California lawmakers are joining other states in trying to restrict health insurers' use of artificial intelligence tools in coverage decisions.
According to the National Conference of State Legislatures, at least 40 states have introduced or passed AI regulation bills in 2024, including six bills specifically related to the healthcare industry. A common focus of the healthcare-specific bills is increasing oversight of insurers' use of AI tools to speed up coverage decisions.
This initiative is being implemented by major insurance companies.
California's bill, which passed the Senate with overwhelming bipartisan support and is scheduled to be presented to the state Assembly for a hearing on July 2, would require all health insurers in the state to have a licensed physician oversee the use of AI tools in the utilization review process.
Bills in New York and Pennsylvania would require insurers to disclose their use of AI to health care providers and individuals, and at least 11 states have issued broad guidelines for AI standards in health insurance, based on a model released in December by the National Association of Insurance Commissioners.
State efforts to rein in AI are growing but face constraints: It would be difficult for states to change AI systems used by insurance companies around the world, and states have no authority over employer health insurance plans covered by the federal Employee Retirement Income Security Act of 1974 (ERISA).
But lawmakers and health care analysts said in interviews that even if the federal government were to make further efforts to establish comprehensive standards for insurers' use of AI, states would still have a key role to play as primary regulators of insurance.
The introduction of the bill has drawn national attention to the issue as many state legislatures finish their sessions for the year, drawing backlash from insurance companies groups that say it would lead to unnecessary over-regulation of coverage decisions.
Lawmakers say the proposal could trigger a series of state actions to force insurers to provide more information about AI tools and how they're used to determine health insurance coverage.
“These are essentially tools that are used to make life-changing, life-altering decisions for people,” said state Assemblyman Pamela Hunter (D), the New York bill's sponsor and vice chair of the National Council of Insurance Legislatives.
“We can't wait. Consumers need to be protected now,” Hunter said.
“Setting a precedent”
Health insurers have been using AI tools for some time to “identify gaps in evidence-based care to share with providers, improve consumer experience, speed up the claims process, and detect fraud,” industry group America's Health Insurance Plans wrote in July 2023 in response to the Biden administration's request for information on national priorities for AI.
But recent lawsuits against major health insurance companies have alleged that AI tools ignore doctors' recommendations about necessary treatments, leaving many patients paying out of pocket or not receiving care at all.
One in five insured adults surveyed had experienced a denied insurance claim in the past year, according to a KFF survey conducted in September 2023. Of those whose claims were denied, 55% said they paid more than they expected for treatment or services.
“We began working on this particular bill when we learned of lawsuits, allegations and concerns that insurance companies were using algorithms to deny claims without physician oversight,” said California Sen. Josh Becker (D).
Becker's bill (SB 1120) passed the state Senate in May by a vote of 37-0, with three abstentions. The bill applies to both private and state-run health insurance plans and would “ensure that any denial, delay or modification of health care services based on medical necessity is made by a licensed physician or other health care provider qualified to evaluate the specific clinical issues in the case,” Becker said.
Becker said he believes the bill “will set a precedent that could be adopted by other states.”
In Pennsylvania, whose 2024 legislative session runs until November, state Rep. Arvind Venkat (D)'s bill (HB 1663) would require state-regulated health plans to provide the state Department of Insurance with information about the AI tools they use in the utilization review process. It would also require them to disclose their use of AI to providers and patients.
Venkat, the emergency physician, said the bill will likely be debated in the fall for a vote in the next legislative session, but he noted that his colleagues have questioned whether the bill should be broadened to address the use of AI across the health care industry.
“Do we really need a global regulatory effort around the entire healthcare system and the use of artificial intelligence? I think that's a valid point,” Venkat said.
Hunter said New York's bill (A 9149) will have to be reintroduced in the next session, as New York's 2024 legislative session ended earlier this month.
Similar bills were introduced in Georgia and Oklahoma this year, but the legislative sessions ended without a vote on the bills.
In Oklahoma, a bill from Republican state Rep. Daniel Pay would require insurers to disclose their use of AI in utilization reviews and submit algorithms and training datasets to the Oklahoma Department of Insurance.
“AI doesn't discriminate based on political stance,” Pei said, adding that there is “very strong bipartisan support” for regulating the technology.
Pei said the bill passed the Oklahoma House Government Modernization and Technology Committee in February along with other AI-related bills, but that states are in the early stages of grappling with the question of how health insurers and others might use AI, and that the issue is likely to gain more momentum next year.
A patchwork of rules
Insurance groups argue that state proposals would overly regulate insurers' decision-making, because states' limited authority could make it difficult to establish common standards on the issue.
AHIP, along with the California Health Insurance Association and the California Association of Life and Health Underwriters, oppose the current version of SB 1120. In a May 16 letter to Becker, the associations wrote that the bill “would significantly limit who can conduct utilization reviews by requiring that reviewing providers be in the same or similar specialty as the requesting provider.”
“If a treatment provider disagrees with a review provider, current law already allows for a process to resolve the dispute through an independent medical review,” the groups wrote, adding: “This provision is unnecessary and will only add additional costs to the health care system.”
There is also the issue of regulating AI within national boundaries.
While consumer protection and insurance laws are primarily state functions, AI “doesn't seem like an issue that should be regulated on a state-by-state basis because the same things are being promulgated across the country or the world,” said Michelle Mello, a professor of law and health policy at Stanford University.
Mello said this could result in a “patchwork of different rules” that health insurers would have to comply with across states.
There are also signs of movement at the federal level.
A bipartisan group of about 50 members of Congress sent a letter Tuesday to the Centers for Medicare & Medicaid Services asking it to clarify the agency's February guidelines that said Medicare Advantage plans must make prior authorization decisions based on specific individual circumstances, not solely on algorithms or software tools.
“We believe CMS should more aggressively monitor plans' use of AI and algorithm-driven tools,” the lawmakers wrote, adding that “plans continue to use AI tools to erroneously deny care and inconsistently report provider evaluation findings.”
