Diagnostic risk factors to differentiate tuberculous and acute bacterial meningitis

Abstract

The objective of this study was to identify independent predictor factors for diagnosis of tuberculous meningitis and develop a clinical prediction tool based upon a set of simple clinical and laboratory parameters in our local population. Clinical and laboratory features were compared in 68 patients with tuberculous meningitis and 123 cases of acute bacterial meningitis in 3 referral centres for tuberculosis in south-eastern Iran. Twenty-two clinical and laboratory features were analysed. Based on the best-fitted model a receiver operating characteristic curve with the highest surface under the curve was constructed. Disease duration before diagnosis (>or=5 d) had the highest odds ratio of 21.9. Age over 30 y, CSF leukocyte count <or=1000 x 10(3) cells/ml and CSF lymphocytosis >or=70% were placed after disease duration with odds ratios of 5.1, 3.7 and 2.6, respectively. Sensitivity, specificity and likelihood ratio for a positive test in this model were 84%, 88% and 7.4, respectively. The area under the ROC curve was 0.92. It appears that a single model can not predict TBM diagnosis in different populations. Using clinical and laboratory parameters may facilitate empirical diagnosis of TBM in endemically low income countries with limited microbiological diagnostic facilities.