This startup says its AI can speed insurance approval of health treatments

In recent years, doctors and patients have reported serious frustration with requirements from insurers for prior authorization before a laundry list of medical treatments and procedures can take place.

In a June survey by the American Medical Association, 90% of physicians reported that the sometimes arduous process negatively affects outcomes for their patients, and nearly one in four doctors said prior authorizations “led to a serious adverse effect” for at least one of their patients. A widely reported study by the health policy research nonprofit KFF found the practice widespread among Medicare Advantage insurers, with an average of 1.7 prior authorization requests submitted per Medicare Advantage enrollee in 2022—a total of 46 million requests.

Prior authorization requirements can also be time-consuming for doctors and medical facilities, which need to document their patients’ need for specific care. Medical providers sometimes must engage in lengthy fax and phone correspondence with insurers before treatment begins, taking time and resources away from actual medical care.

Research has shown that the vast majority of prior authorization requests—including more than 92% of those examined in the KFF report—are ultimately approved by insurers. But these still come at a cost for patients who experience treatment delays, doctors who need to work with health plans to see treatment authorized, and insurers themselves, who typically employ nurses and doctors to review those treatment requests.

A startup called Anterior looks to make the process smoother, deploying artificial intelligence technology to help frontline employees at insurance companies review prior authorization applications. Anterior recently successfully concluded case studies with large organizations that collectively cover roughly 50 million people, with extended deals likely to be announced in the coming months, cofounder and CEO Abdel Mahmoud says. In June, the company announced it had completed a $20 million Series A funding round. 

Mahmoud emphasizes that Anterior’s software doesn’t ever reject an authorization request. Rather, it’s intended to process doctors’ requests and corresponding insurance company requirements to guide reviewers—often nurses—through the steps, reducing the need for medical professionals to spend time at providers’ offices and insurance companies handling prior authorization of care.

“Why are we having to have a nurse sit down working on a laptop reviewing 10 faxes a day?” Mahmoud says he and his colleagues asked upon reviewing the existing system. “And can we do something about that to try and reduce that burden?”

The AI software, named Florence after pioneering nurse Florence Nightingale, can comb through hundreds of pages of medical documentation for a particular patient and procedure. It highlights where insurers’ requirements to demonstrate the treatment is medically necessary are satisfied and any areas where documentation may be spotty. The AI is designed to show its work, and the user interface lets reviewers toggle between Florence’s analysis and specific sections of relevant documents cited by the AI, Mahmoud says. 

“Nurses and doctors are very diligent—they’re going to check everything,” he says. “If something feels off to you, you can check.”

Mahmoud says some insurers have said they’re particularly interested in a tool that can speed up approvals of easy cases. If information is lacking, the AI can help request additional information from the doctor necessary to approve treatment. The software can in some cases help save days or even weeks of administrative processing, getting patients treated quicker, he says. 

Future versions may even be able to contact medical offices by phone, conducting automated conversations with providers to get more information, Mahmoud says. Meanwhile, there are companies working on the other side of the problem, creating tools to assist medical practices in filing prior authorization requests. 

“PAs [prior authorizations] becoming more efficient should mean that they become invisible,” Mahmoud writes in an email to Fast Company. “Just like when you swipe your credit card, you don’t think about the multiple processes happening in the background.”

Meanwhile, outright denials are handled outside the Florence system, and are typically issued only after a second-tier review by a physician, not just a nurse. “AI should not do denials,” Mahmoud says. “We don’t automate denials.”

Still, algorithmic and AI-powered tools for making medical coverage decisions have so far earned a mixed reputation. A recent article by three legal scholars in Health Affairs Forefront arguing for greater regulatory oversight of such software pointed to a 2022 McKinsey study that found AI could automate 50% to 75% of “manual tasks” in prior authorization and specific examples of insurers using such technology to speed the authorization process. But it also cited a Stat News series and class-action lawsuit claiming seniors in UnitedHealthcare Medicare Advantage plans were denied needed rehabilitative care thanks to the use of an algorithm with a high error rate. UnitedHealthcare has denied those allegations.

“It’s not that using AI to process prior authorization or utilization management or claims coverage reviews is inherently ethical or unethical,” says Carmel Shachar, an assistant clinical professor of law at Harvard Law School and one of the authors of the Forefront report. “But if you don’t implement it right, it can have really negative repercussions.”

Another risk, Shachar says, is that streamlining prior authorization processing may encourage insurers to seek preapproval for an even wider range of treatments. 

“We can see that prior authorization denials and claims coverage denials have skyrocketed in the last several years, which is also the time period in which AI has become more and more incorporated,” she says. “When they have an AI program doing that initial pass, it becomes cheaper and cheaper for them to run a PA process, which means they’re more likely to require PA for lower and lower cost services.”

So far, Mahmoud says, he hasn’t seen insurers using the technology to expand the list of treatments requiring prior authorization. With other companies working on AI to aid medical providers in filing requests, it’s possible prior authorization could become a speedier and less painful process. And, he says, Anterior may benefit from potential regulations that require efficient processing of authorization requests or mandate greater transparency about how authorization requests are handled.

“The problem with payers is not that they don’t want to do that,” Mahmoud says. “It’s just that they are struggling in admin hell.”

https://www.fastcompany.com/91255586/this-startup-says-its-ai-can-speed-insurance-approval-of-health-treatments?partner=rss&utm_source=rss&utm_medium=feed&utm_campaign=rss+fastcompany&utm_content=rss