Key takeaways
- AI and other algorithms can be used as part of making Medicare Advantage coverage decisions.
- CMS reminds Medicare Advantage organizations that coverage must be based on individualized determinations.
- AI and other algorithms should not perpetuate or create discriminatory and biased results.
Medicare Advantage (MA) organizations that use modern technology (such as artificial intelligence (AI)) to aid in making efficient coverage determinations are in the crosshairs of both Congress, which has held hearings and requested documents from MA organizations about AI and other algorithm use, and members who have filed class action lawsuits alleging certain MA organizations are using the technology to wrongly deny coverage. We have previously written about these lawsuits and made recommendations to help reduce risk. Last November, members of the House of Representatives urged the Centers for Medicare & Medicaid Services (CMS) to assess how MA organizations use artificial intelligence and algorithms to guide coverage decisions.
On February 6, 2024, CMS released an FAQ memo clarifying how a new Medicare rule (Final Rule) and the Affordable Care Act impact the use of AI and algorithms by MA organizations in making Medicare Advantage coverage determinations. The final rule (CMS Final Rule, CMS-4201-F) became applicable to coverage beginning January 1, 2024. It updates Medicare Advantage coverage criteria, including for basic benefits, prior authorization protocols, and the routine review of utilization management tools.
The FAQ states that MA organizations may use AI and other algorithms “in making coverage determinations,” as long as coverage determinations are made in compliance with applicable laws and regulations. It also provides guidance for MA providers and Medicare-Medicaid Plans navigating the evolving landscape of AI regulations, discussed below. It elaborates on CMS’ previous comments that accompanied the final rule, which recognized that MA organizations could use AI or algorithms to make medical necessity determinations insofar as they are “based on the circumstances of the specific individual…as opposed to using an algorithm or software that doesn't account for an individual's circumstances.” We have previously discussed this final rule in more detail in our January 17, 2024 Managed Care Outlook article.
This latest FAQ provides additional detail describing how the final rule applies to MA organizations’ use of AI as it relates to coverage determinations. Some noteworthy FAQ guidance includes:
- Decision-making should be based on the “individual patient’s circumstances,” not only on larger data sets. The FAQ underscores again the importance of basing coverage decisions on individual patient circumstances, not only on comprehensive data sets used by AI. This means that, according to the final rule, determinations should not be made based upon characteristics of broader pools of data alone, but should also rely on factors such as an individual’s medical history, recommendations from physicians, or clinical notes. While certain predictive outputs from AI could be “used to assist providers or MA plans,” the FAQ makes clear that such predictions alone are insufficient in the decision-making process. CMS draws examples from decisions regarding requests for post-acute care services, or denials or downgrades of observational stays during inpatient admissions.
- Limit denials based on publicly available coverage criteria. Under the final rule, MA organizations may deny coverage for basic benefits solely based on publicly available coverage criteria or other permissible bases. CMS recognized the potentiality of AI to “drift” as it is trained or used, which could result in altering an MA plan’s coverage criteria over time. As a result, the FAQ states that any algorithm or software tool must not alter static coverage criteria or apply other “internal coverage criteria” not otherwise publicly available or otherwise in compliance with established guidelines.
- Addressing discrimination and bias. The FAQ also reminds MA organizations that the Affordable Care Act mandates MA organizations to remain vigilant against discrimination and bias in AI by ensuring that their algorithms and software tools do not perpetuate or exacerbate existing biases or introduce new biases. This requirement may necessitate ongoing monitoring and evaluation of AI and other algorithms to detect and mitigate any instances of discrimination or bias that may arise.
Overall, the guidance provided by CMS through the FAQ is another step in shaping the regulatory landscape surrounding AI utilization by MA organizations. It may also provide some helpful guidance to other managed care organizations that look to CMS rulemaking when developing policies and procedures for other lines of business.
Client Alert 2024-035