Utilizing natural language processing to improve abdominal aortic aneurysm surveillance.

Abstract

Screening and surveillance are essential to prevent aneurysm rupture. We used natural language processing (NLP) software to evaluate efficacy of aneurysm surveillance. NLP software was used to review 7 years of imaging reports at a single institution to identify patients with an abdominal aortic aneurysm. After identification of a patient cohort, review of the electronic medical record was completed to collect patient demographics and information about compliance with nationally recommended aneurysm surveillance protocols. NLP identified a cohort of 1424 patients with an abdominal aortic aneurysm; 1105 (77.6%) were male and the mean age was 74 ± 10 years. Of these patients, 76% were White, 6% were Black, 7% were Asian, 1% was Native American or Pacific Islander, and 11% identified as another race. At the time of data collection, 552 patients (39%) were under active surveillance, 346 patients (24%) had previously been under surveillance and subsequently lost to follow-up, and 523 patients (37%) had incidental findings without initiation of surveillance. In total, more than one-half of the cohort (869 patients [61%]) were not participating in recommended aneurysm surveillance. Findings did not differ significantly by gender, race, or ethnicity. In the cohort, 39% of patients were compliant with recommended surveillance. Incorporating NLP into traditional screening and surveillance practices can allow providers to identify patients outside of recommended surveillance, bridging gaps in care.