Abstract:Objective To establish a multi-index linear discriminant model for the diagnosis of bacterial meningitis after neurosurgery operation. Methods A retrospective analysis was performed on 14 laboratory examinations of bacterial meningitis (n=226) and aseptic meningitis (n=255) in patients in a hospital from 2012 to 2016. Receiver operating characteristic (ROC) curve for significant variables was drawn, items with area under the curve (AUC)>0.7 were conducted linear regression fitting, and a linear discriminant model for combined multi-index diagnosis was established, data of 26 patients with meningitis were selected to verify the discriminant capacity of the discriminant model. Results There were significant differences in 11 indexes between two groups of patients with meningitis (all P<0.05). AUC of cerebrospinal fluid(CSF) white blood cell count (C-WBC), CSF glucose concentration (C-Glu), blood glucose concentration (B-Glu), CSF blood glucose ratio (C/B-Glu) and CSF lactic acid (C-Lac) were all>0.7. The discriminant model was obtained by linear fitting:Y=-0.268×C-Glu+0.127×B-Glu+0.24×C-Lac-0.722×C/B-Glu+0.00000638×C-WBC-0.866, AUC of ROC curve of five indexes for diagnosis of bacterial meningitis was 0.907, sensitivity, specificity, positive predictive value, and negative predictive value were all>80.0%. Data of 26 patients with meningitis validated the discriminant capacity of discriminant model, the accuracy and specificity discriminated by the model were high (90.0%, 81.2%), consistency rate was 84.6%. Conclusion The combined diagnosis of multiple indexes can effectively distinguish bacterial meningitis from aseptic meningitis, and promote the diagnosis of bacterial meningitis.