AI can help streamline prior authorizations, but denials should be reviewed by a qualified clinician
August 4, 2023
Illustration by Dall-E
An editorial last week in the New England Journal of Medicine sums up prior authorization well. It says:
Prior authorization is one of the most enduring, infuriating, and effective tools in the United States for managing health care spending.
We hate prior authorization because it wastes the time of patients and providers, sometimes unnecessarily delays medically necessary care, and comes in between a patient and their provider. But prior authorization for non-emergency services is a bulwark against medically unnecessary medical care that could hurt patients and drives up costs. Without prior authorization, we’d face even higher medical inflation, making health insurance less affordable and further swelling out-of-pocket costs. Prior authorization and post-service claim reviews discourage low value care, and help be sure we don’t all pay for it.
Prior authorization should be standardized so the same information is required for all payers. Determinations should be rapid, and appeals should be fair and expeditious. All of this cries out for artificial intelligence (AI). The volume of medical claims is enormous, and an insurance plan can use information from demographics and diagnoses from past medical and pharmacy claims to determine whether a prior authorization should be granted quickly, or whether it needs further review. Clinical reviewers can limit their review to the small portion of prior authorizations that are far more likely to be justifiably rejected. Plans could also identify providers whose requests are rarely rejected and avoid requiring approvals from them altogether.
There are (at least) two controversies around using AI in prior authorization that may impact members of employer sponsored health plans. The first is AI-calculated lengths of stays. StatNews reported that Medicare Advantage plans used AI to determine the likely length of stay for skilled nursing care, and issued denials based on these projections. Clearly, terminations of benefits for frail elderly people in skilled nursing facilities need to be reviewed by a clinician. The second item is en masse denial of benefits. Propublica reported in March that Cigna was using its prior authorization engine, DXPX, to issue bulk denials. Cigna, responding to a lawsuit filed in July in California, has said that it was not using artificial intelligence, and stated that its streamlined approach to authorizations allowed for rapid authorizations, benefiting patients and doctors.
There is plenty of room to use artificial intelligence to identify which prior authorizations should sail through, and which patients are likely to need enough skilled care that their ongoing care in a facility need not be reviewed. But the role of human clinicians remains critical when services are being denied.
Implications
- Employers should expect their carriers to use tools like artificial intelligence to streamline prior authorization. This will give a competitive advantage to large national carriers, which have the resources to invest in AI, and the large databases on which to train their models.
For example, PBMs are currently reviewing how to leverage AI to analyze authorization request attributes with historical member, prescriber, and PA detail trends, and predict the likelihood of PA approval
- Employers can check with their carriers to be sure that denials are reviewed by clinicians before being finalized, and that members are fully informed and offered appropriate appeals rights.
- Employers can also ask for reporting on denials, appeals, and rates of overturning initial decisions. I would be cautious to request reporting on use of AI since I think it will be used quite widely.
- Carriers can increase their educational efforts around low value or out of network services so that members are not surprised when inappropriate laboratory tests are not covered by their insurance.
Hope you have a great weekend.
This weekend, I’ll be cycling the two-day, 194-mile Pan Mass Challenge to raise money for cancer research. Here’s a link if you want to know more about or sponsor my ride. I am fundraising for the Center for Global Cancer Medicine. This Center, a collaboration of the Dana Farber Cancer Institute and Partners in Health, offers free cancer care in both Haiti and Rwanda, and is focused on improving cancer care in low resource environments.
Good luck on your ride!