Optimizing Antibiotics Prophylaxis in Neurosurgery through Machin Learning: Predicting Infections and Personalizing Treatment Strategies

Introduction: Preventing postoperative infections in neurosurgery is crucial to reducing morbidity. Machine learning (ML) models have shown potential in predicting infections and optimizing antibiotic use. Methods: Patient data from neurosurgical procedures were analyzed to develop and evaluate ML models for predicting postoperative infections. Various algorithms, including logistic regression, Random Forest, Gradient Boosting Machine (GBM), SVM, and neural networks, were compared. Performance metrics such as accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC-ROC) were calculated. Results: The GBM model achieved the best performance, with an accuracy of 89.1% and an AUC-ROC of 0.91. The most important predictors of infection were surgical duration (27.3%), preoperative CRP levels (21.8%), and blood loss (18.5%). Patients who developed infections had significantly longer surgeries and elevated CRP levels. Conclusions: ML models demonstrated high accuracy in predicting postoperative infections in neurosurgery. Early identification of high-risk patients may optimize antibiotic prophylaxis and reduce complications. Further validation is required for clinical implementation.