A prognostic model for early risk stratification in adult community-acquired suspected CNS infections: multicenter development and external validation

Abstract

Objective: Community-acquired central nervous system (CNS) infections remain a major cause of morbidity and mortality, particularly in resource-limited settings. Early prognostication is critical but challenged by delays in definitive diagnosis. This study aimed to develop and externally validate a simple prognostic model to support early risk stratification and intervention. Methods: We conducted a prospective multicenter cohort study in China (NCT04722328), enrolling 1,060 adults with suspected CNS infections. Patients from four hospitals were included in the training group (n = 742) for model development, while patients from three independent hospitals formed the external validation group (n = 318). Independent predictors were identified using least absolute shrinkage and selection operator (LASSO) and multivariable logistic regression. A nomogram was constructed to estimate individualized risk. Model performance was assessed via area under the receiver operating characteristic curve (AUC), concordance index (C-index), calibration plots, decision curve analysis (DCA), and clinical impact curves (CICs). Results: Six predictors were independently associated with poor outcomes: absence of headache, altered consciousness, respiratory failure, hypoproteinemia, low hemoglobin, and hyperglycemia. The model demonstrated strong discrimination in the training group (AUC 0.811; 95% CI 0.774–0.849) and excellent calibration, with DCA indicating clear clinical benefit. External validation confirmed robust performance (AUC 0.855; 95% CI 0.800–0.910), supporting the model’s generalizability. Conclusion: We established a prognostic model to support early identification of severe cases and guide timely comprehensive management in adult patients with suspected community-acquired CNS infections. Clinical registration: ClinicalTrials.gov (NCT04722328).