What question did this study set out to answer?

The study aimed to develop an explainable machine learning model to predict multidrug-resistant organism (MDRO) infection risk by combining clinical and environmental factors.

June 5, 2026Open Access

Explainable Machine Learning Integrating Patient and Environmental Factors for Predicting Multidrug-Resistant Organism Colonization or Infection on ICU Admission

Key Points

The study aimed to develop an explainable machine learning model to predict multidrug-resistant organism (MDRO) infection risk by combining clinical and environmental factors.
Analyzed data from 420 ICU patients, split into 210 MDRO-positive cases and 210 controls.
Used LASSO regression to select predictors and developed six machine learning models.
Evaluated models and interpreted the best-performing one (XGBoost) using SHAP analysis.
Five significant predictors identified, including long-term care residence and prior MDRO-positive status.
XGBoost model achieved an AUC of 0.926 for training and 0.862 for validation sets.
A web-based tool for real-time risk assessment was developed.

Abstract

Objective: Multidrug-resistant organisms (MDROs) pose a serious threat to global public health, particularly in intensive care units (ICU). Few studies have employed machine learning (ML) to capture complex clinical interactions. This study aimed to develop an explainable ML model for early risk stratification of MDRO colonization or infection by integrating patient-specific clinical features with environmental exposure factors. Methods: We analyzed the data of 420 ICU patients (210 MDRO-positive cases and 210 matched controls) admitted between January 2020 and October 2023. Predictors were selected using least absolute shrinkage and selection operator (LASSO) regression. Six ML models—Logistic Regression, Random Forest, Gradient Boosting, AdaBoost, XGBoost, and LightGBM—were developed and evaluated using internal validation on a randomly split test set. The best performing model was interpreted using SHapley Additive exPlanations (SHAP), and a web-based tool was developed for clinical applications. Results: Five predictors were identified through LASSO regression and were independently associated with the composite endpoint in subsequent multivariable logistic regression, including residence in a long-term care facility, MDRO-positive status of the prior bed occupant, central venous catheterization, surgery prior to infection, and duration of arterial catheterization. The XGBoost model demonstrated the highest performance, with an area under the curve of 0.926 for the training set and 0.862 for the validation set. SHAP analysis improved interpretability by quantifying feature contributions and illustrating the rationale behind individual predictions. A web-based tool was developed to facilitate real-time clinical risk assessment. Conclusion: This study demonstrates the utility of integrating environmental risk factors into a ML framework for improved MDRO prediction, resulting in a web-based tool with the potential for clinical decision support and enhancing infection control workflows. Keywords: machine learning, ICU, MDRO, prediction model, SHapley Additive exPlanations

Ask AI

Helpful

Bookmark

View Full Paper