Development of Machine Learning Models for Predicting Surgical Site Infection After Spinal Surgery

Area of Science:

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

Background:

• Surgical site infection (SSI) is a significant complication following spinal surgery.
• Predicting SSI is crucial for patient management and resource allocation.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting postoperative SSI using preoperative clinical data.
• To assess model calibration and clinical applicability for risk stratification.

Main Methods:

• Retrospective single-center study involving patients undergoing spinal surgery.
• Development and comparison of logistic regression, random forest, gradient boosting, and XGBoost models.
• Performance evaluation using AUC, AUPRC, sensitivity, precision, F1 score, Brier score, and calibration slope; SHAP analysis for interpretability.

Main Results:

• The incidence of SSI was 16.6%.
• Logistic regression achieved the highest AUC (0.806) and sensitivity (0.758) in the test set, comparable to machine learning models.
• Key predictors included C-reactive protein, hemoglobin, albumin, and white blood cell count; perioperative variables offered limited additional value.

Conclusions:

• Machine learning models demonstrate acceptable performance for SSI prediction in spinal surgery.
• Conventional logistic regression offers comparable performance, validating its clinical utility with structured datasets.
• Preoperative clinical and laboratory variables are primary predictors, suitable for routine risk stratification.