Construction and verification of early diagnosis model of tuberculous meningitis

This study aims to develop a diagnostic model capable of accurately distinguishing between TBM and non-TBM in adults using basic clinical and laboratory parameters.

Demographic information, clinical symptoms, imaging characteristics, and cerebrospinal fluid (CSF) parameters were collected. Consecutive patients diagnosed with TBM or non-TBM who met the inclusion and exclusion criteria were enrolled from June 2019 to June 2024 at Ningbo No.2 Hospital (training set). A random sample of 100 eligible patients from the same period at The First Affiliated Hospital of Wenzhou Medical University served as the validation set. LASSO regression with 10-fold cross-validation was applied for variable selection, generating a set of 53 variables with non-zero coefficients. These variables were then ranked by the magnitude of their coefficients, and the top 7 variables with the highest importance were retained for the final multivariate logistic regression model. Model performance, including discrimination, calibration, and clinical utility, was validated in both cohorts.

Multivariate logistic regression identified seizures at admission, mental and behavioral abnormalities, thalamic abnormalities on brain MRI, a positive Pandy test, cerebrospinal fluid (CSF) pressure, CSF glucose levels, and serum albumin as independent predictors of TBM (p < 0.05). The model showed promising discrimination (AUC 0.876 vs. 0.958) and good calibration in both cohorts. Decision curve analysis confirmed its clinical utility, suggesting potential to enhance diagnostic practice.

This multivariable diagnostic model can assist clinicians in early differentiation between TBM and non-TBM using simple clinical and laboratory parameters.