Predictive effect of postoperative recovery in general anesthesia patients using interpretable models based on swarm...

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Anesthesiology

Background:

• Postoperative recovery is crucial for patient outcomes and hospital resource management.
• Predicting recovery can optimize patient care pathways and reduce complications.
• Current prediction methods may lack accuracy or interpretability.

Purpose of the Study:

• To develop and validate an interpretable swarm intelligence machine learning model for predicting postoperative recovery in patients undergoing general anesthesia.
• To identify key clinical variables associated with poor postoperative recovery.
• To assess the clinical utility of the predictive model.

Main Methods:

• Retrospective collection of clinical data from 1,128 patients undergoing general anesthesia.
• Development of a predictive model using swarm intelligence machine learning (XGBoost base learner).
• Validation of the model using internal (test set) and external (dataset B) datasets.

Main Results:

• LASSO regression identified seven key predictors: surgery duration, anesthesia duration, neutrophil-to-lymphocyte ratio (NLR), C-reactive protein (CRP), serum creatinine, body mass index (BMI), and age.
• The model achieved high performance, with an F1 score of 0.8447 and AUC of 0.9265 on the training set, and AUC of 0.8383 on external validation.
• The model demonstrated significant predictive capability for postoperative recovery.

Conclusions:

• An interpretable swarm intelligence machine learning model effectively predicts postoperative recovery in patients after general anesthesia.
• The model's predictions can assist clinicians in identifying at-risk patients and planning interventions.
• This approach enhances clinical decision-making for improved patient outcomes.