Improving Follow-Up of Incidental Pulmonary Nodules in the Emergency Department Using an Artificial Intelligence-Supported Workflow.
Authors
Affiliations (9)
Affiliations (9)
- Department of Medicine, Sub-Department of Pulmonary and Critical Care, University of Maryland Baltimore Washington Medical Center, 301 Hospital Drive, Second Floor, Glen Burnie, MD 21061, United States. Electronic address: [email protected].
- Touro College of Osteopathic Medicine, 230 W 125th St, New York, NY 10027, United States.
- Department of Medicine, Sub-Department of Pulmonary and Critical Care, University of Maryland Baltimore Washington Medical Center, 301 Hospital Drive, Second Floor, Glen Burnie, MD 21061, United States.
- Department of Clinical Research, University of Maryland Baltimore Washington Medical Center, 305 Hospital Drive, Third Floor, Glen Burnie, MD 21061, United States.
- Department of Thoracic Surgery, University of Maryland Baltimore Washington Medical Center, 305 Hospital Drive, Third Floor, Glen Burnie, MD 21061, United States.
- Population Health and Data Analytics, University of Maryland Baltimore Washington Medical Center, 301 Hospital Drive, Glen Burnie, MD 21061, United States.
- Department of Medicine, Division of Pulmonary, Critical Care & Sleep Medicine, University of Maryland School of Medicine, 655 W. Baltimore Street Baltimore, MD 21201.
- Department of Radiology, University of Maryland Baltimore Washington Medical Center, 301 Hospital Drive, Glen Burnie, MD 21061, United States.
- Department of Radiation Oncology, University of Maryland School of Medicine, 655 W. Baltimore Street Baltimore, MD 21201.
Abstract
Incidental pulmonary nodules (IPNs) are frequently identified on computed tomography (CT) scans but are often associated with poor rates of patient notification and follow-up, limiting opportunities for early lung cancer detection. To evaluate whether implementation of an artificial intelligence (AI)-supported workflow improves patient notification and follow-up of IPNs detected in the emergency department (ED). We conducted a retrospective pre-post cohort study at an academic-affiliated community hospital. The pre-intervention cohort included ED patients undergoing chest CT between January and March 2023. The post-intervention cohort included ED patients undergoing chest CT between June and August 2025, in which an AI-based natural language processing system identified potential IPNs from radiology reports, and patients were contacted to facilitate follow-up. Primary outcomes were rates of patient notification about their IPN and nodule-specific follow-up. With implementation of the AI-supported workflow, patient notification increased from 171/228 (75%) to 223/252 (88.4%) p = 0.0001, and nodule-specific follow-up increased from 122/228 (53.5%) to 171/252 (67.9%) p = 0.0012. Inability to reach patients by phone after their ED visit was identified as a significant barrier to follow-up. There was no significant difference in lung cancer stage at diagnosis between cohorts. An AI-supported IPN identification and outreach workflow improved patient notification and follow-up. Proactive communication strategies, facilitated by AI, represent a feasible approach to addressing care gaps and enhancing early lung cancer detection.