Abstract
BACKGROUND: Accurately predicting mortality in patients with skin and soft-tissue infections (SSTIs) remains challenging. Machine learning models offer rapid processing, algorithmic impartiality, and strong predictive accuracy, which may improve early risk stratification in the emergency department (ED). METHODS: We retrospectively analyzed clinical data from 1,294 ED patients diagnosed with SSTIs between March 2015 and December 2020. Five machine learning algorithms-logistic regression (LR), k-nearest neighbours (KNN), support vector machine (SVM), random forest (RF), and Extreme Gradient Boosting (XGBoost)-were developed using 20 candidate variables, with model performance evaluated in independent runs. A simplified XGBoost model using only the six most influential predictors was also derived for bedside application. RESULTS: Among the five models, XGBoost achieved the highest performance (AUC = 0.892, sensitivity = 86.9%, specificity = 93.4%). The streamlined six-variable XGBoost model further improved predictive metrics (AUC = 0.922, sensitivity = 88.5%, specificity = 95.4%), matching or slightly surpassing the full model while reducing data requirements. CONCLUSIONS: XGBoost outperformed LR, KNN, SVM, and RF in predicting SSTI mortality, offering both higher accuracy and operational efficiency. Its sequential tree-building, regularization, and robust handling of missing data enable superior discrimination in tabular clinical datasets. The simplified model, requiring only standard admission variables, provides a fast, cost-effective, and highly accurate tool for early identification of high-risk patients in the ED.