Predicting Postoperative Circulatory Complications in Older Patients: A Machine Learning Approach

Area of Science:

• Geriatric Medicine
• Anesthesiology
• Data Science

Background:

• Postoperative circulatory complications (PCCs) pose a significant risk for elderly patients undergoing major non-cardiac surgery.
• Accurate prediction of PCCs is crucial for timely intervention and improved patient outcomes.

Purpose of the Study:

• To utilize machine learning algorithms to identify key determinants for predicting PCCs in elderly patients.
• To develop and validate a predictive algorithm for PCCs in this demographic.

Main Methods:

• Secondary analysis of data from a randomized controlled trial involving 1,720 elderly participants (60-90 years).
• Analysis of 67 candidate variables, including baseline characteristics, laboratory tests, and scale assessments.
• Application of machine learning, specifically a Random Forest model, for feature selection and prediction.

Main Results:

• Identified significant predictors of PCCs: ICU and surgery duration, APACHE-II score, intraoperative heart rate, blood loss, opioid use, patient age, VAS-Move-Median score, Charlson comorbidity score, fluid volumes, red blood cell transfusion, and endotracheal intubation duration.
• The Random Forest model achieved a high accuracy of 0.9872 in predicting PCCs.
• Key determinants were effectively identified through feature selection.

Conclusions:

• An effective algorithm for predicting PCCs in elderly patients has been developed and validated.
• Identification of key risk factors allows for targeted risk stratification and management strategies.
• Machine learning offers a powerful tool for predicting complex postoperative complications in geriatric populations.