CMS Proposed Rule for Medicare Advantage Plans Details Equity Requirements for Use of AI

The Centers for Medicare & Medicaid Services (CMS) released a proposed rule putting forward new policies for the Medicare Advantage (MA) and Part D programs for contract year (CY) 2026. While certainly not the focus of this proposed rule, the rule also provides guidance for the use of artificial intelligence (AI) by MA plans.

Specifically, to address the growing integration of AI tools in MA plan operations, CMS proposes to require MA organizations ensure that services are delivered equitably, regardless of whether they are delivered through human or automated systems. This proposed provision would be added to the MA regulation that govern MA’s obligations to provide access to services in a culturally competent manner to enrollees with limited English proficiency or reading skills, with disabilities, diverse gender identities among other enrollee attributes.¹ While recognizing the efficiencies AI can bring, CMS highlighted concerns, which CMS has raised previously,^2,3 about “algorithmic discrimination,” where AI usage may deepen biases and inequities within the health care system. CMS’s proposed requirement “intends to make clear that MA organizations must provide all enrollees, without exception, equitable access to services, including when MA organizations use AI or other automated systems to aid their decision-making” and “requires MA organizations to ensure services are provided equitably irrespective of delivery method or origin, whether from human or automated systems.”

CMS provides  example of how MA plans could demonstrate compliance, including ensuring (1) that they understand, recognize, and limit the impact of biased data inputs within any AI and/or automated system they utilize; (2) that they create and follow a process to regularly review any automated system they utilize to confirm that it use is non-discriminatory; and (3) "that outputs with a known discriminatory bias are not used within a MA organization’s automated system in a manner that discriminates.” Consistent with CMS’s oversight authority, CMS may audit and perform other compliance activities to assess MA organizations compliance with this rule. CMS further states that “In the event that an MA organization licenses an AI or automated system, or contracts with a third party for services that are furnished using one of these tools, the MA organization will be held responsible ... [and consistent with existing regulations] an MA organization is ultimately responsible even if it uses an First Tier, Downstream, and Related Entity (FDR) to fulfill obligations and responsibilities.”

As part of the regulated proposal, CMS also plans to adopt the definition of “automated system” from the White House Office of Scient and Technology Policy’s Blueprint for an AI Bill of Rights⁴ and “patient care decision support tool” from 45 CFR 92.4.⁵

With a change in administrations looming, it is not assured that this proposal will be finalized. Comments on the proposed rule are due January 27, 2025, after the January 20 inauguration of president-elect Donald Trump. This means it will be left to the Trump Administration to finalize—or ignore—what the agency has just proposed.

If you have questions about equity requirements for the use of AI, please reach out to Randi Seigel (rseigel@manatt.com) or Annie Fox (afox@manatt.com).