Analysis of Factors Affecting Postoperative Delirium in General Surgery: Applying Machine Learning.

Area of Science:

• Medical Informatics
• Surgical Outcomes Research
• Artificial Intelligence in Healthcare

Background:

• Postoperative delirium is a common complication in general surgery.
• Early identification and intervention are crucial for improving patient outcomes.
• Predictive models can aid in risk stratification for delirium.

Purpose of the Study:

• To develop and validate a machine learning model for predicting postoperative delirium.
• To identify key pre-surgical, surgical, and post-surgical variables associated with delirium risk.
• To assess the performance of the XGBoost model in delirium prediction.

Main Methods:

• Utilized a dataset of general surgery patients from 2017-2023.
• Employed machine learning techniques, specifically the XGBoost algorithm.
• Analyzed pre-surgical, surgical, and post-surgical data to identify risk factors.

Main Results:

• The XGBoost model achieved high predictive accuracy (0.92) and area under the receiver operating characteristic curve (AUROC) of 0.94.
• Identified significant pre-surgical, surgical, and post-surgical variables contributing to delirium risk.
• Demonstrated the model's capability for early identification of high-risk patients.

Conclusions:

• Machine learning, particularly the XGBoost model, is effective in predicting postoperative delirium.
• The developed model can facilitate early risk identification and personalized interventions.
• This approach has the potential to improve patient outcomes in general surgery.