Early Prediction of Delirium in Postcardiac Surgery Patients: Machine Learning Model Development and External...

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Critical Care Medicine

Background:

• Postoperative delirium is a common complication after cardiac surgery.
• Delirium is linked to increased hospital stays, cognitive decline, and mortality.
• Early prediction is crucial for timely interventions.

Purpose of the Study:

• Develop and validate a machine learning model for predicting delirium in cardiac surgery patients.
• Facilitate early detection of high-risk individuals in the ICU.
• Support targeted preventive strategies for delirium.

Main Methods:

• Utilized data from MIMIC-IV 2.0 and eICU-CRD databases for postoperative cardiac surgery patients.
• Developed predictive models using logistic regression, support vector classifier, extreme gradient boosting (XGB), and random forest.
• Assessed model performance using AUC, accuracy, sensitivity, PPV, NPV, and F1-score.

Main Results:

• The XGB model showed the best performance in both internal and external validation.
• External validation achieved an AUC of 0.75, indicating strong predictive ability.
• Key predictors included hospital length of stay, Glasgow Coma Scale score, and mean blood pressure.

Conclusions:

• An externally validated XGB model demonstrates strong predictive capability for ICU delirium post-cardiac surgery.
• The model can support real-time delirium alert systems.
• Enables risk stratification and evidence-based decision-making in ICUs.