Abstract
BACKGROUND: VE is a central nervous system infection of viral origin, and remains an important disease burden in recent years. Early identification of VE patients and timely interventions are crucial for optimizing clinical outcomes. OBJECTIVE: This study aims to construct a predictive model for early detection of VE patients. METHODS: The study retrospectively analyzed clinical data of 160 VE and 131 non-VE patients from China between January 2022 and March 2025. Data were split into training (70%, 203 cases) and validation (30%, 88 cases) cohorts. Predictor variables were identified via logistic regression analyses, and predictive models were established and validated. Model discrimination was assessed using ROC curves, calibration via H-L test and calibration curves, and clinical applicability via DCA. A nomogram was developed for result visualization. RESULTS: Six covariates (ALB, β (2)-MG, IFN-γ, CSF WBC, CSF LYM, CSF P) were identified as independent VE risk factors (OR >1), and IFN-α as a protective factor (OR <1). In the training cohort, a visualized nomogram was built based on the stepwise multivariate logistic regression. AUC of the nomogram in the training and validation cohorts was 0.86 (95% CI, 0.80-0.91) and 0.85 (95% CI, 0.77-0.93). The calibration curves for the probability of VE infection showed optimal agreement between prediction by nomogram and actual observation. DCA indicated that nomogram conferred high clinical net benefit. CONCLUSION: This study's predictive model reliably identifies VE patients, offering a scientific basis for clinical decision-making and improving patient outcomes.