A Novel Diagnostic Prediction Model for Distinguishing Between Tuberculous and Cryptococcal Meningitis.

Background and aim: Tuberculous meningitis (TBM) and cryptococcal meningitis (CM) are easily misdiagnosed due to atypical clinical symptoms. It is difficult for physcians to achieve a rapid and accurate differential diagnosis of TBM in the early stages of disease onset. The aim of this study was to construct a diagnostic prediction model for TBM and CM.

Methods

In this retrospective study, 194 patients with TBM and CM were divided into two groups: training group (163 patients) and validation group (31 patients). Univariate and multivariate analyses were performed on the training group patients. The diagnostic factors were selected to construct the differential diagnostic prediction model for TBM and CM, and the prediction model was verified. A receiver operating characteristics curve (ROC) was constructed and used to evaluate the diagnostic value of the novel model.

Results

Univariate analysis of clinical characteristics revealed differences in eight parameters (P<0.05) between tuberculous meningitis and cryptococcal meningitis. The multivariate analyses showed there were five independent differential factors including age, disease course, albumin-to-globulin ratio, cerebrospinal fluid protein, and cerebrospinal fluid sugar to blood sugar ratio in this study between TBM and CM, while there was no significant difference in the number of nucleated cells in CSF (P=0.088). A differential diagnosis model for TBM and CM was constructed based on those factors. A ROC was constructed with an area under curve [AUC] of 94.5%, a sensitivity of 85.71%, and specificity of 94.59% in the training group.

Conclusion

The novel diagnostic scoring model for TBM and CM has greater differential diagnosis potential than previous reports, which can provide more reliable preliminary diagnosis results for primary hospitals, effectively reduce misdiagnosis, and provide references for early treatment.