"Could She/He Walk Out of the Hospital?": Implementing AI Models for Recovery Prediction and Doctor-Patient...

Area of Science:

• Trauma Care
• Medical Informatics
• Artificial Intelligence in Medicine

Background:

• Major trauma is a leading cause of mortality and disability globally, necessitating improved predictive tools.
• The unpredictable nature of trauma requires advanced methods for forecasting patient outcomes.
• Existing models for trauma patient management are continually being enhanced.

Purpose of the Study:

• To develop and validate an AI/ML model for predicting adverse events in major trauma patients.
• To integrate a predictive tool into the hospital information system for clinical use.
• To assess the performance of AI/ML models against traditional clinical scoring systems.

Main Methods:

• Retrospective analysis of 5521 major trauma patients admitted between 2010 and 2019.
• Evaluation of five AI models, including XGBoost, for predictive accuracy.
• Deployment of the best-performing model as a web-based application integrated with the HIS.

Main Results:

• XGBoost achieved the highest performance (AUC 0.748), surpassing Injury Severity Score (ISS) and Glasgow Coma Scale (GCS).
• The integrated model demonstrated improved efficiency and interpretability via SHAP analysis.
• Healthcare professionals provided positive feedback on the tool's usability and clinical value.

Conclusions:

• An AI-powered predictive model for major trauma patients has been successfully developed and integrated into clinical workflow.
• The model enhances efficiency and quality of care in trauma management.
• Future work will expand predictive capabilities to include complications and treatment outcomes for broader clinical application.