Front Cell Infect Microbiol. 2026 May 15;16:1743976. doi: 10.3389/fcimb.2026.1743976. eCollection 2026.
ABSTRACT
BACKGROUND: Respiratory syncytial virus (RSV) bronchiolitis is a leading cause of pediatric intensive care unit (PICU) admissions worldwide, with septic shock representing a severe and potentially fatal complication. Early identification of high-risk patients remains challenging in clinical practice. This study aimed to develop and validate a reliable prediction model for septic shock in children admitted to PICU with RSV bronchiolitis using readily available clinical and laboratory parameters.
METHODS: A retrospective cohort study was conducted on 224 pediatric patients with RSV bronchiolitis admitted to PICU. Demographic characteristics, clinical features, and laboratory findings upon admission were systematically collected. Given the low event rate, Firth logistic regression was employed for univariate and multivariate analyses to mitigate small-sample bias. Model performance was rigorously evaluated using 5-fold cross-validation, bootstrap resampling, calibration metrics (Hosmer-Lemeshow test, Brier score, calibration intercept and slope), and discrimination measures (area under the receiver operating characteristic curve). Development of web calculator based on Shiny app to facilitate clinical implementation.
RESULTS: Four independent predictors of septic shock were identified through multivariate Firth logistic regression: fungal co-infection (adjusted odds ratio [aOR]: 9.01, 95% confidence interval [CI]: 2.26-36.49, P = .003), elevated admission glucose (aOR: 1.23, 95% CI: 1.04-1.48, P = .02), decreased antithrombin III (aOR: 0.96, 95% CI: 0.94-0.99, P = .004), and elevated interleukin-6 (aOR: 1.00, 95% CI: 1.00-1.01, P = .008). The final model demonstrated excellent discrimination with an area under the curve (AUC) of 0.892 (95% CI: 0.870-0.914) in 5-fold cross-validation and 0.906 (95% CI: 0.838-0.949) in bootstrap validation. Calibration was optimal (Hosmer-Lemeshow χ²=5.991, P = .648; Brier score=0.065; calibration slope=0.957; calibration intercept=0.002). At the optimal threshold of 0.154, the model achieved sensitivity of 75.0%, specificity of 90.5%, and overall accuracy of 88.8%.
CONCLUSIONS: We developed and internally validated a prediction model incorporating fungal co-infection, admission glucose, antithrombin III, and interleukin-6 for early identification of septic shock risk in pediatric RSV bronchiolitis patients. The model demonstrated good discrimination (AUC 0.892) in this single-center cohort. However, given the modest sample size, external validation in larger, multicenter populations is essential before clinical implementation. This tool may support early risk stratification and clinical decision-making, though its impact on patient outcomes requires further evaluation.
CLINICAL TRIAL REGISTRATION: http://www.chictr.org.cn/index.aspx, identifier ChiCTR2200057182.
PMID:42221582 | PMC:PMC13219336 | DOI:10.3389/fcimb.2026.1743976
