Prior authorization has ripple effects on patients and clinicians, but artificial intelligence (AI) has the potential to simplify the process, health policy experts said during an hosted by the Kaiser Family Foundation on Thursday.
Recently the backlash against prior authorization requirements has been growing. "Even when appropriate, prior authorization creates delays to care, and that can [worsen] outcomes, and can affect things like cancer survival," said Fumiko Chino, MD, a radiation oncologist at Memorial Sloan Kettering Cancer Center in New York City.
One in three physicians blamed the process for resulting in a serious adverse event, such as hospitalizations, permanent impairment, or even death, according to a 2023 survey from the American Medical Association. Even just one day of delay can potentially mean uncontrolled pain for a patient, said Chino. For patients with cervical cancer, each day of delay "equates to a 1% decreased local control rate," she said. "If you have a 5-day delay, that's a 5% decreased local control rate."
Anna Schwamlein Howard, JD, principal for policy development at the American Cancer Society Cancer Action Network (ACS/CAN) in Washington, D.C., said delays in care can have financial costs as well. A patient waiting days for a pain medication to be approved may wind up in the emergency room, which increases costs for both the patient and the payer, an outcome Howard argued is "penny-wise and pound foolish."
But Troyen Brennan, MD, a former executive at CVS Care and an adjunct professor at the Harvard T.H. Chan School of Public Health in Boston, defended the process, arguing that it cuts down on unnecessary care. About 15% to 30% of all care in the U.S. healthcare system is ineffective, Brennan said.
Plus, there are "really not any good studies ... showing actual harm," he argued. "There are a lot of surveys from physicians, in particular, that say that there are tremendous delays, but there's obviously a response bias associated with this."
Also, prior authorization is "fairly heavily regulated," Brennan added. If private health plans don't meet certain requirements -- 7 days to a decision for standard requests, and 14 days for expedited -- they can be fined by the Centers for Medicare & Medicaid Services (CMS) or the Department of Labor.
Estimating costs for both the provider and the insurer, the return on investment of such policies is about 10 to 1, he said. "Does it seem like a reasonable thing for the insurance [company] to continue to do? It does," he said.
However, due to criticism from legislators, patients, and physicians, he's seeing a "retreat back away from utilization management" -- another term for prior authorization -- in cases where there isn't the same return on investment, and a sharpened focus on more "non-controversial" requests, Brennan said.
In January, CMS issued a final rule aiming to streamline the prior authorization process which called on Medicare Advantage plans, state Medicaid and Children's Health Insurance Program (CHIP) fee-for-service programs, Medicaid managed care plans, and CHIP managed care entities to file prior authorization decisions within 72 hours for expedited requests and 7 calendar days for standard requests. The rule also mandates that insurers must publicly report their prior authorization metrics. Prescription drugs are excluded from the rule.
Howard, speaking for ACS/CAN, said her organization "would like to see those timelines sharpened" to 72 hours for non-expedited requests, and 24 hours for expedited requests.
Brennan went further, suggesting that "this whole process could be completely automated" through computer programs that would take a prior authorization request and "would basically interrogate the electronic medical record and come back with a decision immediately."
He noted that in dozens of places throughout the final rule, CMS called on the Office of the National Coordinator for Health Information Technology (ONC) to "do your part" in calling for changes to the electronic medical record.
Asked whether AI could improve prior authorization and prevent delays, panelists were cautiously optimistic. "I wouldn't be afraid of it, sort of ex ante, but you would want to make sure you've got complete transparency," Brennan said, adding that ultimately these decisions come down to a discussion between a physician like Chino and another radiation oncologist. (One large insurer, UnitedHealth, was by two families who claimed that the insurer used a faulty AI algorithm to deny necessary care to their since-deceased elderly family members.)
Chino said she welcomed AI with "some caveats," noting that marginalized populations are more likely to be "missing key elements" in their electronic medical records. "And then you've trained a machine based on a data set that is essentially racist." In an email to app, Chino explained that an AI prior authorization system might "inadvertently flag charts for denials" based on "flawed datasets," which might then determine that "other patients" -- including those coded as high income and white -- are "low risk," and speed up those approvals.
"So if one group gets faster approvals, it can perpetuate inequity and disparate care," she said.