Interpretable early mortality prediction in oncology ICU patients: A dual-cohort validation of a LASSO–XGBoost–SHAP framework

1

The increasing number of critically ill cancer patients in ICUs necessitates effective early risk stratification within the first 0–24 hours of admission.
2

Traditional severity scoring systems often fail to accurately assess risk in cancer ICU patients due to their unique pathophysiology and treatment-related factors.
3

Machine learning models have shown promise in improving mortality prediction in oncology ICU populations by capturing complex, high-dimensional data.
4

External validation and calibration of machine learning models are crucial for their bedside utility and acceptance in clinical decision-making.
5

Open-access databases like MIMIC and eICU provide valuable resources for developing and validating mortality risk prediction models in cancer ICUs.

Digital Health

by Xinyi Chen, Lu Wang, Wan Qin, Mu Yang, Yuanmei Yan, Xiaoxiao Luo, Xianglin Yuan
July 7, 2026

1