A machine learning-based prediction model for delirium risk in malnourished elderly ICU patients with SHAP...

Area of Science:

• Geriatric Medicine
• Intensive Care Medicine
• Computational Medicine

Background:

• Delirium is a common complication in elderly ICU patients, particularly those with malnutrition.
• Early identification of delirium risk is crucial for effective management and improved patient outcomes.

Purpose of the Study:

• To identify risk factors for in-hospital delirium in malnourished elderly ICU patients.
• To develop and validate a machine learning (ML) model for early delirium risk stratification.

Main Methods:

• Utilized data from MIMIC-IV and eICU-CRD databases for model development and external validation.
• Compared multiple ML models, selecting eXtreme Gradient Boosting (XGBoost) for its superior performance.
• Interpreted the final model using SHapley Additive exPlanations (SHAP) to identify key predictors.

Main Results:

• Included 6,449 malnourished elderly ICU patients; delirium was associated with higher disease severity and worse outcomes.
• The XGBoost model demonstrated strong discrimination, calibration, and clinical utility, validated externally.
• Key predictors included SOFA score, GCS score, body temperature, SpO2, GNRI, pH, and mechanical ventilation.

Conclusions:

• Developed and validated an interpretable ML model for delirium risk prediction in malnourished elderly ICU patients.
• The model shows potential for early identification of high-risk patients.
• Facilitates targeted prevention and individualized delirium management strategies in clinical settings.