Development of a prediction model for hypotension after induction of anesthesia using machine learning

Area of Science:

• Anesthesiology
• Medical Informatics
• Machine Learning

Background:

• Arterial hypotension during anesthesia induction is a significant risk factor for adverse patient outcomes.
• Predicting this complication is challenging due to its multifactorial etiology.
• Early identification of patients at risk is crucial for timely intervention.

Purpose of the Study:

• To evaluate the feasibility of developing machine learning models for predicting postinduction hypotension.
• To compare the performance of different machine learning algorithms in forecasting hypotension.
• To identify key patient and anesthesia parameters predictive of hypotension.

Main Methods:

• Trained four machine learning models (Naïve Bayes, logistic regression, random forest, artificial neural network) to predict hypotension.
• Utilized data from anesthesia monitors, infusion pumps, anesthesia machines, and electronic health records (demographics, comorbidities).
• Data collected from the start of induction up to immediately before tracheal intubation.

Main Results:

• The random-forest model achieved the highest predictive performance with an AUC of 0.842.
• This outperformed Naïve Bayes (0.778), logistic regression (0.756), and artificial neural network (0.760) models.
• Key predictors included lowest systolic blood pressure, lowest mean blood pressure, and mean systolic blood pressure prior to tracheal intubation.

Conclusions:

• Machine learning models can effectively predict postinduction hypotension using data available before tracheal intubation.
• The random-forest model shows superior performance in this predictive task.
• This approach offers a potential tool for improving patient safety during anesthesia induction.