Machine Learning Algorithm Using Electronic Chart-Derived Data to Predict Delirium After Elderly Hip Fracture...

Area of Science:

• Artificial Intelligence in Medicine
• Geriatric Surgery
• Anesthesiology

Background:

• Elderly patients undergoing hip fracture repair are highly susceptible to postoperative delirium.
• Existing delirium identification methods lack accuracy and require additional patient assessments.
• Electronic health records (EHRs) offer a potential data source for predictive modeling.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting postoperative delirium in elderly hip fracture patients.
• To identify significant risk factors contributing to delirium development using ML.
• To assess the accuracy and potential clinical utility of ML-based delirium prediction.

Main Methods:

• Retrospective case-control study of 245 elderly patients (≥65 years) undergoing hip fracture repair.
• Data collected from anesthesia records and medical charts, including demographics, surgical, and anesthetic variables, and frailty index.
• Four ML models (Random Forest, XGBoosting, SVM, MLP) were trained and validated using K-fold cross-validation to predict delirium incidence (assessed via CAM).

Main Results:

• Postoperative delirium occurred in 12.2% of patients.
• ML models achieved prediction accuracies ranging from 83.67% to 87.75%.
• Key identified risk factors included dementia/history of stroke, blood transfusion, preparation time, frailty index, vasopressor use, surgery duration, and anesthesia type.

Conclusions:

• Hospital-specific ML models derived from EHR data can effectively predict delirium risk in elderly hip fracture patients.
• ML models quantify the contribution of various risk factors, offering insights into delirium pathophysiology.
• Further validation is necessary to establish the clinical applicability of these ML models for delirium detection.