A new decision support model for preanesthetic evaluation.

Area of Science:

• Anesthesiology and Intensive Care Medicine
• Medical Informatics
• Decision Analysis

Background:

• Reducing anesthetic risks and mortality is a key challenge in anesthesia and intensive care.
• The American Society of Anesthesiologists (ASA) score is crucial for preanesthetic patient evaluation.
• Current methods may lack interpretability or optimal accuracy.

Purpose of the Study:

• To develop a methodology for deriving simple, interpretable rules to classify patients into ASA score categories.
• To support the assignment of ASA scores and inform decisions regarding patient surgical acceptance.
• To improve the accuracy and transparency of preanesthetic risk assessment.

Main Methods:

• Utilized the MR-Sort multiple criteria decision analysis model.
• Developed a diagnosis system to assign ASA scores and guide surgical acceptance decisions.
• Employed a database of 898 patients for model training and validation.

Main Results:

• Achieved high accuracy in predicting ASA scores and surgical acceptance decisions.
• Demonstrated superior performance compared to other machine learning methods.
• Generated simple, human-readable decision rules interpretable by clinicians.

Conclusions:

• The proposed MR-Sort model accurately predicts ASA scores using available patient attributes.
• The model's interpretability enhances clinical decision-making in preanesthetic evaluations.
• The system offers a valuable tool for improving patient safety in anesthesia.