AI-Powered Prediction of Postoperative Wound Infections in Diabetic Patients
Summary: This study develops DiabCompSepsAI, a Random Forest Classifier using NSQIP data to predict postoperative wound infections and sepsis in diabetic patients with high accuracy (94%+). By analyzing factors like surgical duration, wound classification, and comorbidities, the model enables early interventions to mitigate healing complications, reduce hospital stays, and lower costs in this high-risk group.
Key Highlights:
- DiabCompSepsAI achieves 94% accuracy in predicting wound infections and sepsis, outperforming traditional risk assessments.
- Top predictors include prolonged surgery, contaminated wounds, and patient weight, highlighting modifiable factors for better healing.
- A user-friendly Streamlit app supports real-time predictions, integrating into clinical workflows for proactive wound care.
- AUC scores of 0.92–0.95 demonstrate strong discriminative power, aiding in reducing morbidity from diabetic complications.
- Future validation could expand its use to diverse surgical settings, enhancing postoperative management.
Keywords:
diabetic wound infection,
postoperative sepsis,
AI wound prediction,
surgical site infections,
diabetic wound healing