Improving mortality prediction in critically ill cancer patients with a multidimensional machine learning model

BackgroundPrognostic assessment in critically ill patients with cancer remains challenging, as conventional ICU severity scores often perform suboptimally in this population. Machine learning (ML) approaches may improve outcome prediction by integrating acute physiology, organ dysfunction, and oncologic variables. We aimed to develop and validate ML-based models to predict ICU mortality and 30-day survival in critically ill cancer patients. MethodsWe conducted a retrospective cohort study including 997 critically ill cancer patients admitted to the ICU. Forty-eight demographic, oncologic, physiological, laboratory, and therapeutic variables collected at ICU admission were used to train and validate ML models. Eight algorithms were evaluated using stratified cross-validation with feature selection and hyperparameter optimization. Model performance was assessed using discrimination, calibration, and classification metrics. Model interpretability was explored using Shapley additive explanations (SHAP). ResultsCatBoost achieved the best performance for ICU mortality prediction (AUROC 0.96), showing excellent discrimination and calibration, and outperforming other ML models. Prediction of 30-day survival was less accurate (best AUROC 0.75), reflecting the influence of post-ICU factors not captured at admission. Key predictors of ICU mortality included severity of organ dysfunction, therapeutic objectives, vasopressor and methylene blue use, SAPS III score, lactate, platelet count, and blood urea nitrogen. For 30-day survival, baseline physiological status, admission type, SAPS III, lactate, creatinine, age, and body mass index were most relevant. SHAP analysis demonstrated that acute physiology and organ dysfunction, rather than cancer diagnosis alone, primarily drove short-term outcomes. ConclusionsML-based models, particularly CatBoost, outperformed traditional prognostic tools for predicting ICU mortality in critically ill cancer patients. Cancer was not an independent predictor of short-term mortality; outcomes were primarily driven by pre-ICU conditions, acute physiology, and severity of organ dysfunction. External validation is needed to confirm generalizability and support future integration of ML-based prediction tools into clinical decision-making in oncologic critical care.