An AI-assisted approach to lung cancer screening (LCS) can significantly improve patient uptake of low-dose CT (LDCT) exams and help detect cancer earlier, according to research presented at the annual meeting of the American Society of Clinical Oncology (ASCO) in Chicago.
Using a centralized patient navigation model along with AI, Peoria, IL-based OSF Healthcare System, a health system with 17 hospitals in both rural and urban areas, was able to lift its LCS rate for eligible patients from 18% to 42% over a five-year period. What’s more, the interventions led to an increase in detection of earlier-stage cancers.
“A multifaceted approach combining centralized navigation and innovative screening technologies can further improve screening uptake,” said first author and presenter Jun Zhang, MD, PhD, of the OSF HealthCare Cancer Institute in Peoria.
Although screening with LDCT has decreased lung cancer mortality in high-risk individuals, screening uptake in the U.S. remains at approximately 20% -- far short of the 40% to 50% threshold likely needed for meaningful population-level mortality reduction, according to Zhang.
Barriers to LDCT screening include primary physician uncertainty about eligibility criteria, limited patient awareness, geographic and insurance-related access challenges, and organizational fragmentation that disconnects screening from diagnosis and treatment, Zhang said.
To help, the health system wanted to see if an AI-assisted multifaceted approach could aid in screening uptake. They hypothesized that a phased, AI-integrated health system intervention that included AI-assisted electronic health record (EHR) alerts for identifying high-risk eligible patients, centralized navigator coordination, quality dashboard integration, radiology workflow automation, and incorporation of blood-based screening could together achieve high and sustainable screening rates while improving stage at diagnosis, Zhang said.
These measures were implemented over time. In October 2019, dedicated patient navigators were consolidated into a centralized team that manages screening registries, patient scheduling, and care coordination. Next, a radiology natural language processing (NLP) software engine was deployed in mid-2021 to automate reading-based actions from radiology reports using the Lung-RADS classification system. Later that year, AI-assisted best-practice alerts were added for their EHR software.
In early 2022, system-wide adoption of the 2021 expanded U.S. Preventive Services Task Force (USPSTF) lung cancer screening guidelines was implemented. A quality dashboard displaying screening metrics was then integrated mid-year into the health system’s Ambulatory Quality Dashboard. The final intervention consisted of a blood-based ctDNA screening pilot targeting non-compliant and high-risk populations.
These interventions yielded significant improvements in screening rates.
Lung cancer screening rates at OSF after interventions | ||||
| OSF annual LDCT volume | National screening rate* | OSF screening rate | Gap in percentage points | |
| 2020 | 2,223 | 14.5% | 18.2% | 3.7% |
| 2021 | 2,647 | 14.8% | 21.4% | 6.6% |
| 2022 | 3,280 | 15.5% | 27.8% | 12.3% |
| 2023 | 3,406 | 16% | 33.6% | 17.6% |
| 2024 | 3,522 | 18.2% | 38.2% | 20% |
| 2025 | 4,108 | 19.5% | 42.8% | 23.3% |
*Benchmarks from the American Lung Association
Zhang noted that these improvements were also resilient during COVID. The program also experienced stage migration, in which detection of stage I cancers increased and stage IV cancer detections decreased. These results also outpaced national averages.
Furthermore, benefits were particularly prominent in rural areas; some rural centers saw an over 21% increase in stage I diagnoses over the six-year period, according to the researchers.
“A multifaceted hub-and-spoke model can dramatically benefit cancer screening in rural areas,” Zhang said.
For more of AuntMinnie’s coverage of ASCO 2026, click here.
