Can the feds keep up with AI in health care, as insurers automate more claims denials?

Artificial intelligence (AI) has proliferated the health care sector, sparking litigation and increased regulatory activity. The impact of AI in health care has catapulted regulatory agencies’ efforts to keep up. Health care entities, particularly in the Medicare Advantage space, may need to calibrate their business technologies and compliance structures to adapt to this evolving legal landscape.

Pending litigation against Medicare Advantage plans: Uncharted territory?

In late 2023, two class action complaints were filed against two large Medicare Advantage organizations, Humana Inc. and UnitedHealthcare Inc. in federal court. The complaints’ core allegations and theories of recovery are largely the same, centering on the insurers using AI to make allegedly improper coverage determinations for Medicare Advantage enrollees.

Specifically, the complaints allege that each insurer deploys a particular AI tool to “prematurely and in bad faith discontinue payment for health care services” based on “rigid and unrealistic predictions for recovery.” The plaintiffs’ claims for relief are grounded in a host of state insurance law and common law contract claims, including breach of contract, bad faith and unjust enrichment.

After speculation about the lawsuits’ broader implications, the insurers each recently filed motions to dismiss. The arguments common to both motions are the court lacks subject matter jurisdiction because the plaintiffs’ claims arise under the Medicare Act and the plaintiffs failed to exhaust their administrative remedies; and the plaintiffs failed to state a claim because their state law claims are preempted by the Medicare Act.

On the first point, the defendants argue that, although the plaintiffs try to “dress up” or “disguise” their claims under state law causes of action, the “fundamental premise” remains unchanged: the plaintiffs disagreed with an MA plan’s coverage determination and process. Thus, the defendants argue, the plaintiffs’ claims “arise under” the Medicare Act, and the plaintiffs must exhaust a mandatory administrative appeal process before seeking judicial review. In any event, the defendants assert that the Secretary of Health and Human Services (HHS)—not the MAOs—would be the proper defendant.

Second, the defendants argue that, at bottom, the plaintiffs’ claims relate to Medicare coverage, and their state law claims are preempted by Medicare’s “vast regulatory scheme.”

At first blush, the complaints may appear to present novel legal issues because of the seeming newness of Medicare Advantage plans integrating AI into their processes. But the dispositive legal questions at this stage—i.e., whether the plaintiffs’ claims arise under the Medicare Act and whether the Medicare regulatory scheme preempts the plaintiffs’ state law claims—rely on decades of precedent.

In short, while the AI angle may be relatively new, the legal questions posed by the defendants’ motions are not. Notably, this month the plaintiffs are expected to file amended complaints, which likely will aim to address the legal issues raised in the motions to dismiss.

AI developments in the regulatory arena

On the other hand, the Centers for Medicare & Medicaid Services (CMS) issued new guardrails germane to the AI tools at issue in the insurer lawsuits. The Contract Year 2024 MA and Part D Final Rule (CMS-4201-F) (Final Rule) went into effect January 2024—just weeks after the insurer complaints were filed. However, the proposed rule was published back in December 2022, on the heels of an April 2022 HHS Office of Inspector General report (OIG Report) finding that 13% of the reviewed MA plan prior authorization denials issued by 15 selected  insurers were for services that met traditional Medicare coverage rules. OIG Report at 9. Consequently, CMS proposed “certain guardrails” for how “utilization management tools are used, and associated coverage decisions are made.” Proposed Rule at 79498.

The Final Rule squarely addresses certain aspects of using AI tools in MA coverage determinations—requiring, for example, insurers to “ensure that they are making medical necessity determinations 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.” Final Rule at 22195.

Related: Denied by AI: Medicare Advantage’s ‘predictive’ software cuts off care, say feds

Further, if using AI in coverage decisions, the Final Rule requires MA plans to “understand the external clinical evidence relied upon” by AI and “how that evidence supports the coverage criteria applied by these tools. The MA plan must make the evidence that supports the internal criteria…publicly available, along with the internal coverage policies themselves.”

After the Final Rule went into effect, CMS issued a Feb. 6, 2024, memorandum (CMS Memo) to address how CMS “expect[s] MA plans to comply with these new rules.” Notably, the CMS Memo instructs that an “algorithm or software tool can be used to assist MA plans in making coverage determinations,” but “the algorithm or artificial intelligence [must] compl[y] with all applicable rules.”

Echoing the Final Rule, the CMS Memo reiterates that MA plans must make medical necessity determinations based on “the individual patient’s circumstances, so an algorithm that determines coverage based on a larger data set instead of the individual patient’s medical history, the physician’s recommendations, or clinical notes would not be compliant” with Medicare regulations.

As relevant to the insurer class actions, the CMS Memo affirms that AI tools can be used to assist Medicare Advantage plans in making coverage determinations but cautions against relying on AI as the sole basis for their decisions.

Although the Final Rule and CMS Memo seem to leave certain standards ambiguous, what appears clear, however, is that CMS will be evaluating Medicare Advantage plans’ compliance with the Final Rule’s coverage and utilization management requirements in both “routine and focused audits.”

Practice tips and practical implications

It may be a long time before litigation generates merits determinations as to the proper use of AI in Medicare Advantage plan administration. In the meantime, insurers utilizing AI in coverage determinations should prepare for audits. Preparation may include (1) reviewing AI data sources and methodology for potential publication, and (2) evaluating whether the AI is introducing, perpetuating, or exacerbating discriminatory bias.

Valerie Cohen is a partner and Sophia Porotsky is an associate in the New York office of Venable LLP.