Association Between Red Blood Cell Distribution Width and Liver Injury in Children with Epstein-Barr Virus Infection: A Retrospective Study with Incremental Predictive Value Assessment

Background: Liver injury is a common complication of Epstein-Barr virus (EBV) infection in children, yet reliable predictive biomarkers remain limited. Red blood cell distribution width-coefficient of variation (RDW-CV), an emerging inflammatory marker, has shown prognostic value in various infectious diseases but its role in EBV-related hepatic impairment is poorly characterized. Methods: This retrospective study included 123 hospitalized children with laboratory-confirmed EBV infection at Nanjing Lishui People's Hospital between July 2020 and December 2025. Liver injury was defined as ALT >80 U/L. The dose-response relationship between RDW-CV and liver injury was assessed using restricted cubic spline analysis. Multivariable logistic regression with progressive adjustment evaluated the independent association. Five variable selection methods (R², adjusted R², BIC, Mallows' Cp, LASSO,) integrated with clinical expertise identified nine predictors for model construction. The incremental value of RDW-CV was quantified by comparing AUC, NRI, and IDI between models with and without RDW-CV. Results: Of 123 patients, 48 (39.0%) developed liver injury. RDW-CV was significantly elevated in the liver injury group (13.6 ± 1.0% vs. 13.2 ± 0.8%, P = 0.046). Restricted cubic spline analysis revealed a significant linear association (P for overall = 0.007). In fully adjusted multivariable analysis, each 1% RDW-CV increment conferred threefold higher odds of liver injury ( OR 3.06, 95% CI 1.59–5.91, P= 0.001), with consistent effects across age and sex subgroups. The 9-predictor model including RDW-CV demonstrated superior discrimination compared to the model without RDW-CV (AUC: 0.800 vs. 0.718, P = 0.035; categorical NRI 0.274, P = 0.015; continuous NRI 0.617, P < 0.001; IDI 0.130, P < 0.001). Furthermore, the 9-predictor model showed adequate calibration (MAE = 0.059) and favorable net benefit in decision curve analysis (DCA). Conclusions: RDW-CV is an independent, robust predictor of liver injury in children with EBV infection. Incorporating RDW-CV into clinical prediction models significantly enhances risk stratification accuracy, supporting its potential as a routine adjunctive biomarker for early identification of hepatic complications.