CMS Issues Proposed Changes to MA Risk Adjustment Process and RADV Audit Update

CMS closed January with two significant developments affecting risk adjustment policy in the Medicare Advantage (MA) space. On January 26, 2026, CMS released the Calendar Year (CY) 2027 Medicare Advantage and Part D Advance Notice (2027 Advance Notice), which proposed notable risk adjustment policy changes. One day later, CMS released a new HPMS memorandum providing updates on its accelerated risk adjustment data validation (RADV) audits. Together, these actions reinforce CMS’s continued emphasis on payment accuracy and program integrity in the MA program.

2027 Advance Notice: Proposed Limits on Diagnosis Sources

MA plans receive risk-adjusted payments based on the diagnoses they submit for enrollees, which translates into higher payments for those enrollees with more serious or chronic conditions. In the 2027 Advance Notice, CMS proposed to make several changes to the MA risk adjustment model.¹ Of particular note is CMS’s proposal to exclude from the risk score calculation “all diagnoses submitted on unlinked chart review records (CRRs).”

Currently, MA organizations (MAOs) can report risk adjustment data to CMS’s encounter data system on an encounter data record (EDR) or as a CRR. A CRR allows MAOs to “add risk adjustment eligible diagnoses or delete diagnosis codes previously reported for plan enrollees.” Linked CRRs are diagnoses that are linked to a specific beneficiary encounter previously submitted on an EDR. Unlinked CRRs, in contrast, are diagnosis codes added based on a chart review that are not associated with a specific, previously accepted claim or encounter for a face-to-face service. Based on findings from HHS-OIG and MedPAC that unlinked CRRs contribute to differential coding across MAOs and may inflate payments, CMS anticipates that “excluding unlinked CRRs from risk adjustment may reduce differences in payment resulting from differences in differential coding across MA organizations” and result in “more accurate payments.” CMS estimates that this change, if adopted, would have a significant impact on program costs, reducing MA payments by 1.53 percent and resulting in an estimated net savings of $7.12 billion.

The exclusion of diagnoses from unlinked CRRs for risk adjustment scoring purposes is a significant change that has been under consideration for years and, if finalized as adopted, would likely have a disproportionate impact on larger MAOs, which generally have more robust retrospective review programs. Among other things, eliminating unlinked CRRs will require plans to shift resources toward ensuring that diagnoses identified through medical record reviews are tied to an actual service record if they want the CRR to count for risk adjustment purposes.    

Updates to RADV Audit Strategy

Turning from risk adjustment methodology to oversight, CMS also provided important updates to its expanded RADV audit strategy. In May 2025, CMS issued a press release declaring its new—and self-proclaimed “aggressive”—enforcement strategy for RADV audits. (The press release is summarized in our previous blog post.) According to CMS, the agency’s revised strategy set the goal of concluding audits for payment years 2018 through 2024 on an expedited timeframe, in part by (i) shortening the medical record submission window, (ii) reducing the number of medical records per hierarchical condition category (HCC) code that may be submitted, and (iii) introducing new technologies to accelerate audits. The agency also announced in the press release that it was establishing variable sample sizes and would expand audits to all RADV eligible contracts. 

In its January 27, 2026 HPMS memo, CMS emphasized its ongoing commitment to a “robust, comprehensive audit strategy” and provided long-awaited updates on how that strategy will be implemented. In announcing the updates, CMS continued to emphasize that it remains several payment years behind in completing RADV audits and reiterated that strengthening oversight of MA payments remains a top priority for the Trump administration. Based on feedback from stakeholders, CMS announced the following updates to its strategy:

• CMS restored the previous five-month medical record submission window and extended the hardship exception request submission window for the payment year 2019 audits, allowing MAOs additional time to secure medical records from providers.

• To “reduce audit overlap and provider burden,” the agency plans to initiate future RADV audits every three months and publish a calendar describing the audit initiation cadence so that industry stakeholders can better plan for their business needs. CMS anticipates beginning the payment year 2020 RADV audits by February 2026.

• CMS clarified that statistically valid, variable sample sizes (ranging from 35 to 200 enrollees) will be used based on contract size or similar criteria for payment year 2020 and later RADV audits. Accordingly, smaller contracts are much less likely to be subject to a 200-enrollee sample.

• Balancing CMS’s goal to accelerate completion of the RADV audits with MAO and provider burden, CMS confirmed that the limit of two medical records per audited HCC will remain for future RADV audits.  

• Finally, CMS continued its promise to “secure new technology, powered by artificial intelligence” to be used as “a medical coder support tool to streamline human coding reviews.” The agency clarified that any decisions that could result in overpayment decisions will be made by human coders. CMS did not provide any further detail on the technology but promised to share more information in the future.

Notably, the memo references the September 2025 decision from the U.S. District Court for the Northern District of Texas in Humana v. Becerra, 801 F.Supp.3d 624 (N.D. Tex. 2025), which vacated certain portions of the RADV audit regulation (including with regard to extrapolation from RADV audits). The memo states that during the agency’s appeal of the decision, CMS will “fully comply with the district court’s order as long as it remains in effect, while continuing to pursue upcoming RADV payment year audits.”

Implications for MAOs

In the 2027 Advance Notice Fact Sheet, CMS explained that it continues “working towards a MA risk adjustment system guided by” the following three principles:

1. simplicity to reduce day-to-day administrative burden for both plans and providers;
2. competition on creating value for patients where risk adjustment facilitates such competition equally for all varieties of plans irrespective of size or resources; and
3. payments that accurately reflect beneficiary health risk and facilitate the efficient use of healthcare resources, enhanced program integrity, and greater accountability.

Consistent with these goals, CMS suggests that its proposed exclusion of diagnoses from unlinked CRRs is targeted at leveling the playing field among MAOs and ensuring payments accurately reflect beneficiary health risk. Likewise, the updates to the RADV audit strategy seek to reduce provider and MAO burden in submitting documentation for audits, while still permitting CMS to efficiently conduct RADV audits and recoup any associated overpayments.

Taken together, CMS’s proposed modifications with regard to appropriate diagnosis sources and its accelerated RADV audit strategy underscore the agency’s broader effort to tighten payment accuracy controls across the MA program. MAOs should evaluate their reliance on unlinked CRRs, strengthen encounter data processes, and prepare for continued increased audit activity as CMS carries forward its multi‑year RADV audit strategy.

1. In addition to proposing the exclusion of diagnoses from unlinked CRRs, CMS also proposed to exclude from its 2027 risk score calculation diagnoses from audio-only encounters (as indicated by modifiers “93” or “FQ,” which were first implemented on January 1, 2022). Based on longstanding policy implemented during the pandemic, diagnoses resulting from “telehealth visits using interactive audio and video telecommunications system that permits real-time interactive communication” can be considered for risk adjustment purposes. Given that the 2024 risk adjustment model was calibrated using 2018 diagnoses and 2019 costs, and was phased in over three years, this is CMS’s first opportunity to calibrate the risk adjustment model to exclude diagnoses from audio-only services that are identified using the specified modifiers. ↩︎