A Machine Learning-Based Algorithm for the Prediction of Intensive Care Unit Delirium (PRIDE): Retrospective Study

Area of Science:

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

Background:

• Delirium is a common complication in intensive care units (ICUs).
• Limited evidence exists for treating established ICU delirium.
• Early recognition and prevention are crucial for managing critically ill patients.

Purpose of the Study:

• To develop and validate a machine learning model for predicting ICU delirium within 24 hours of admission.
• Utilize electronic health record (EHR) data for early delirium prediction.
• Name the developed algorithm the Prediction of ICU Delirium (PRIDE).

Main Methods:

• Retrospective cohort study using EHR data from a tertiary referral hospital.
• Included adult patients admitted to medical or surgical ICUs.
• Developed and validated prediction models (RF, XGBoost, DNN, LR) using internal and external datasets (MIMIC-III).

Main Results:

• The Random Forest (RF) model achieved the highest Area Under the Receiver Operating Characteristic (AUROC) curve for internal validation (0.916) and external validation (0.721).
• The RF model demonstrated good calibration with a Brier score of 0.168.
• Machine learning models effectively predicted delirium using EHR data.

Conclusions:

• Machine learning models can predict ICU delirium within 24 hours of admission using EHR data.
• The PRIDE algorithm shows potential for assisting ICU physicians in delirium prevention.
• Prospective studies are needed to confirm the algorithm's clinical utility and performance.