Developing a machine learning-based prediction model for postinduction hypotension.

Area of Science:

• Anesthesiology
• Medical Informatics
• Cardiovascular Medicine

Background:

• Arterial hypotension during general anesthesia leads to significant postoperative complications like kidney injury, myocardial injury, and stroke.
• Postinduction hypotension (PIH) is complex, influenced by patient factors, medications, and anesthetic choices, challenging traditional prediction models.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for predicting PIH.
• To identify key clinical predictors of PIH using ML.

Main Methods:

• Retrospective cohort study of 20,309 adult patients undergoing non-obstetric surgery with intravenous induction.
• ML model developed using k-fold cross-validation; performance evaluated by AUC; feature importance assessed using SHAP values.
• PIH defined as mean arterial pressure (MAP) < 55 mmHg within 10 minutes post-induction.

Main Results:

• PIH occurred in 24.4% of patients (4,948/20,309).
• Key predictors identified: preinduction systolic and mean arterial pressures, propofol dose, and beta-blocker use.
• The ML model achieved an AUC of 0.732 for PIH prediction.

Conclusions:

• The ML-based model shows significant predictive capability for PIH.
• This model can enhance preoperative planning and patient risk stratification.
• Prospective studies are needed for further validation.