New CMS Guidelines Say AI Can’t Deny Medicare Advantage Patients Coverage

Artificial intelligence (AI) has been making headlines across sectors for well over a year now — at least since the groundbreaking launch of ChatGPT in November 2022. And the health care industry is no exception: As AI continues to sharpen and evolve, health care officials of many stripes have found ways to use the technology to their benefit.

In one study, for example, ChatGPT showed an accuracy rate of greater than 98% in choosing the correct screening method for breast cancer by analyzing patient data and clinical guidelines. But according to a policy memo issued by the Centers for Medicaid & Medicare Services (CMS) to Medicare Advantage insurers in early February, AI cannot be relied on whole cloth to make health insurance coverage or care decisions like denying hospital admission or downgrading a patient to an observation stay.

In other words, if you have a Medicare Advantage plan, your coverage or care cannot be denied by a computer — or at least, not by a computer alone.

How is AI being used in health care settings?

As in other industries, the presence of AI in health care settings has been steadily increasing. For physicians and other clinic officials, the technology can be used to aid in diagnostics, treatment recommendations, administrative duties and even answer patient questions. (In fact, one study found that patients actually preferred chatbot-crafted answers over those written by human physicians, rating them highly for both quality and empathy.)

In these ways, AI tools promise to lighten the load and free up valuable human work hours in a high-burnout field. For payers, too, AI can be a boon, speeding up the underwriting process and detecting insurance fraud more quickly and efficiently than human investigators alone.

But AI isn’t always right — and when medical care is on the line, even minor inaccuracies can have major effects.

The CMS memo comes on the heels of two recent lawsuits in which patients claimed their insurers — UnitedHealth and Humana, respectively — used faulty artificial intelligence algorithms to wrongfully deny elderly patients care determined by their doctors to be medically necessary.

According to allegations of the Humana case, over 90% of patient claims denied by the company’s AI model, nH Predict, were later found to be in error and reversed.

What the CMS memo means for Medicare Advantage patients

While it’s unlikely that AI will be divested entirely from its growing place in the health care industry, the CMS memo — along with an open letter issued by the House of Representatives in November 2023, asking the agency to increase its oversight on AI tools used in Medicare Advantage plans — indicates that legislators are aware of the potential downfalls and frank dangers of AI-directed health care decisions.

To that end, the memo allows that AI "can be used to assist MA plans in making coverage determinations," while putting the responsibility of ensuring those decisions are commensurate with coverage-determination rules squarely in the hands of the humans who manage the software. (The CMS specifically defines artificial intelligence, for its purposes, as "a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations or decisions influencing real or virtual environments.")

While these guidelines do not, as yet, apply to patients whose insurance coverage is not under a Medicare Advantage plan, they may be an opening foray to a broader conversation about patient rights in the world of AI — and may lead to new legislation in the same way the Affordable Care Act changed the way insurers could respond to preexisting conditions.

Time will tell how both AI technologies and the laws that govern them continue to grow and change. For now, Medicare Advantage patients can rest easy knowing that, even if AI is involved to some extent, their file’s final review will still fall to a real person.