DiabCompSepsAI: Integrated AI Model for Early Detection and Prediction of Postoperative Complications in Diabetic...

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Surgical Outcomes Research

Background:

• Diabetic patients face higher risks of postoperative complications like wound infections and sepsis.
• These complications lead to extended hospital stays and increased morbidity.
• Traditional risk assessments may not fully capture the complexity of these risks.

Purpose of the Study:

• To evaluate the predictive accuracy of a Random Forest Classifier for postoperative wound infections and sepsis in diabetic patients.
• To enable early clinical interventions through improved risk stratification.
• To leverage ensemble learning for enhanced predictive performance.

Main Methods:

• Retrospective analysis of the National Surgical Quality Improvement Program (NSQIP) database.
• Utilized demographic, clinical, and surgical variables, with one-hot encoding and normalization.
• Employed an 80/20 train-test split for a Random Forest Classifier model.

Main Results:

• Random Forest model achieved >94% accuracy for both wound infection and sepsis prediction.
• Precision and recall metrics exceeded 94%, demonstrating high true positive identification.
• AUC values of 0.92 (wound infection) and 0.95 (sepsis) indicate strong discriminative power.

Conclusions:

• The Random Forest Classifier effectively predicts postoperative complications in diabetic patients.
• High performance metrics suggest potential for real-time clinical risk stratification.
• Clinical integration could improve patient outcomes and reduce healthcare expenditures.