An influential group that advises Congress on Medicaid is recommending increasing transparency into artificial intelligence-backed prior authorization and boosting human oversight over automated pre-approvals for care.
The recommendations come as states and the federal government say they have limited insight into payers’ use of the technology in the safety-net insurance program, which can make it challenging for regulators to monitor for data bias or inaccuracies, analysts said during the meeting.
The Medicaid and CHIP Payment and Access Commission, or MACPAC, voted Thursday on several recommendations to lawmakers, including:
- Suggesting the CMS issue guidance to state Medicaid agencies and managed care plans that requires a human with appropriate expertise to review automated care denials for medical necessity.
- Recommending the agency change regulations to ensure medical necessity denials in fee-for-service Medicaid are made by a person with a background in the enrollee’s medical, behavioral or long-term care needs.
- Recommending the CMS issue guidance to states on how they can use existing regulatory authority to oversee insurers’ use of automation in utilization management.
- Suggesting state Medicaid agencies update their contracts with health plans to require them to disclose how they’re using the technology for coverage and authorization decisions.
The recommendations could help ameliorate some of the challenges associated with AI in Medicaid prior authorizations, analysts said during MACPAC’s May meeting.
For example, with limited state or federal guidance into how AI should be used in prior authorization, some stakeholders are hesitant to adopt the technology, said Katherine Rogers, deputy director and congressional liaison at MACPAC. That means benefits from AI — like speedier decision-making or less administrative burden on providers — could be missed.
Meanwhile, states and the federal government have “limited visibility” into how managed care plans are currently using the technology in Medicaid prior authorizations, she said. Though states could use existing oversight to determine how plans are utilizing AI, not all do — creating risks of inaccurate decisions from technical problems or data errors, she added.
“Transparency and disclosure are important tools in documenting and assessing the use of automation, including the nature of emerging risks,” Rogers said. “With limited transparency, states and the federal government are less able to monitor for those potential risks posed by automation tools, including inaccurate coding, data bias and more.”
Prior authorization in the spotlight
The recommendations come as prior authorization, where providers have to receive approval from their patients’ health plans before offering certain services or drugs, has faced increased criticism from lawmakers and regulators.
Providers have long slammed the practice, arguing prior authorizations are time consuming and delay medically necessary care. Payers say the pre-approvals prevent wasteful medical care and help control ever-rising healthcare spending.
The advent of AI has created a new wrinkle in the prior authorization debate. While the technology could reach decisions more quickly and limit some of the paperwork burden on providers, critics say AI allows payers to deny claims en masse. Some patients could lose access to needed care, given few Medicaid enrollees appeal prior authorization decisions.
“It’s so rare that people appeal,” Commissioner Dennis Heaphy, a health justice advocate and researcher at the Massachusetts Disability Policy Consortium, said during the MACPAC meeting. “I think we have an obligation to ensure that we do what we can to write language that will support the least number of denials as possible.”
Meanwhile, federal oversight over AI in healthcare is limited. The Trump administration has largely taken a deregulatory approach to the technology, in a bid to speed the rollout of AI across the economy.
MACPAC will have to consider a “federal structure” on Medicaid decision-making that allows for flexibility on addressing AI, given how quickly the technology has advanced, said Commissioner Michael Nardone, a health policy consultant who previously worked at the Center for Medicaid and CHIP Services.
“Some of the things we’re talking about here, like regulations, they take time to develop,” he said. “And in the meantime, AI is barreling full speed ahead.”
