Development and validation of a machine learning-based prediction model for surgical site infection after spinal metastasis surgery

Surgical site infection (SSI) is a severe postoperative complication in patients with spinal metastases. This study retrospectively enrolled 460 patients with spinal metastases who underwent surgery at a single center. Stepwise variable screening using univariate analysis, LASSO regression, and multivariate Logistic regression identified seven independent risk factors for SSI: operative time, age, ECOG score, diabetes mellitus, open surgery, preoperative chemotherapy, and hypoproteinemia. Six machine learning models were constructed, and 10-fold cross-validation showed that the Logistic regression model achieved the best performance (AUC = 0.906), with an AUC of 0.866 in the test set. This study developed and validated a risk prediction model for SSI following spinal metastasis surgery with good discrimination and clinical interpretability, providing a reference for individualized perioperative infection risk assessment.