Oncology Experts Debate the Potential for AI to Improve Timely Access to Care

The process of submitting insurance claims and appealing coverage denials related to oncology treatments has been growing increasingly complex as treatment paradigms expand. Incorporating AI tools into daily workflows may allow oncology professionals to save time contending with insurance denials and re-allocate attentions back toward face-to-face patient care. However, larger issues are also at play that indicate the need for insurance companies, legislators, and oncologists to collaboratively determine a balance optimal resource utilization and prioritizing patient outcomes.

OncLive gathered insights from the following experts about the challenges of prior authorization, the potential use of AI tools for claims management, and the need for structural re-evaluation of the insurance claims process within oncology:

• Kate Baker, MD, MMHC, medical director of value-based care at Tennessee Oncology in Lebanon and Gallatin
• Samyukta Mullangi, MD, MBA, a medical oncologist at Tennessee Oncology in Dickson
• Youngmin Kwon, PhD, a postdoctoral fellow at the Vanderbilt University Medical Center in Nashville, Tennessee

OncLive: ⁠In your practice, approximately how much time would you say you and/or your team spend appealing insurance denials related to patient care?

Mullangi: I practice in a large community oncology practice where we have a robust and well-staffed revenue cycle management [RCM] group. They have prior authorization coordinators and people who work with physicians to appeal denials. [They] also manage a lot of the financial considerations, like assisting patients with getting connected to grants, copay assistance programs, etc. This offloads a lot of burden for me, but in general, it's a labor-intensive and exceedingly manual process. It creates a lot of delays.

If a patient has an urgent, rapidly growing cancer presentation, and I need to start them rapidly on treatment, I often have to accept the possibility that I won't be reimbursed for that because I'm initiating a treatment plan maybe the next day or within the same week. For the most part, I try to budget in patients at least starting 1 week out, sometimes even 2 weeks out, to allow the whole process of submitting the prior authorization, potentially navigating an initial denial, and submitting the re-appeal that the team is doing on my behalf. However, before I can get a patient started on treatment, I often need them to see another physician, or I need to get an additional scan, or I need them to get a port. Sometimes it's okay that we're pushing out the start of treatment by a week or 2, but sometimes it's not clinically defensible. We're doing it because we need the time to navigate the prior authorization and the eventual risk of insurance denial. It's not a perfect system, and it needs a lot of work.

Have you seen instances where insurance denials delay patient care? How would you say that affects the overall outcomes of those patients?

Baker: I worked with the Tennessee Oncology Practice Society to get legislation passed in a past legislative session to prohibit step therapy protocols for patients with metastatic cancer, for both supportive care medications, as well as anti-oncolytics.1 As part of that process for preparing to meet with legislators around that bill, we've collected a few patient stories where FDA-approved and guideline-concordant care was ordered by physicians for patients with metastatic, incurable cancer. In one case, a patient who had metastatic colon cancer [was] on a third line of treatment. [Patients in that circumstance do not] have time to wait for insurance denials to go through. For a patient with third-line metastatic colon cancer, even a week makes a big difference. That patient went on to experience worsening symptoms and ended up passing away before she could receive therapy, just due to prior authorization and step edit denials. We see this frequently.

Thankfully, the whole team of people that we have [helps] us to get medicines through that process. But the process takes time. It's at least days, and sometimes, if there are appeals, [it takes] weeks. If there are outright denials, that's potentially months that the patient's lost, especially with oral oncolytic drugs.

How could AI tools help appeal insurance denials?

Mullangi: AI tools have a lot of promise, because so much of submitting prior authorization requests involves manual abstraction. I'm putting together a note, I'm entering some codes, but then somebody on the RCM team is combing through what I've documented and extracting from there to fill out a form that can be a PDF or a paper copy. It's not an automated, seamless process. [This is] a truly administrative burden. [It would be helpful to have] the ability to potentially upload and have natural language processing automatically pull in all the justifications I've already documented in my note, as well as create a more seamless process [that potentially runs the note] against logic that the insurance company is holding into place for approvals or denials. [It would also be helpful to receive guidance saying]: since you are ordering this test, would you consider adding this diagnosis to support that? An automated clinical decision support around that would be helpful.

However, as a first step, before [exploring AI as a] solution, it would be good to revisit the whole concept and ask: Is this [process] meaningfully deterring excessive waste or fraud in health care? Are the risks that this system is imposing worth the benefits of cutting down this purported waste?

Kwon: The concern is that a lot of times these decisions are not being made on the basis of clinical efficacy, but just purely [because] a doctor forgot to attach documentation, or a patient forgot to attach a critical piece of evidence that demonstrates evidence of clinical necessity [for a treatment]. It's important to acknowledge that, although in a lot of cases it may seem like we're making decisions off clinical factors, in a lot of other cases, administrative mishaps are falling through the cracks.

Baker: These processes are iterative and repetitive. If you're submitting prior authorization letters for the same medicine for many patients, that's a repetitive process. One thing we've considered as a practice, [which] we're not using quite yet but a lot of practices are also looking into, is using AI to help make that process quicker, particularly when it's a repetitive process.

AI is good at that. We're still at the stage where a human would need to double check the work being done by AI. But that could take a lot off the plates of the humans and allow us to be a bit more efficient in getting those prior authorizations through.

What larger-scale changes should also be considered when aiming to improve the insurance coverage process for patients with cancer?

Kwon: In the absence of more structural solutions, we're ultimately imposing these burdens onto the patients and providers. They're all already dealing with a lot of issues related to burnout, whether it’s COVID-19 or other factors. Doctors are having to operate in a system that is fundamentally bureaucratized.

We can say to the doctors: Next time you appeal for a letter, maybe include better evidence, more documentation, etc. But that ultimately puts a lot of pressure on the doctors to have to do all that. Some of the ASCO statements on prior authorization show that doctors, on average, are filling out approximately 40 to 50 prior authorizations per week.2 Sometimes the time spent on filling out prior authorizations is almost equivalent to the full-time load of a physician. It’s tough that providers are the ones who have to do all this to advocate on behalf of patients.

By attempting to improve the ease of this system, we're offloading a larger conversation that we need to have about the role of prior authorization to begin with. Rather than trying to perfect the process itself, maybe we need to have a conversation on cancer care. It's an inelastic service, meaning that regardless of how costly [cancer treatments] are, [patients are] still going to need them.

If the demand for treatment is so high that regardless of how you price it, there's a demand for these services, perhaps we need to have a separate conversation about whether prior authorization is the right tool to improve care in oncology. Maybe it's more applicable in elective procedures, but in the context of oncology, there is something uncomfortable about the idea that someone other than a patient’s doctor is making decisions for their care. This isn't to say that there aren't examples of low-value care in oncology; that does happen. However, prior authorization might not be the correct tool to deal with that. We need to have more, larger societal conversations about what we are willing to fund and reimburse, rather than subjecting these decisions to the adjudicator at the insurer. They may not even be a real person; it could be AI making these decisions.

References

1. HB 0850. Tennessee General Assembly. Updated May 27, 2025. Accessed October 6, 2025. https://wapp.capitol.tn.gov/apps/Billinfo/default.aspx?BillNumber=HB0858&ga=114
2. ASCO Prior Authorization Survey Summary. ASCO. Accessed October 6, 2025. https://cdn.bfldr.com/KOIHB2Q3/as/wt2thjbxbj24ncwnwjwzvk9c/2022-ASCO-Prior-Auth-Survey-Summary