18 million reduction in the expected shared savings under the current REACH methodology and the new framework.

2. Reduce Ceiling for Regional Blend Adjustment to Benchmark for Standard ACOs

Description: The maximum upward adjustment from blending regional expenditures into the historical benchmark will be reduced from 5% to 3% of the adjusted FFS US Per Capita Cost (USPCC).

Impact: This change is the most financially impactful to the few ACOs affected. It will only impact REACH Standard ACOs that are very regionally efficient, specifically those whose regional benchmark blending currently results in at least a 3% increase to the historical benchmark.

Using the PY2024 REACH ACOs and the new regional weights for PY2025 (i.e., 55%/45%), we can estimate the revised AD benchmark in Milliman ACO Builder® and find that five REACH Standard ACOs have their upward regional adjustments capped based on the 5% ceiling. We estimate that, with the update to a 3% ceiling, an additional nine REACH Standard ACOs would be capped in PY2025. Those that were previously capped at 5% would see their benchmarks further reduced under the 3% cap. For these 14 REACH ACOs, we estimate this methodology change will result in a benchmark impact of -$16 million (averaging -0.9% of benchmark per impacted ACO). Because most of the ACOs impacted are under the Global option, the impact of this change on the benchmark (and thus the shared savings of ACOs) will be a reduction in excess of $15 million.

3. Increase benchmark discount for ACOs in Global risk-sharing option

Description: Although the benchmark discount for PY2025 is unchanged, in PY2026 it will increase from the currently scheduled 3.5% to 4.0% of benchmark.

Impact: This change will impact all REACH ACOs participating in the Global risk option, (approximately 80% of the REACH ACOs in PY2024).

The discount is a defining component of REACH and the most direct way to ensure savings from CMS’s perspective. While CMS is not increasing the discount in PY2025 (i.e., remaining steady at 3.5%), there will still be a meaningful hurdle to overcome in PY2026 (i.e., the discount of 4.0%).

Although this applies beginning in PY2026, we will contextualize this change using the PY2025 benchmark. There are 86 Global risk option Standard ACOs in PY2024 with approximately $25 billion in aggregate benchmark in PY2025. On this basis, the changes represent an estimated $125 million reduction in the expected savings for REACH ACOs (i.e., additional 0.5% multiplied by aggregate benchmark).

4. Improve statistical reliability of benchmark for voluntarily aligned beneficiaries

Description: Voluntarily aligned beneficiaries will receive a benchmark starting in PY2025 that aligns with the methodology used for claims-based beneficiaries as long as the ACO has at least 500 (Standard and New Entrant ACOs) or at least 250 (High-Needs ACOs) voluntarily aligned beneficiaries in at least one of the base years (BY) of the baseline period (i.e., 2021-2023 for PY2025).

Impact: This change will only impact REACH ACOs that had significant voluntarily aligned populations during the historical baseline period.

For those ACOs, the impact of this change will depend on how the cost profile of the voluntarily aligned beneficiaries in the base years compares to those aligned in the performance year. If the strategy to drive voluntary alignment of the ACO has shifted significantly over time, causing fluctuations in voluntary alignment (i.e., credibility), then this methodology is likely to yield an insufficient benchmark for this population.

As an example, consider an ACO utilizing voluntary alignment to a limited degree in 2021 and 2022 (e.g., fewer than 20 beneficiaries). Then, in 2023, the ACO implemented a policy in which it would voluntarily align all the beneficiaries that the physicians were seeing (including those who were not claims-based aligned), which yielded approximately 1,000 beneficiaries. In this scenario, the historical experience from the first two base years (BY1 and BY2) would not necessarily be representative of the voluntarily aligned population the ACO would expect to see in a performance year (close to 1,000) and could lead to an inappropriate benchmark.

5. Increase ceiling for regional blend adjustment to benchmark for High-Needs population ACOs

Description: High-Needs ACOs will have their historical benchmarks calculated for PY2025 and PY2026 using simulated historical alignment (consistent with the methodology used for Standard ACOs) from the 2021-2023 base years.

Additionally, the ceiling for the regional blend adjustment will be set at 9% of the adjusted FFS USPCC for both years.

Impact: This change will impact all High-Needs ACOs (there were 14 in PY2024) and will have varying impacts on each ACO. The previous methodology utilized actual alignment to calculate each historical base year instead of simulated alignment, which could have led to insufficient benchmarks for some High-Needs ACOs. This new methodology will bring the High-Needs ACO methodology in line with the Standard ACO methodology (albeit with different BYs) and create a more appropriate benchmark for most High-Needs ACOs. However, one remaining consideration is that, for a high-needs population, a prospective claims-based alignment methodology will exclude a large percentage of the population who may have expired before the performance period and will introduce greater volatility in the historical spending and thus the benchmark for these ACOs.

The increase in the ceiling of the regional blend adjustment from 5% to 9% of FFS USPCC will be favorable for ACOs that were successful in the High-Needs program in PYs 2021, 2022, and/or 2023, but it will not have a direct impact on all High-Needs ACOs.

6. Ensure budget neutrality of stop-loss reinsurance

Description: A uniform multiplier adjustment will be applied to ensure budget neutrality across the entire stop-loss program (i.e., to ensure that stop-loss premiums collected equal stop-loss payments for the program overall).

The stop-loss program has always been intended to be budget-neutral, but, given differences between the measurement periods and emerging experience, it has not historically been budget-neutral. CMS implemented this change to ensure budget neutrality.

Impact: Overall, this change could yield either a positive or negative adjustment to the stop-loss performance of those ACOs that elect to participate in the stop-loss program.

Historically, High-Needs ACOs were charged a virtual ”premium” (by CMS where the calculation is performed entirely behind the scenes with no money changing hands within the settlement calculations) based on the average stop-loss charge in the county instead of examining their own historical experience as is the case for Standard ACOs. High-Needs ACOs typically have more charges in the stop-loss bands than the average county’s population. This has led to most High-Needs ACOs receiving more stop-loss payout than they were charged and causing at least a portion of the stop-loss model’s imbalance. With the change to the High-Needs ACO benchmarking methodology to utilize historical experience, it is possible that CMS may modify the stop-loss charge calculation for High-Needs ACOs to leverage their own providers’ historical experience, which has the potential of alleviating the imbalance previously observed within this population.

7. Reconcile total monies owed (TMO) in provisional settlement

Description: The calculation of the optional provisional settlement will now take into account total monies owed (TMO), including Enhanced PCC (EPCC) payments. The financial guarantee requirement is reduced from 4.0% to 3.75% for ACOs electing the EPCC/Advanced Payment Option.

Impact: This change to the provisional settlement may make this option less appealing for many ACOs. If EPCC payback is included in the provisional settlement, then most ACOs (with the exception of those expecting very large, shared savings payments) would not expect to receive a positive settlement at this point in the year, which would result in the ACO simply having to pay back a portion of the EPCC early. This may prove problematic for smaller ACOs utilizing EPCC to support other investments such as bonds or infrastructure.

8. Update policy to accommodate Guiding an Improved Dementia Experience (GUIDE) model overlaps

Description: GUIDE model payments for overlapping beneficiaries will be included in the calculation of expenditures and benchmark trend for REACH ACOs. At a very high level, GUIDE model payments represent more expenses than the beneficiaries would have incurred under Medicare FFS.

Impact: For REACH ACOs with dementia populations of similar distribution to those observed nationwide, this change is likely to have a minimal impact on ACO performance, as there will be a similar upward impact on ACO expenditures as well as REACH reference expenditures (which ultimately inform the retrospective trend adjustment and benchmark trend).

However, for REACH ACOs with higher-than-typical prevalence of dementia patients within their populations (e.g., High-Needs ACOs), this could represent a significant additional expense for their ACOs that will not be fully accounted for in the benchmark.

• If overlapping providers (i.e., those in a REACH ACO as well as participating in GUIDE) can capture some of those additional GUIDE costs as revenue this may not be problematic.
• If an ACO’s providers are not participating in GUIDE but other non-ACO affiliated providers are, and these providers are treating the ACO’s attributed population, the overall performance of the ACO may be negatively impacted. This is due to the additional costs to the total cost of care from these GUIDE-attributed lives.

9. Update risk adjustment model

Description: For PY2025, AD risk scores will be calculated as a blend of 33% of the 2020 CMS-Hierarchical Condition Category (HCC) risk model (V24) and 67% of the 2024 CMS-HCC risk model (V28), with the coding intensity factor (CIF) capped at 1%.

Impact: Continuing to cap the CIF in PY2025 (as done in PY2024) reduces uncertainty in what would otherwise be an unknown assumption.

ACOs have been aware of the three-year transition toward the v28 CMS-HCC risk model since its implementation in PY2024. As PY2025 follows the scheduled increase to 67%, the impact on the ACOs will be dependent on how much focus and effort they have put toward acclimating to the new v28 model.

10. Apply standardized Area Deprivation Index (ADI) for health equity benchmark adjustment

Description: The national/state blended ADI will be replaced with a standardized area-level socioeconomic deprivation measure.  

Impact: CMS has yet to provide details on this update but has stated the intention is to improve the accuracy of the adjustment and capture socioeconomic deprivation in areas with high housing values.

11. Adjust PY2023 expenditures for significant, anomalous, and highly suspect (SAHS) billing

Description: CMS observed SAHS billing for two urinary catheter Current Procedural Terminology (CPT) codes in CY2023 (A4352 and A4353), which it plans to exclude from CY2023 expenditures (for both the ACO and the REACH reference population) when they contribute to future benchmarks (e.g., High-Needs ACO benchmarks).

CMS plans to release future guidance regarding how SAHS will be addressed more generally moving forward.

Impact: By removing the experience for these two specific CPT codes from both the ACO experience as well as the REACH reference population, future benchmarks will be calculated as if the catheter claims were not part of the at-risk claims in PY2023. Given the low volume of expenditures of these claims historically prior to the SAHS billing in 2023 (historically about 0.1% of FFS costs), we do not expect the removal of these claims in CY2023 to have a material impact on the benchmark of the ACOs on average.

What’s next

ACOs have at this point submitted and finalized their participant lists for PY2025. ACOs looking to monitor their emerging 2025 experience and beginning to prepare for future performance years should plan to update their projections and understand the impact of these changes in PY2025 and PY2026.

CMS has recently published the PY2025 REACH Rate Book.⁷ Given its release, ACOs should incorporate these model updates into their projections to gain better insight into their expected performance in PY2025. It is worth noting that additional iterations for the PY2024 REACH Rate Book were released as late as December, due to late methodology changes to accommodate fee schedule updates. The PY2025 Rate Book methodology document⁸ for the PY2025 REACH Rate Book released in September 2024 indicates it does not account for the 2025 fee schedules. We recommend that ACOs carefully monitor CMS announcements for any potential revisions to the rate book and update their projections accordingly.

Additionally, monitoring individual provider performance is crucial to ensure all participants contribute to a successful ACO and ultimately achieve shared savings.

For the illustrations of this analysis, we used Milliman ACO Builder Forecast REACH model to quantify these impacts program wide. Deeper provider-level analytics can be performed to help you understand the impact to your ACO.

---

Below is an abbreviated description of the methodology, data, and disclosures utilized in this discussion. For additional information, please reach out to the authors.

Methodology

Milliman ACO Builder Forecast REACH replicates the REACH program methodologies defined by CMMI (https://innovation.cms.gov/innovation-models/aco-reach). The inputs, outputs, and calculations are consistent with CMMI guidance.

Data relied upon

• CMS research identifiable files (RIFs), 100% Medicare claims and beneficiary data (e.g., claims data described at https://www.resdac.org/cms-data/files/carrier-ffs and https://resdac.org/file-availability-vrdc).
• PY2024 REACH/Kidney Care Choices (KCC) Adjusted Medicare Rate Book was initially released in August 2023, updated in October 2023, updated again in December 2023, and is available at https://innovation.cms.gov/media/document/aco-reach-kcc-py24-ratebook.

Caveats and Limitations

Milliman developed Milliman ACO Builder and other related models to estimate the values included in this white paper. The intent of the models is to project settlement estimates under CMMI’s ACO REACH program. We have reviewed the models, including their inputs, calculations, and outputs, for consistency, reasonableness, and appropriateness to the intended purpose and in compliance with generally accepted actuarial practice and relevant actuarial standards of practice (ASOP).

In developing our models, we relied upon data and program guidance supplied by CMS and CMMI. Milliman ACO Builder utilizes the 100% Medicare claims and beneficiary data from CMS research identifiable files (RIFs) of the Virtual Research Data Assistance Center (VRDC). We performed a limited review of the data used directly for this report but have not audited or verified this data and other information. If the underlying data or information is inaccurate or incomplete, then the results provided in this white paper may likewise be inaccurate or incomplete. CMMI has the ability to make unanticipated changes to program specifications and revise its previously issued data. Different specifications or data will produce different estimates. We used approximate methods and data that may be different from what CMMI would use in these calculations.

The values contained within our model are based on a large set of historical nationwide healthcare data. Estimates based on or derived from this data and supporting files are dependent upon values and options entered by the user. Milliman ACO Builder is based on historical data and may not reflect current and future financial relationships. Actual results will vary from estimates. Actual results should be monitored and corrective steps taken as necessary.

---

¹ CMS. https://www.cms.gov/aco-reach-model-performance-year-2025-model-update-quick-reference.

² See https://www.cms.gov/priorities/innovation/data-and-reports/2024/gpdc-2nd-ann-report p36, which is supported by Exhibit ES.5 of https://www.cms.gov/priorities/innovation/data-and-reports/2024/gpdc-2nd-ann-report.

³ An ACO or provider’s financial efficiency can be measured as the ratio of its expenditures relative to the regional expenditures (e.g., the ACO REACH rate book) on a risk-adjusted basis. An efficient ACO has a ratio less than 1.0; while an inefficient ACO has a ratio greater than 1.0.

⁴ Ibid.

⁵ For estimates provided we relied on PY2024 Standard ACO participant lists. We projected to PY2025 with data through calendar year (CY) 2023 using the 2024 ACO REACH Rate Book, USPCC trends from the 2025 Medicare Advantage Rate Announcement, and Milliman ACO Builder®.

⁶ An ACO or provider’s financial efficiency can be measured as the ratio of its expenditures relative to the regional expenditures (i.e., the ACO REACH rate book) on a risk- adjusted basis. An efficient ACO has a ratio less than 1.0, an inefficient greater than 1.0.

⁷ CMS. https://www.cms.gov/aco-reach-kcc-py25-ratebook.

⁸ CMS. https://www.cms.gov/priorities/innovation/files/aco-reach-and-kcc-models-py25-rate-book-development.pdf.