The payer prior authorization process is already fairly mechanical in nature. In theory, physicians succeed in winning such battles time and again.
Still, it’s my distinct impression that such wins are more a concession in the interest of speeding things up rather than an acknowledgment that the doctor may have a point. The truth is that in the interests of payers to occasionally give in rather than make it clear how little individual clinician judgment means to them. No point in upsetting the troops too much, eh?
I was thinking about this the other day when I read an article arguing that payers are likely to use AI to handle prior authorization processes in the future. In his piece for insideBigData, Myndshift CEO and founder Ron Wince notes that not only are manual prior authorization processes frustrating for physicians, they can lead to postponed care, with 75% of providers recently surveyed by the AMA reporting that they wait up to three days for responses to prior auth requests.
Of course, if payers have put prior authorization requirements in place, they’ve done so for a reason — most likely trends they have seen emerge and claims data over time. Unfortunately, this doesn’t leave much room for frequent updates to these guidelines, as they don’t want to change their standards without sufficient evidence.
But AI could change the game. As Wince points out, given its capacity for monitoring and refining processes and making predictions to drive greater efficiencies, AI is well suited to performing prior authorization tasks.
Even better, AI-driven prior authorization processes could actually prove to be far more flexible than existing human-ones. “AI has the potential to transform prior auths into an entirely patient-get process,” Wince writes. “For example, AI can mine data from lab, medication and claims data to recommend appropriate treatments and evaluate the patient outcomes.”
In other words, while prior authorization guidelines may not be trumped up out of thin air, AI could bring sophisticated data analytics and predictive modeling to the process of addressing a patient’s needs. Rather than simply denying authorization for a particular treatment, medication or service, AI could conceivably make recommendations based on recent, relevant data rather than historical financial or outcomes trends.
Of course, for all of this to work, both payers and physicians would have to have some degree of trust in the AI’s governing algorithms. Neither side will settle for a “black box” AI tool whose results are mysterious even to those who are generating them.
Also, there still some work to do in determining which data to use, seeing to that such data is relevant and of adequate quality and make sure that these choices lead to appropriate results. Obviously, the stakes are high, so high that despite the billions of dollars that might be saved, it will probably be a while before payers or providers get comfortable automating the prior auth process.
Still, I do believe that someday payers will find an approach to automating prior authorization with AI that works for everyone involved. Hopefully, this will allow doctors to spend less time arguing with payers and more time doing their thing.