Related Experiment Video
Updated: Jun 11, 2025
12:18
A Machine Learning Approach to Design an Efficient Selective Screening of Mild Cognitive Impairment
Published on: January 11, 2020
7.5K
Development and Validation of Machine Learning Models to Predict Postoperative Delirium Using Clinical Features and
Woo-Seok Ha1, Bo-Kyu Choi1,2, Jungyeon Yeom1
1Department of Neurology, Severance Hospital, Yonsei University College of Medicine, Seoul 03722, Republic of Korea.
Journal of Clinical Medicine
|September 28, 2024
Summary
Machine learning accurately predicts postoperative delirium by integrating clinical data with polysomnography (PSG) sleep patterns. Identifying high-risk patients early using sleep factors can improve outcomes and reduce complications.
Area of Science:
- Anesthesiology
- Sleep Medicine
- Artificial Intelligence
Background:
- Postoperative delirium impacts up to 50% of patients after high-risk surgeries, leading to poor long-term prognoses.
- Identifying predictive factors for delirium is crucial for developing targeted interventions and improving patient outcomes.
Purpose of the Study:
- To predict postoperative delirium using machine learning models that integrate polysomnography (PSG) and sleep-disorder questionnaire data.
- To identify key sleep-related factors contributing to delirium risk.
- To enhance early detection and management of delirium.
Main Methods:
- A cohort of 912 adult patients undergoing surgery under general anesthesia with available PSG data were studied.
- Delirium was assessed clinically within 14 days postoperatively.
- Machine learning models, including extreme gradient boosting, were trained and tested to predict delirium, analyzing feature importance.
Main Results:
- The integrated model combining clinical and PSG data achieved an AUC of 0.84, outperforming models using only clinical data (AUC 0.81) or PSG data alone (AUC 0.60).
- Significant predictors included surgery duration, age, midazolam use, oxygen saturation nadir, periodic limb movement index, and reduced rapid eye movement (REM) episodes.
- Fewer REM episodes and increased daytime sleepiness were associated with higher delirium risk.
Conclusions:
- An AI model integrating clinical and sleep variables reliably predicts postoperative delirium.
- Sleep-related factors are significant contributors to delirium risk and its prediction.
- Early identification of high-risk patients through AI-driven analysis of sleep patterns can potentially reduce delirium-related costs and complications.
