A novel diagnostic strategy of differential diagnosis of tuberculous meningitis and non-tuberculous meningitis: a retrospective observational cohort study

Tuberculous meningitis (TBM) leads to severe disability and mortality rates, making early diagnosis critical. However, it is challenging to distinguish it from other common forms of non-tuberculous meningitis (non-TBM), including bacterial meningitis, cryptococcal meningitis, and viral meningitis. This study aims to construct a diagnostic model between TBM and non-TBM. A total of 543 patients were enrolled, and 405 subjects remained and were subsequently divided into a training set and a validation set. Basic information, laboratory results, and imaging results of patients were collected, and R4.1.0 was used to construct and validate the diagnostic model. Subsequently, 30 patients were recruited as an independent validation cohort to verify the diagnostic efficacy of the model further. Finally, 10 patients with suspected TBM were prospectively observed, and the model was applied for diagnosis, with results compared to the final clinical diagnosis. The differential model of TBM and non-TBM was composed of the systemic symptoms of tuberculosis, altered consciousness, neurological deficits, meningeal irritation, cerebrospinal fluid protein, positive T-cell spot test for tuberculosis infection, and C-reactive protein. The areas under the receiver operating characteristic curve for the model in the training and validation sets were 0.872 (95% confidence interval [CI] = 0.833-0.913) and 0.844 (95% CI = 0.751-0.937), respectively. Furthermore, the validation cohort also shows good diagnostic performance with a sensitivity and specificity of 88.9% and 85.7%, respectively. Notably, 9 out of 10 patients had diagnoses consistent with model predictions. A novel diagnostic model was developed and validated using common clinical indicators and laboratory results to distinguish between TBM and non-TBM effectively.IMPORTANCETuberculous meningitis is a serious disease. Currently, there is no effective way to perform early differential diagnosis, particularly in resource-limited settings. This article presents a new, simple method.