Date of Award

January 2026

Document Type

Open Access Thesis

Degree Name

Medical Doctor (MD)

Department

Medicine

First Advisor

Mridu Gulati

Abstract

Background: Low dose computed tomography for lung cancer screening frequently identifies incidental findings. The most frequent and consequential thoracic and upper abdominal incidental findings include interstitial lung disease (ILD), interstitial lung abnormalities (ILA), emphysema, coronary artery calcification, aortic calcification, aortic aneurysm, and adrenal lesions. Many of these abnormalities carry prognostic and therapeutic implications. However, the prevalence of these findings, variation in reporting practices, and downstream management within integrated health systems remain incompletely characterized, particularly for ILA and ILD.

Research Aims: We aimed to quantify the prevalence of selected incidental findings on lung cancer screening CT across a large multi-hospital health system and to evaluate demographic and institutional variations in their reporting. Additionally, we assessed the factors associated with the assignment of the Lung-RADS category S modifier for clinically significant incidental findings. Finally, for the subset of individuals with ILA and ILD, we determined the frequency and independent predictors of pulmonary referral and PFT orders within 6 months of screening.

Hypothesis: We hypothesized that incidental findings are common in lung cancer screening CT, that substantial inter-institutional variability exists in reporting practices, and that patients with ILD and ILA experience limited downstream pulmonary evaluation despite radiographic detection.

Methods: We conducted a retrospective study of adults (≥18 years) undergoing lung cancer screening CT between January 1, 2022 and May 31, 2025 at seven delivery networks sites within the Yale New Haven Health System. A rule-based natural language processing (NLP) algorithm identified seven common incidental findings from radiology reports. Algorithm performance was validated through manual review of all positive cases and 375 randomly selected negative reports. The prevalence of incidental findings and Lung-RADS category were reported. Predictors of category S designation, pulmonology referral within 6 months, and PFT orders within 6 months were identified. Multivariable logistic regression adjusted for age, gender, race, delivery network, insurance status, chronic obstructive pulmonary disease (COPD), and incidental findings.

Results: 11283 patients had at least one lung cancer screening CT during the study period. The mean age was 65.9 years and 48.2% were female. Emphysema was present in 68.1%, coronary artery calcification in 57.0%, aortic calcification in 17.8%, ILA in 3.5%, adrenal mass/nodule in 2.6%, aortic aneurysm in 2.3%, and ILD in 2.0%. The Lung-RADS category S modifier was applied in 12.2% of all scans and was most commonly applied for aortic aneurysm (40.4%), ILA (34.0%), and ILD (29.6%). Category S designation was more strongly associated with delivery network than incidental findings. Within 6 months, pulmonology referrals significantly increased in ILD patients (19.9%; p<0.001) but not ILA (7.8%; p=0.062) patients compared to neither group (5.5%). PFT orders within 6 months was significantly increased in ILD (15.9%; p<0.001) but not ILA (6.5%; p=0.520) patients compared to neither group. The strongest predictors of referrals were COPD (aOR 4.34; 95% CI 3.58-5.27; p<0.001) and ILD (aOR 3.58; 95% CI 2.51- 5.10; p<0.001).

Conclusion: Incidental findings are common on lung cancer screening CT, with marked institutional variation in reporting practices and downstream management. ILD detection increases pulmonary referral and testing, however, no difference was observed for ILA. Overall, 80.1% of patients with ILD findings and 92.2% of patients with ILA findings were not referred within 6 months. These findings highlight opportunities to standardize reporting, improve care pathways for ILA and ILD, and implement system-level interventions, including structured nomenclature and decision support tools to optimize follow-up in lung cancer screening programs.

Comments

This is an Open Access Thesis.

Download
Open Access

This Article is Open Access

Share

COinS