Abstract
OBJECTIVE: The diagnosis of post-neurosurgical bacterial ventriculitis/meningitis (BV/M) remains challenging, particularly in the patients with leukocytic pleocytosis and hypoglycorrhachia. This study aimed to establish a nomogram identifying the high-risk and low-risk post-neurosurgical BV/M and evaluate the utility of this risk scoring system. METHODS: Adult patients with CSF leukocytes ≥ 100/mm(3) and glucose ≤ 2.2 mmol/L experienced neurosurgical or invasive procedures in three hospitals were divided into training and validation cohort, patients from Medical Information Mart for Intensive Care (MIMIC)-III and MIMIC-IV were also used as validation cohort. Multivariate logistic regression was performed to identify independent predictors and establish a nomogram to predict the occurrence of BV/M. RESULTS: Totally, 271 patients (150 confirmed BV/M and 121 confirmed non-BV/M) selected from 711 suspected post-neurosurgical BV/M patients with leukocytic pleocytosis and hypoglycorrhachia were used as training cohort. CSF glucose, CSF leukocytes, CSF erythrocytes, CSF neutrophil proportions, blood lymphocyte proportions, and external ventricular drainage filtered out from 20 easily available parameters were used as independent predictors to develop the nomogram for BV/M. The high discriminative power of this nomogram was assessed by two independent validation cohorts (84 and 58 patients respectively). CONCLUSION: The nomogram uses six easily available indexes to predict BV/M risk. High-risk patients should receive immediate antibiotics, increasing CSF examination frequency is recommended before antibiotic treatment in low-risk patients.