Predicting Short-Term Mortality after Endovascular Aortic Repair Using Machine Learning-Based Decision Tree Analysis

Toshiya Nishibe; Tsuyoshi Iwasa; Masaki Kano; Shinobu Akiyama; Toru Iwahashi; Shoji Fukuda; Jun Koizumi; Masayasu Nishibe

doi:10.1016/j.avsg.2024.10.009

Predicting Short-Term Mortality after Endovascular Aortic Repair Using Machine Learning-Based Decision Tree Analysis

Ann Vasc Surg. 2024 Nov 22:111:170-175. doi: 10.1016/j.avsg.2024.10.009. Online ahead of print.

Authors

Toshiya Nishibe¹, Tsuyoshi Iwasa², Masaki Kano³, Shinobu Akiyama³, Toru Iwahashi³, Shoji Fukuda³, Jun Koizumi⁴, Masayasu Nishibe⁵

Affiliations

¹ Department of Medical Informatics and Management, Hokkaido Information University, Ebetsu, Japan; Department of Cardiovascular Surgery, Tokyo Medical University, Tokyo, Japan. Electronic address: toshiyanishibe@yahoo.co.jp.
² Department of Medical Informatics and Management, Hokkaido Information University, Ebetsu, Japan.
³ Department of Cardiovascular Surgery, Tokyo Medical University, Tokyo, Japan.
⁴ Department of Radiology, Chiba University School of Medicine, Chiba, Japan.
⁵ Department of Surgery, Eniwa Midorino Clinic, Eniwa, Japan.

PMID: 39580030
DOI: 10.1016/j.avsg.2024.10.009

Abstract

Background: Endovascular aneurysm repair (EVAR) has revolutionized the treatment of abdominal aortic aneurysms by offering a less invasive alternative to open surgery. Understanding the factors that influence patient outcomes, particularly for high-risk patients, is crucial. The aim of this study was to determine whether machine learning (ML)-based decision tree analysis (DTA), a subset of artificial intelligence, could predict patient outcomes by identifying complex patterns in data.

Methods: This study analyzed 169 patients who underwent EVAR to identify predictors of short-term mortality (within 3 years) using DTA. Data included 23 variables such as age, gender, nutritional status, comorbidities, and surgical details. The Python 3.7 was used as the programming language, and the scikit-learn toolkit was used to complete the derivation and verification of the decision tree classifier.

Results: DTA identified poor nutritional status as the most significant predictor, followed by chronic kidney disease, chronic obstructive pulmonary disease, and advanced age (octogenarian). The decision tree identified 6 terminal nodes with a risk of short-term mortality ranging from 0% to 79.9%. This model had 68.7% accuracy, 65.7% specificity, and 79.0% sensitivity.

Conclusions: ML-based DTA is promising in predicting short-term mortality after EVAR, highlighting the need for comprehensive preoperative assessment and individualized management strategies.