Development and validation of a decision tree model to distinguish bacterial meningitis from viral meningitis and/or encephalitis in children

Early differentiation between bacterial meningitis (BM) and viral meningitis and/or encephalitis (VM/E) in children remains challenging due to overlapping clinical features and limited availability of rapid aetiological diagnosis. We aimed to develop and internally validate an interpretable decision tree model based on routinely available variables for early discrimination between BM and VM/E. We conducted a single-centre retrospective cohort study in which paediatric patients were classified as BM or VM/E according to predefined criteria. Candidate variables were first identified through univariate analysis and subsequently incorporated into a decision tree model. Model performance was evaluated in both the training and test sets. A total of 149 children were included, comprising 94 (63.09%) BM and 55 (36.91%) VM/E cases. The decision-tree model included three predictors, with procalcitonin (PCT) as the primary splitting variable, followed by cerebrospinal fluid (CSF) protein and C-reactive protein (CRP). In the training set, the model demonstrated excellent performance, with an area under the curve (AUC) of 0.979 (95% CI: 0.912–0.994), sensitivity of 96.97%, and specificity of 97.44%. In the test set, the model achieved an AUC of 0.753 (95% CI: 0.622–0.885), with sensitivity of 82.14% and specificity of 68.75%. Empirical antimicrobial therapy was frequently initiated prior to definitive diagnosis, with a substantial proportion of patients with VM/E receiving antibacterial treatment and some patients with BM receiving antiviral therapy. We developed an interpretable and parsimonious decision-tree model based on routinely available clinical variables for early differentiation between BM and VM/E in children. This model may serve as a practical tool for early clinical assessment before microbiological confirmation.