Machine learning for prediction of in-hospital mortality in lung cancer patients admitted to intensive care unit

Abstract

Backgrounds The in-hospital mortality in lung cancer patients admitted to intensive care unit (ICU) is extremely high. This study intended to adopt machine learning algorithm models to predict in-hospital mortality of critically ill lung cancer for providing relative information in clinical decision-making. Methods Data were extracted from the Medical Information Mart for Intensive Care-IV (MIMIC-IV) for a training cohort and data extracted from the Medical Information Mart for eICU Collaborative Research Database (eICU-CRD) database for a validation cohort. Logistic regression, random forest, decision tree, light gradient boosting machine (LightGBM), eXtreme gradient boosting (XGBoost), and an ensemble (random forest+LightGBM+XGBoost) model were used for prediction of in-hospital mortality and important feature extraction. The AUC (area under receiver operating curve), accuracy, F1 score and recall were used to evaluate the predictive performance of each model. Shapley Additive exPlanations (SHAP) values were calculated to evaluate feature importance of each feature. Results Overall, there were 653 (24.8%) in-hospital mortality in the training cohort, and 523 (21.7%) in-hospital mortality in the validation cohort. Among the six machine learning models, the ensemble model achieved the best performance. The top 5 most influential features were the sequential organ failure assessment (SOFA) score, albumin, the oxford acute severity of illness score (OASIS) score, anion gap and bilirubin in random forest and XGBoost model. The SHAP summary plot was used to illustrate the positive or negative effects of the top 15 features attributed to the XGBoost model. Conclusion The ensemble model performed best and might be applied to forecast in-hospital mortality of critically ill lung cancer patients, and the SOFA score was the most important feature in all models. These results might offer valuable and significant reference for ICU clinicians’ decision-making in advance.

Machine learning links T cell function and spatial localization to neoadjuvant immunotherapy and clinical outcome in pancreatic cancer

“…Finally, SHAP values were used to interpret cellular biology underpinning model predictions (43). SHAP values denote the relative importance of a given feature in driving a model’s prediction – this method has been used to explain machine learning predictions in prior cancer studies (38, 61, 62).…”

Interpretable Mortality Prediction Model for ICU Patients with Pneumonia: Using Shapley Additive Explanation Method

“…In future research, incorporating real-world data and continuous model updates could enable the development of a dynamic prediction tool, responsive to evolving patient conditions and treatment protocols. Collaboration with clinicians is crucial for re ning the model and ensuring its seamless integration into clinical work ows [52][53][54].…”

Neural networks based on attention architecture are robust to data missingness for early predicting hospital mortality in intensive care unit patients

“…XGB was widely applied in previous researches aiming to early predict hospital mortality for ICU patients and showed improved predictive performance over other ML models. 5 – 8 , 10 – 13 As mentioned before, XGB owned a built-in mechanism to handle missing data, which made it competent for our dataset. For optimizing hyperparameters of XGB, we performed a grid search on different combinations of the following hyperparameter settings: n_estimators (400, 600, 800), learning_rate (0.01, 0.05, 0.1), colsample_bytree (0.6, 0.8), subsample (0.4, 0.6, 0.8), max_depth (4, 6, 8), min_child_weight (1.0, 2.0), gamma (0.2, 0.4), and determined the optimal setting to achieve the highest average AUROC in the 5-fold cross-validation on the training set.…”

“…The most frequently applied ML algorithms for early prediction of hospital mortality include classification and regression tree (CART). 5 – 7 Naive Bayes model, 5 , 8 , 9 support vector machine, 5 , 8 random forest, 5 – 11 extreme gradient boosting (XGB) 5 – 8 , 10 – 13 and artificial neural network. 7 , 11 , 14 Compared to conventional severity scores, ML models have more sophisticated algorithm for mining data pattern and show improved predictive performance.…”