Abstract TP287: Use of Artificial Intelligence Driven Natural Language Processing Platform to Identify and Optimize Care for Patients with Intracranial Brain Aneurysms

Background: Intracranial aneurysms (IA) are routinely identified incidentally in patients who undergo brain imaging for unrelated pathology. Additionally, intracranial aneurysms are re-identified in patients who are lost to follow up or are unaware of their preexisting diagnosis. Herein we evaluate a software driven and nurse navigator supported Cerebral Aneurysm Detection and Surveillance Program. Methods: An artificial intelligence (AI) driven natural language processing (NLP) system (Illuminate, Overland Park, KS) was implemented on an electronic medical record (Epic, Verona, WI) at a large tertiary healthcare system and reviewed radiographic image reports over 2 years (7/1/2023-7/1/2025) looking for patients with intracranial aneurysms. Beginning 7/2024, reports were evaluated in real time. A nurse-navigator was utilized to review these predictions for accuracy and clinical relevance. The nurse navigator facilitated appropriate follow up by communicating with both patients and providers. The majority of patients were placed in a software enabled surveillance program and higher risk patients were the referred to the Neurovascular Clinic to initiate or re-establish care. Results: 309, 469 reports were reviewed over the study period yielding 3, 781 patients with an IA. Of these, 1, 886 were closed (no follow up indicated). 1, 351 were closed after initial review and 535 after a period of surveillance. 1151 patients with IA were added to the surveillance program and of these, 512 were post treatment, 173 were greater than 5 mm and 447 were less than 5 mm (19 unknown). 215 new higher risk IA patients were engaged into the health care system because of this program. The program was responsible for 259 clinic visits, 100 diagnostic imaging studies, 31 diagnostic angiograms and 21 IA open or endovascular treatments representing 883, 273 of net new patient revenue added to the system over the fiscal year. The program is tracking 783 clinic visits, 431 diagnostic imaging studies, 109 diagnostic angiograms and 82 IA open or endovascular treatments. Conclusion: The application of an AI driven NLP model with effective patient navigation assistance successfully identified (IA) requiring additional follow up and treatment. This platform allowed for identification and engagement of patients, improved operational efficiency, helped improve patient throughput and revenue in a large health care system, and shows promise for optimizing health care efficiency at scale.