Machine learning-based diagnostic model combined with Chinese natural language processing for surgical-site...

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Surgical Patient Monitoring

Background:

• Surgical site infections (SSI) are a significant healthcare-associated complication.
• Early detection of SSI remains a critical challenge in clinical practice.
• Predictive modeling offers a potential solution for proactive SSI management.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for predicting SSI in Chinese surgical patients.
• To assess the model's performance using established metrics like AUC, sensitivity, and specificity.
• To provide a tool for real-time, automated risk assessment of SSI.

Main Methods:

• A multicenter cohort study involving 118,314 surgical patients in China.
• Utilized clinical, microbiological, and demographic data for model development.
• Applied and optimized multiple ML algorithms, including decision trees, with cross-validation.

Main Results:

• The decision tree ML model achieved high predictive performance (AUC 0.92 development, 0.90 validation).
• Demonstrated strong sensitivity and specificity in identifying patients at risk for SSI.
• Model performance was consistent across various patient subgroups.

Conclusions:

• Successfully developed and externally validated an ML-based model for SSI prediction.
• The model exhibits robust and stable performance, suitable for clinical application.
• This tool can enhance clinical decision-making through automated SSI risk assessment.