Volume 11,Issue 5
An Interpretable XGBoost Model with LASSO Feature Selection for Predicting In-Hospital Mortality in Patients with Sepsis
Background: Sepsis remains a leading cause of in-hospital death in intensive care units (ICUs). This study developed and internally validated an interpretable machine-learning model for predicting in-hospital mortality in patients with sepsis. Methods: This retrospective cohort study used the Medical Information Mart for Intensive Care IV (MIMIC-IV, version 2.2) database. Adult patients meeting the Sepsis-3 criteria were included. Candidate predictors were extracted from the first 24 hours of ICU admission. The least absolute shrinkage and selection operator (LASSO) regression was used for feature selection. Extreme gradient boosting (XGBoost), logistic regression, and a decision tree were trained on a 70% training set and evaluated on a 30% held-out test set by the area under the receiver operating characteristic curve (AUC). SHapley Additive exPlanations (SHAP) were used to interpret the final model. Results: A total of 4, 262 patients with sepsis were included; the in-hospital mortality rate was 19.9%. LASSO retained 15 predictors. The XGBoost model achieved the highest discrimination (AUC 0.86; 95% confidence interval 0.84–0.88), outperforming logistic regression (0.80) and the decision tree (0.78). SHAP analysis identified lactate, mean arterial pressure, age, SOFA score, and PaO2 as the dominant predictors, with risk rising steeply at lactate > 2 mmol/L and mean arterial pressure < 65 mmHg. Conclusions: An interpretable XGBoost model based on routinely collected early ICU variables accurately predicted in-hospital mortality in sepsis and may support early risk stratification.
[1] Singer M, Deutschman C, Seymour C, et al., 2016, The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA, 315(8): 801–810.
[2] Rudd K, Johnson S, Agesa K, et al., 2020, Global, Regional, and National Sepsis Incidence and Mortality, 1990–2017: Analysis for the Global Burden of Disease Study. Lancet, 395(10219): 200–211.
[3] Evans L, Rhodes A, Alhazzani W, et al., 2021, Surviving Sepsis Campaign: International Guidelines for Management of Sepsis and Septic Shock 2021. Intensive Care Medicine, 47(11): 1181–1247.
[4] Seymour C, Gesten F, Prescott H, et al., 2017, Time to Treatment and Mortality During Mandated Emergency Care for Sepsis. New England Journal of Medicine, 376(23): 2235–2244.
[5] Vincent J, Moreno R, Takala J, et al., 1996, The SOFA (Sepsis-Related Organ Failure Assessment) Score to Describe Organ Dysfunction/Failure. Intensive Care Medicine, 22(7): 707–710.
[6] Knaus W, Draper E, Wagner D, et al., 1985, APACHE II: A Severity of Disease Classification System. Critical Care Medicine, 13(10): 818–829.
[7] Gao J, Lu Y, Ashrafi N, et al., 2024, Prediction of Sepsis Mortality in ICU Patients Using Machine Learning Methods. BMC Medical Informatics and Decision Making, 24(1): 154.
[8] Chen S, Fan J, Pishgar E, et al., 2025, XGBoost-Based Prediction of ICU Mortality in Sepsis-Associated Acute Kidney Injury Patients Using MIMIC-IV Database with Validation from eICU Database. medRxiv, 2025.02.24.25322816.
[9] Lundberg S, Lee S, 2017, A Unified Approach to Interpreting Model Predictions. Advances in Neural Information Processing Systems, 30: 4765–4774.
[10] Johnson A, Bulgarelli L, Shen L, et al., 2023, MIMIC-IV, a Freely Accessible Electronic Health Record Dataset. Scientific Data, 10(1): 1.
[11] Collins G, Reitsma J, Altman D, et al., 2015, Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD): The TRIPOD Statement. BMJ, 350: g7594.
[12] van Buuren S, Groothuis-Oudshoorn K, 2011, mice: Multivariate Imputation by Chained Equations in R. Journal of Statistical Software, 45(3): 1–67.
[13] Tibshirani R, 1996, Regression Shrinkage and Selection via the Lasso. Journal of the Royal Statistical Society Series B, 58(1): 267–288.
[14] Chen T, Guestrin C, 2016, XGBoost: A Scalable Tree Boosting System. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 785–794.
[15] Zabihi M, Kiranyaz S, Gabbouj M, 2019, Sepsis Prediction in Intensive Care Unit Using Ensemble of XGBoost Models. Computing in Cardiology, 45: 1–4.
[16] Bakker J, Nijsten M, Jansen T, 2013, Clinical Use of Lactate Monitoring in Critically Ill Patients. Annals of Intensive Care, 3: 12.
[17] Moons K, Kengne A, Grobbee D, et al., 2012, Risk Prediction Models: II. External Validation, Model Updating, and Impact Assessment. Heart, 98(9): 691–698.