Predicting delirium in intensive care unit patients every 8 hours with machine learning: Model development and...

Area of Science:

• Critical Care Medicine
• Artificial Intelligence in Healthcare
• Nursing Informatics

Background:

• Delirium in the intensive care unit (ICU) is linked to adverse patient outcomes.
• Predicting ICU delirium risk remains challenging for healthcare providers.
• Early identification is crucial for implementing timely interventions.

Purpose of the Study:

• To develop and evaluate a machine learning (ML) model for predicting delirium occurrence in ICU patients.
• To identify key variables for delirium prediction using routinely collected data.
• To enable proactive nursing interventions for delirium prevention.

Main Methods:

• Retrospective analysis of electronic medical records from a mixed ICU.
• Inclusion of adult patients admitted between January and December 2023, with ICU stays >24 hours.
• Development of prediction models using XGBoost, LightGBM, CatBoost, and random forest algorithms.
• Delirium defined by an Intensive Care Delirium Screening Checklist score ≥4.

Main Results:

• The study included 273 patients; 62.3% experienced delirium.
• CatBoost model achieved an Area Under the Curve (AUC) of 0.886, with high precision (0.804) and accuracy (0.816).
• Routinely collected nursing variables, including the Intensive Care Delirium Screening Checklist and Glasgow Coma Scale scores, were significant predictors.

Conclusions:

• The developed ML model shows promise for identifying patients at high risk of ICU delirium.
• The model utilizes readily available nursing variables, facilitating practical clinical application.
• Further validation with larger, diverse populations is recommended to guide nursing interventions and optimize patient care.