Machine Learning Model Based on Baroreflex Sensitivity for Predicting Post-Induction Hypotension in Elderly Patients

Overview

The purpose of this study is to develop a high-performance machine learning model combining dynamic baroreflex sensitivity (BRS) metrics and multi-dimensional static clinical features to predict the risk of post-induction hypotension (PIH) in elderly patients undergoing elective non-cardiac surgery under general anesthesia.

Aging significantly alters cardiovascular autonomic function, characterized by elevated sympathetic and decreased parasympathetic tone, rendering elderly patients highly vulnerable to post-induction hypotension (PIH). While existing machine learning models heavily rely on static data (e.g., baseline blood pressure, demographics, medication history), they lack real-time dynamic regulatory inputs, limiting their predictive performance in individualized care. This single-center, prospective cohort study aims to bridge this gap by introducing preoperative BRS parameters-calculated via the continuous non-invasive arterial pressure (CNAP) method-into machine learning frameworks. A total of 500 patients aged over 65 years scheduled for elective non-cardiac surgery will be enrolled. Preoperative data, including autonomic indices, frailty assessments, and static clinical factors, will be mapped alongside intraoperative events and 30-day postoperative complications. Multiple machine learning algorithms (Logistic Regression, Random Forest, GBDT, XGBoost, LightGBM, and LSTM) will be leveraged and optimized using cross-validation to construct a robust clinical decision-support pipeline.

Conditions

Eligibility

Locations (1)

Outcomes

Central Contacts