Identification of Staphylococcus aureus spondylitis in patients with spinal infections using machine learning based on routine laboratory tests: a multicenter retrospective study.

Rapid etiological identification of Staphylococcus aureus in spinal infections can be challenging, often delaying targeted therapy. We developed a machine learning model leveraging XGBoost to predict S. aureus etiology in spinal infections directly from routine laboratory indicators. The XGBoost model demonstrated superior predictive performance (AUC 0.812; 95% CI: 0.728-0.896) among four algorithms, with SHAP analysis identifying D-dimer, Monocyte Percentage, Albumin, and Alanine Aminotransferase as crucial predictors.