AI cancer screening is booming, most not covered by insurance

Artificial intelligence for cancer screening has taken off.

Yet most of those new programs aren’t covered by Medicare or private insurers, which creates headwinds for companies looking to boost adoption and for patients who could benefit from the new technology.  

“Traditionally, for medical devices it takes up to seven years after a product that’s approved by the FDA to get reimbursed. So, it is quite a challenge,” said Brittany Berry-Pusey, co-founder and COO of AI screening startup Avenda Health.

As AI capabilities accelerate, the Food and Drug Administration has authorized 882 AI and machine learning-enabled devices and programs. Nearly 600 of them have been radiology AI applications approved in the last five years. Most do not yet have billing codes that would allow them to get reimbursement and prevent patients from paying out of pocket.

While some tools have shown early promise in helping to improve diagnosis and care for cancer patients, more data may be needed to determine whether they are more effective than conventional screening before major insurers will be willing to cover them.

One of Avenda’s products illustrates the complex process that has to take place before insurers cover AI tools.

The company’s Unfold AI prostate cancer platform helps urologists find more cancer cells than traditional MRI screenings. It can aid in identifying the best treatment to reduce the risk of prostate cancer surgery side effects like incontinence and impotence.

The FDA approved the program for medical decision support last year. Just as important, the American Medical Association designated a provisional billing code for it — which most AI radiology products have not yet received.

Now, Avenda is working on getting Medicare and insurers to provide coverage, which can take years in many cases.

“If there’s no payment, that means patients have to pay out of pocket, which can be challenging … especially for our patients. This is an older patient population,” said Berry-Pusey.

The American Medical Association, the medical professional organization that assigns the Current Procedural Terminology codes that allow reimbursement, issued guidelines for establishing AI CPT codes last fall. The group said different medical specialties should help determine the standards for use in their fields.    

The lack of reimbursement is hindering adoption of new AI programs for cancer screening, especially for smaller hospitals and physician practices, said Dr. William Thorwarth, CEO of the American College of Radiology, which represents thousands of professionals in the field. Yet, in a letter to a congressional committee assessing the use of AI in health care, he cautioned against moving too quickly.

Thorwarth wrote that AI reimbursement is complex and establishing billing codes for every approved AI tool is “problematic.”  He added that it is “unclear” whether the AI platforms currently covered are “adding value to patients or the health system.”

Medicare and private health insurers have expressed similar caution. A spokesperson for the Centers for Medicare & Medicaid Services told CNBC that the agency takes CPT codes into account for reimbursement and “continually assesses opportunities to leverage new, innovative strategies and technologies safely and responsibly, including Artificial Intelligence.”

Part of that caution may stem from an earlier experience with computer-aided mammography in the late 1990s. Doctors since have said it led to false positives and unnecessary biopsies.  

Independence Blue Cross Chief Medical Officer Dr. Rodrigo Cerda said the verdict is still out on the effectiveness of the newest programs.

“The evidence hasn’t quite met the bar to say it clearly makes a benefit of positive difference for our members and doesn’t introduce other risks that might be false positives or sort of gives confidence to the false negatives,” Cerda said.

Without insurance reimbursement, radiology provider RadNet has resorted to charging patients a fee for its proprietary Enhanced Breast Cancer Detection AI screening, which was launched in 2022. RadNet has published data indicating the tool helps to improve cancer detection.

The company recently cut the price of the test from $59 to $40. It said its AI digital health revenue more than doubled in the first quarter from a year ago, and patient adoption of AI screening increased from around 25% to 39% of mammogram patients.

RadNet’s executives compare the process with AI screening to the radiology industry’s experience with digital breast Tomosynthesis, known as 3D mammography. It was approved by the FDA in 2011, and women were initially offered the screening for an out-of-pocket fee. By the end of the decade, it was widely covered by insurers.

“The question is, can we eventually get the [insurers] to step up for that? And I think driving the adoption and the value propositions of finding more cancers, I think will eventually convince them,” said Dr. Greg Sorensen, RadNet’s chief science officer.  

Sorensen said RadNet has enrolled an employer in New Jersey, which will start covering the breast cancer scans for its workers.

The company will also soon launch an AI-enhanced prostate MRI screening for $250. But at that price, it may pose a bigger barrier to adoption — and access for patients who can’t afford it.

UCLA neurology professor Josh Trachtenberg was willing to pay for an AI prostate cancer screening, which he feels made a world of difference to his own care.  

Trachtenburg says when he was diagnosed with prostate cancer last year, several doctors told him he would need to have his prostate removed, a procedure that would have left him with incontinence and impotence problems.

He turned to a urologist at the UCLA medical school who was using Avenda Health’s Unfold AI program. The program more accurately measured the scope of his tumor, which allowed the doctor to get at the cancer cells in surgery while preserving healthy tissue.

Trachtenberg worries that patients who can’t afford the extra costs for certain AI tools will pay the price with poorer outcomes.

“I think most men who aren’t faculty in a medical school … are just put through the meat grinder because that’s what insurance covers and that’s the ‘go to’ procedure,” he said.

Avenda Health’s Berry-Pusey worries that patients will lose out on new technologies altogether because the uncertainty of reimbursement could stymie funding for innovation.

“As a startup, we’re always looking for investors, and so making sure that there is a clear path to revenue — it’s important for us to survive,” she said.  

Investors are backing health-care AI developers despite the payment hurdles. Alex Morgan, a partner at Khosla Ventures, is upbeat about the sector, and recently participated in large funding round for a radiology AI firm.

“If you just have a human do a bunch of activities, and then you stick AI on … you’re not getting any efficiency gains,” Morgan said, adding that the key to getting paid is to “provide differentiated, powerful outcomes.”

He said that in the end, the technology that improves the quality of care and outcomes for patients will win out.

Correction: Brittany Berry-Pusey is the COO of Avenda Health. An earlier version of this story misstated her position.