Predicting reoperation and readmission for head and neck free flap patients using machine learning

Area of Science:

• Surgical Oncology
• Plastic Surgery
• Machine Learning in Medicine

Background:

• Head and neck (HN) free flap reconstruction is complex.
• Unplanned readmission and reoperation are significant concerns post-HN surgery.

Purpose of the Study:

• Develop machine learning (ML) models to predict 30-day unplanned readmission and reoperation.
• Utilize demographic and perioperative factors for prediction.

Main Methods:

• Extracted data from the 2012-2019 NSQIP database.
• Employed eXtreme Gradient Boosting (XGBoost) for ML model development.
• Validated models using 2019 data.

Main Results:

• ML model for readmission: 82% accuracy, 63% sensitivity, 85% specificity, AUC 0.78.
• ML model for reoperation: 62% accuracy, 51% sensitivity, 64% specificity, AUC 0.58.
• 10.7% of patients experienced 30-day readmission; 18.3% underwent reoperation.

Conclusions:

• XGBoost effectively predicts readmission and reoperation in HN free flap patients.
• Findings support shared decision-making between clinicians and patients.
• Potential to enhance future database data collection.