Medical News Observer

Key Points:

• The study addressed the challenge of low AAA detection rates by developing a machine-learning model using data from electronic medical records.
• The model, incorporating 41 clinical factors, achieved a 200% increase in AAA detection compared to traditional methods.
• This innovative approach has the potential for early diagnosis and improved patient management.

Overview

The prevalence of abdominal aortic aneurysms (AAA) and the challenges associated with their detection have been long-standing issues in vascular health management. Despite clear recommendations from various public health bodies, the current adoption rate of abdominal aortic aneurysm screening is still low. One reason is its low prevalence in the general population. 

Using a retrospective cohort study, Researchers aimed to increase the detection rate of abdominal aortic aneurysms (AAA) screenings by using machine learning to identify individuals at the highest risk. Despite low screening adoption rates and AAA prevalence, the researchers developed a machine-learning model using longitudinal medical records, focusing on patients aged 65 to 75 years. 

Machine-learning model for AAA detection

The model was validated using 6-fold cross-validation on data from 18,660 patients, identifying 314 abdominal aortic aneurysms. It incorporated 41 clinical factors from electronic medical records, including several newly identified clinical variables associated with AAA. The model outperformed standard guidelines, achieving a 200% increase in abdominal aortic aneurysm detection. 

Researchers have now integrated this novel algorithm into the institution’s workflow, enhancing targeted screening and more effectively identifying high-risk individuals. This model has the potential for early diagnosis and management and could further improve patient outcomes.

Reference

Salzler, Gregory G., Evan J. Ryer, Robert W. Abdu, Alon Lanyado Bsc, Tal Sagiv Bsc, Eran N. Choman, Abdul A. Tariq, et al. 2024. “Development and Validation of a Machine Learning Prediction Model to Improve Abdominal Aortic Aneurysm Screening.” Journal of Vascular Surgery 0 (0). https://doi.org/10.1016/j.jvs.2023.12.009.