Development and bootstrap-based internal validation of a prediction model for in-hospital mortality in patients with Fournier’s gangrene

Fournier’s gangrene (FG) remains a high-mortality surgical emergency where early risk-stratification is often complex. We aimed to develop a preliminary prediction model for in-hospital mortality in patients with FG, perform bootstrap-based internal validation, and compare its apparent performance with established scoring systems. This retrospective multicenter study analyzed 133 patients from four tertiary hospitals in South Korea. We retrieved demographic and clinical data from electronic records and calculated FG severity index (FGSI) and Uludag FGSI (UFGSI) for comparison. Independent prognostic factors were identified using multivariate logistic regression, which informed the construction of both a predictive nomogram and a web-based risk calculator. Model performance was comprehensively assessed through discrimination, calibration, and decision curve analysis (DCA), followed by internal validation using a 1,000-iteration bootstrap resampling procedure. The in-hospital mortality rate was 24.1% ( n = 32). Multivariate analysis identified age, lactate, albumin, and hemoglobin as independent predictors of mortality. The proposed Yonsei Fournier’s gangrene prognostic index (YFGPI) showed high apparent discrimination in the development dataset, with an AUC of 0.936. After bootstrap validation incorporating the predictor-selection process, the optimism-corrected C-index was 0.909. Exploratory DCA suggested a potential net benefit of the YFGPI compared with the FGSI and UFGSI across a range of threshold probabilities in the development cohort. The YFGPI is a preliminary, parsimonious prediction model based on readily available clinical parameters. It showed potential utility for individualized mortality-risk estimation in the development cohort; however, external validation and prospective impact assessment are required before it can be used to guide clinical decision-making. not applicable.