Prediction of Postinduction Hypotension by Machine Learning.

Area of Science:

• Anesthesiology
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Post-induction hypotension (PIH) is a common complication following anesthesia induction.
• PIH is linked to adverse post-operative outcomes and influenced by anesthetic agents and clinical factors.
• Current anesthetic dosing strategies require enhancement for predicting and mitigating PIH risk.

Purpose of the Study:

• To develop and validate a machine learning model for predicting the risk of post-induction hypotension.
• To support clinical decision-making regarding anesthetic dosing, specifically propofol.
• To provide personalized anesthetic safety recommendations.

Main Methods:

• Utilized a dataset of 320 cases from the VitalDB database.
• Incorporated demographic data, vital signs, and medication dosages (including propofol) as input features.
• Employed nested cross-validation for robust model performance evaluation.

Main Results:

• Achieved high predictive performance for PIH risk (precision 0.83, recall 0.84).
• The model successfully predicted PIH risk prior to anesthetic induction.
• An advisory model was developed to suggest personalized safe propofol dosage ranges.

Conclusions:

• Machine learning models can effectively predict PIH risk based on pre-induction data.
• Propofol dosage is a critical factor in predicting PIH, enabling risk assessment before administration.
• The developed model and advisory system offer potential for personalized and safer anesthetic management.