Global health systems are overwhelmed and understaffed. Hospitals have squeezed every possible efficiency out of healthcare workers, and doing administrative tasks takes up an alarmingly large percentage of providers’ time compared to that spent treating patients. The use of artificial intelligence tools—such as voice-to-text transcriptions for prescriptions and chart notes, and automated staffing schedulers—is starting to reduce the administrative burden.
Automating insurance coverage verifications and low-risk prior authorizations could be the next major step for healthcare leaders trying to innovate their way through the industry’s labor shortage. Insurance claims and prior authorizations often require healthcare workers to make phone calls, send emails and texts, and, in some cases, fax forms to insurance companies—and then wait for responses, says Sandra Carrico, Vice President of Machine Learning for Sorcero, a life sciences language intelligence platform focused on patient outcomes, increased productivity, and regulatory monitoring.
Healthcare leaders who want to make strides with AI need to get comfortable with moving more quickly than they’re used to—and the strong, established boundaries between providers and insurance companies will have to be broken.
On average, a manual prior authorization takes an average of 21 minutes of staff time—and a single authorization can take up to 45 minutes, according to a 2021 report from the Council for Affordable Quality Healthcare (CAQH), a not-for-profit alliance of health plans, providers, government agencies, and standard-setting bodies. “If we could automate prior authorization, we could get a much better use of our highly trained healthcare professionals,” says Carrico. “From a productivity point of view, it’s a simple win.”
It’s also a clear financial win. Insurance companies and other payers such as Medicare could save as much as $437 million per year by automating prior authorizations, according to the aforementioned CAQH report. The group estimates that the healthcare industry at large could reduce its administrative costs by $13.3 billion annually by fully employing automation.
Breaking Down Walls
One of the main obstacles to automating prior authorizations is the siloing of data between providers and insurance companies, says Carrico, who was the principal AI architect at Anthem during the period when the health insurance giant transformed from a traditional insurance company to a coverage and care platform built on data, AI, and machine learning.
The American Medical Association has been calling for a transition to automated and streamlined prior authorizations since at least 2018. Data uniformity and transparency from all levels of the healthcare system are key reforms principles prescribed in the organization’s guiding documents. Still, prior authorization is one of the least active areas for implementing advanced computing by healthcare organizations, according to the CAQH.
Unfortunately, the healthcare industry moves very slowly, notes Kyle Kotowick, a solutions architect with a doctorate in human systems integration and the founder of technological systems consultancy Invicton Labs. “They take the same approach to IT as they do to new medical techniques and procedures: carefully test it in a controlled environment, ensure it doesn’t do any harm, and then consider adopting it if the benefits outweigh the costs,” he says.
But healthcare leaders who want to make strides with AI need to get comfortable with moving more quickly than they’re used to—and the strong, established boundaries between providers and insurance companies will have to be broken. “If you’re going to solve these problems, you’re going to have to break the rules,” says Carrico.
Bringing in Top Talent
Competing for talent is an ongoing challenge for healthcare leaders trying to implement AI. Traditional healthcare companies have always had trouble competing for the talent needed to make these leaps into advanced patient care, says Raj Vishnu, the Senior Client Partner for Healthcare and Life Sciences at Toptal, who placed skilled tech freelancers who placed skilled tech freelancers at a Fortune 25 healthcare firm when the company developed its center for AI and made other digital innovations.
Now that large tech firms such as Google and AWS have moved into the healthcare space, it’s even harder for healthcare companies to lock down full-time tech workers with advanced skills, he says.
“Over the last few years, technology has been changing exponentially but the talent market is improving on a linear scale, which means the gap is widening on a day-to-day basis,” says Vishnu. “So it’s absolutely critical that healthcare companies have a robust talent model to effectively handle the future state.”
Using contingent talent teams with the capabilities to build advanced computer models is one option. It can be more cost-effective than competing to hire scarce full-time tech workers, he says, and it also keeps the teams nimble and able to pivot and innovate as needed. Firms will also need access to operational workers—known as machine learning operations, or ML Ops—who will be tasked with running automation on a day-to-day basis, says Carrico.
Hiring people to do ML Ops may be even more challenging than hiring teams to build the systems. “The tools aren’t mature yet, people don’t understand what questions to ask, the patterns aren’t well established, and it’s not broadly understood,” she notes.
Moving Forward With AI in Healthcare
The large firm Toptal worked with now uses AI for customer service, billing, care management, and adjudicating claims. And in a virtual shareholders’ meeting in 2021, the company’s chief digital officer said that the company expects to automate 50% of its work over the next few years.
It’s time for other forward-thinking health insurance companies to step up and lead this charge. “All of the information funnels through the insurance company because they are the payer,” says Vishnu. “At the healthcare company we worked with, Toptal talent ended up building their entire AI platform, which is conscious of all the information that they get: the claims information, the clinical data, the demographic information, the wearable data, all of that—and it forms what is called a data lake.”
It is on this deep flow of data that minor prior authorization decisions can be made, among other automated decisions. “It’s a win-win-win for the insurance company and the patient, as well as for the doctor’s office,” he says. “There will be hesitancy in some of the adoption just because of the nature of the existing relationships, but those are all barriers that will be pushed over. The value that the patient is going to see is so huge that the system will adapt—and it will be equally beneficial to all the players.”