Factors contributing to preventing operating room "never events": a machine learning analysis.

Area of Science:

• Medical Informatics
• Patient Safety Research
• Surgical Quality Improvement

Background:

• Surgical
• Never Events
• are preventable errors during surgical procedures.
• Risk factors for these events are not well quantified in relation to surgical characteristics.
• Machine learning offers a novel approach to identify and quantify these risks.

Purpose of the Study:

• To identify and quantify risk factors contributing to major surgical
• Never Events
• using machine learning.
• To improve patient safety and the quality of surgical care.

Main Methods:

• Utilized data from 9,234 safety observations and 101 root-cause analyses of
• Never Events
• (wrong site surgery, retained foreign items).
• Employed three random forest supervised machine learning models.
• Evaluated model performance using 10-cross validation and Gini impurity.

Main Results:

• Identified 24 contributing factors across six surgical departments.
• Quantified significant risk factor impacts, with some exceeding 900% in specific departments like Urology and Orthopedics.
• Found pairs of factors increased event probability, particularly in Gynecology (875-1900%) and General Surgery (2720-13,600%).
• Specific factors like surgery length and nurse involvement showed varying impacts on wrong site surgery and retained foreign body events.

Conclusions:

• Machine learning effectively quantified risk factor impact on specific
• Never Events
• relative to surgical characteristics.
• Recommends adjusting safety standards based on risk assessments tailored to individual operating rooms and surgical types.