Providers lean on AI startups to limit, challenge insurance denials

As more insurance companies use automation tools in coverage decisions, some providers and patients are turning to artificial intelligence to speed up their own approval-request processes.

Several vendors have added capabilities to offer providers relevant medical and insurance policy information in an instant. A flurry of startups have also launched, promising clinicians and patients more seamless methods of limiting and challenging insurance denials.

Related: Generative AI costs adding up for providers

“At the end of the day, it's just about evening the playing field,” said Dr. Michael Gao, co-founder and CEO of AI startup SmarterDx.

Health insurance companies have spent the past several years investing in technologies, including some powered by artificial intelligence, to reduce administrative burdens throughout the claims process for their staff and providers. Insurers, including Humana and CVS Health’s subsidiary Aetna, continue searching for additional ways to automate the process.

Patients and providers have grown critical of some insurers’ alleged use of automation. The American Hospital Association said in a September report automation by insurers contributed to an increase in the number of commercial and Medicare Advantage claims denials between 2022 and 2023. UnitedHealth Group and Cigna are contending with lawsuits alleging they use AI tools and algorithms to deny care. The companies have said their tools are not used to make coverage decisions and, instead, are used to inform providers of care patients might need and accelerate payment.

Health plans are working to protect patients from the full impact of rising costs and connect them with care, the health insurance trade group AHIP said in a statement in response to requests for comment about how its members are incorporating AI and automation.

“In the fragmented and heavily regulated healthcare system, health plans, providers and drugmakers share a responsibility to make high-quality care as affordable as possible and easier to navigate for the people we collectively serve,” AHIP said.

Providers say their problems can be multi-pronged. In June, the American Medical Association said more than 80% of its members don't always appeal refusals because many don’t think they’ll be successful, have limited time, or need to provide urgent care.

Startups are responding by creating tools geared toward helping providers more seamlessly navigate the utilization management process, and the interest they’ve received over the past year may indicate this use of automation could be the new normal, said Dr. Bill Fera, principal at Deloitte Consulting.

“In the past, it was just really hard to get to all of that information because it's in different places, and it's not always apparent and easy to get to. But now, with generative AI, it is,” Fera said.

Evolving insurance policies and nuances baked into self-insured employer plans create additional challenges for providers attempting to determine whether prior authorization for care is needed and the grounds for why a denial should be overturned.

“There's a lot of opacity in the problem, and so what we try to do is highlight what the medical necessity rules are, and then help them submit higher quality prior authorization so they get higher success rates for approval,” said Dr. Jeremy Friese, CEO of Humata Health. Friese founded the company in 2023.

About 40,000 physicians at health systems and medical groups are using Humata's platform to assist with prior authorization in various specialties, according to the company. Humata has helped boost provider approval rates from the 80s to the upper 90s, Friese said. The company also works with payers to automate approvals.

Startups are also developing large language models that are trained using plan policies and clinical data to craft appeal letters.

Take SmarterDx. In October, the startup rolled out a tool that generates appeal letters in a fraction of the time for providers. The company charges customers based on how much money its tools recover, and is currently focused on justifying inpatient hospital admissions and whether the diagnoses associated are valid.

Three health systems are currently using the tool, Gao said, which allows providers to upload a denial letter to a secure website, identifies supporting evidence from the patient record, feeds that data into a large language model, and converts that information into a letter. A clinician may move a sentence or make a minor edit before submitting to the insurer, he said.

Its tools don’t always mean more money for providers.

“We are about the truth, so occasionally, we'll find something that actually lowers payment because it's actually an error the other way around, where there's something there that shouldn't be there, and we represent that to our customers as well,” Gao said.

In addition, some providers have begun building their own AI-powered tools in-house. For example, New York University Langone Medical Center created a team that’s looking for use cases for AI and has used generative AI to improve the claims process. The total cost of ownership long-term is likely lower for providers building this technology versus buying it from vendors, but not all providers have the resources to do this themselves, Fera said.

Patients are also fighting back on denials with AI. In October, the startup Claimable launched its platform to ease the appeal process for consumers. People using its platform spend about 30 minutes answering a questionnaire and then the software generates an appeal letter that includes their personal story and supporting medical information. Patients pay a $39.95 flat fee no matter the size of the claim and regardless of whether their denials get overturned, said Dr. Warris Bokhari, co-founder and CEO of Claimable.

These vendors have reported a surge in interest over the last 18 months as AI and large language models gain steam among providers.

“It makes the doctors’ lives easier, and of course there’s a lot of reimbursement and revenue integrity that goes along with it,” said Nate Wilson, president and cofounder of Regard. The company works with more than 100 hospitals and uses data in the electronic health record to recommend diagnoses with the aim of increasing approval rates.

But industry leaders caution providers could be walking a fine line incorporating automation on their side of the utilization management process given the heat on insurers. The products of generative AI can also contain errors, leading to additional manual work later.

Automation isn't the absolute answer to fixing an overall cumbersome process, said AMA Immediate Past President Dr. Jesse Ehrenfeld. The Centers for Medicare and Medicaid Services' final rule requiring government-sponsored health plans to respond to requests more quickly, provide reasons for denying care and publicly share that data is a step in the right direction, he said. Health insurers, including UnitedHealth Group and Cigna, have also taken steps to cut some prior authorization requirements.

Claimable's Bokhari agrees the system isn't perfect.

“We're in this to solve a problem,” he said. “There needs to be change in this country, and until someone comes up with a solution, this might be the solution.”

Nona Tepper contributed.