Read the full-text of this article for free

Title: An explicable machine learning approach for predicting 30-day septic mortality for ICU patients

Authors: Liang Zhou; Ruiqian Wu; Shanshan Wang; Temitope Emmanuel Komolafe; Jiachen Guo; Zhiping Fan; Zhiwei Zhang

Addresses: Shanghai University of Medicine and Health Sciences, Shanghai, 201318, China ' University of Shanghai for Science and Technology, Shanghai, 200093, China ' Shanghai Jiao Tong University, Shanghai, 200030, China ' Shanghai University of Medicine and Health Sciences, Shanghai, 201318, China ' Shanghai University of Medicine and Health Sciences, Shanghai, 201318, China ' Shanghai University of Medicine and Health Sciences, Shanghai, 201318, China ' Shanghai Chest Hospital, Shanghai, 200030, China

Abstract: This study developed MorSNX, a clinician-friendly model combining neural networks and XGBoost, to predict 30-day mortality risk in ICU patients with sepsis using vital signs and clinical data from the MIMIC-IV database. The top 25 predictive features were identified through backward stepwise regression, and SHAP values and decision curve analysis (DCA) enhanced interpretability. Validation with the eICU database demonstrated superior performance (AUC 0.9563), with significant clinical utility across decision thresholds, outperforming traditional models and scores, particularly at a 0.4 probability threshold. MorSNX offers a robust, interpretable tool for sepsis prognostication in critical care.

Keywords: sepsis; mortality risk; machine learning; XGboost; clinical decision support.

DOI: 10.1504/IJICT.2025.145831

International Journal of Information and Communication Technology, 2025 Vol.26 No.9, pp.1 - 22

Received: 06 Sep 2024
Accepted: 31 Oct 2024
Published online: 28 Apr 2025 *