Abstract
PURPOSE: To develop and validate a predictive model for necrotizing soft tissue infections (NSTIs) in hospitalized patients with skin and soft tissue infections (SSTIs). PATIENTS AND METHODS: In this retrospective single-center study, 131 patients with SSTIs admitted between April 2021 and March 2024 were divided into NSTI (n=72) and non-NSTI (n=59) groups. Demographic information, clinical characteristics, and key severity scores were collected. LASSO regression was used to select variables, and a logistic regression model was then developed. Model performance was evaluated using AUC, sensitivity, specificity, and the Hosmer-Lemeshow test. Internal validation was performed using 1000 bootstrap resamples. Model discrimination, calibration, and clinical utility were assessed through ROC curve, calibration curve, and decision curve analysis (DCA), respectively. RESULTS: Four independent predictors-age, temperature, depth of infection, and SOFA score-were identified. A nomogram constructed based on these variables exhibited good discriminative performance, with an AUC of 0.832 (95% CI: 0.763-0.900), a sensitivity of 0.722, and a specificity of 0.797. Internal validation revealed an AUC of 0.832 (95% CI: 0.752-0.897). The model passed the H-L test (p = 0.322). The calibration curve showed close alignment between predicted and observed outcomes (slope = 0.916), indicating good calibration. DCA demonstrated net clinical benefit across a wide range of threshold probabilities (0.1-0.9). CONCLUSION: The predictive model is simple and practical, with good discriminative and calibration performance. It may serve as a useful decision-support tool for early identification of NSTIs. However, future multicenter prospective studies are still needed for external validation to assess its practical application value.