Use of AI Lets Health System Find Lung Cancer at Early Stages

For Amie J. Miller, MSN, APRN, AOCNP, ACHPN, CTTS, November always brings opportunity, and a little frustration.

She was excited at how many attendees at the Association of Cancer Care Centers’ (ACCC) 51st Annual Meeting & Cancer Center Business Summit knew that November was Lung Cancer Awareness Month. But she admitted it's tough following the annual pink out to promote Breast Cancer Awareness Month that comes every October.

Lung cancer, she notes, kills more Americans than breast, colon, or prostate cancer combined.

Amie J. Miller, MSN, APRN, AOCNP, ACHPN, CTTS | Image: LinkedIn

Miller shared this observation at Thursday’s ACCC session on how artificial intelligence (AI) and business intelligence solutions can help health systems make practice improvements. She runs an AI-fueled effort that saves lives and brings more than $1 million a year to Sarasota Memorial Health Care System in southwest Florida.

At Sarasota’s Brian D. Jellison Cancer Institute, Miller is coordinator of the Lung Cancer Early Detection and Prevention Program, which features an initiative to follow up on incidental pulmonary nodules (IPNs) detected through AI. IPNs are asymptomatic lesions that may be detected on a CT scan during an emergency room visit or other intervention; these could signal early lung cancer but often go unaddressed.

Recent therapeutic advances in the treatment of lung cancer mean survival rates are vastly improved—if cancer is caught early. However, as a 2024 report from the American Lung Association showed, disparities are widening in biomarker testing and lung cancer screening, meaning that that whether one dies of lung cancer may depend on access. Miller said national lung cancer screening rates remain stubbornly low at 18.1%, although there are pockets of the country—such as the catchment area for St. Elizabeth’s in Kentucky—where about 45% of the eligible population is screened.

As Miller explained, understanding who is eligible for screening is key. Recommendations from the US Preventive Services Task Force (USPSTF) call for screening those aged 50 to 80 years who are current smokers or who have quit within the past 15 years, after smoking 20 years or more. “When we look at the data, only about 42% of lung cancer patients qualified for lung cancer screening,” she said. “So, that’s a real issue.”

Thus, 44% of lung cancers are diagnosed at a late stage, where survival rates are significantly lower if cancer is caught at stage I or II. That’s where an IPN program can make a difference, Miller said. According to a 2022 study by Penn Medicine, about 10% of patients with an IPN of more than 8 mm will receiving a lung cancer diagnosis.

So, who are these patients?

“Incidental pulmonary nodules are a different animal than lung cancer screening,” Miller said. “With cancer screening, you have people who are well insured; they're well connected to their primary care doctor, they're educated, and they're advocating for themselves. Whereas, with IPNs, these patients are oftentimes within the emergency room for their health care, oftentimes they don't have a primary care physician, oftentimes they are uninsured.”

Miller took the audience step by step through Sarasota’s journey with its AI-driven lung cancer screening and IPN initiative:

• Sarasota Memorial Healthcare System invested in AI-driven software in 2019, which boosted its lung cancer screening program and allowed the launch of an IPN program designed to catch cancers at earlier stages.
• The AI technology reads close to 150,000 CT scans that come in across the 1100-bed system each year. A radiologist will accurately report what is on the scan; almost always, patients are not receiving a scan for a lung nodule.
• Previously, the patients would discharge to a primary care physician who would be unaware of the IPN; a patient with very early stage lung cancer might return several years later with late-stage lung disease. Miller said, “This is something that is happening in all of your emergency rooms today. I guarantee it.”
• Under the IPN program, the AI flags IPNs that merit follow-up, and information on these patients is forwarded for follow-up. A communication tool kits was developed to help reach out to patients to educate them on what the CT scan showed and next steps.
• After integrating AI, Sarasota’s annual screening volume rose by 99%, from 676 patients in 2019 to 1342 patients in 2023. Screening rates surpass national benchmarks for early detection, with 70% of lung cancer screening and 44% of incidental pulmonary nodule cancer cases diagnosed at stages I and II.

Miller emphasized that the AI program has more than paid for itself: she wrote that the health care system produced a return rate of return of more than 85% for high-risk lung cancer patients, with an average of 91% from 2019 through Q2 2024, compared with the national average of 22.3%.

According to the session abstract, “The AI solution helped generate $8,321,128 in downstream charges and $803,106 in contribution margin from 1702 cases and $5,559,125 in charges and $349,121 in contribution margin from 275 cases, from the lung cancer screening and incidental pulmonary nodule programs, respectively.”

A key to the AI program’s success is that it doesn’t interrupt the workflow; yet it tracks the patients and allows IPNs to be caught in a timely manner. Each day, the AI program scans CT scans and pulls a what Miller calls a “work list” of patients who need follow-up. “And I guarantee you, every single day I call a patient and say, ‘By the way, you were in the in the emergency room last week having your CT scan done because you had a kidney stone—you actually have a spot on your lung and it measures 10 millimeters.’ And I can’t tell you how many times patients [say], ‘I had no idea.’”

There’s now a weekly meeting among a team of pulmonologists and a thoracic surgeon to discuss high-risk cases that emerge from the AI-driven caseload; Miller said they cover 10 cases in 30 minutes, and over time, they are building awareness about IPNs.

“There is ongoing community physician outreach,” she said. “This is something that will generate increases in our screening numbers…. That’s something I talk about every day—how can we increase our screening volumes, by engaging physicians and empowering and educating the communities?”