BackgroundThe gold standard for diagnosis of meningitis is the isolation of a pathogen from cerebrospinal fluid (CSF) by culture or PCR. However, treatment is routinely commenced based on CSF findings prior to microbiological results. This study determined the predictive value of CSF parameters for diagnosing bacterial and viral meningitis in young infants.MethodsMulticentre retrospective (2010–2020) cohort study of 1088 CSF results from infants aged 0–90 days. The predictive value of CSF parameters (white blood cell count (WCC), neutrophil, protein, glucose) was evaluated in 538 meningitis cases (39 bacterial, 499 viral) compared with controls with negative CSF microbiological testing and no prior antibiotics.ResultsFor bacterial meningitis, the sensitivity of the commonly used CSF WCC cut-off of 20×106/L for neonates, 15×106/L for infants 1–2 months old and 5×106/L for infants 2–3 months old was 89%, 91% and 86% and the specificity was 78%, 77% and 61%, respectively. CSF protein levels ≥1 g/L in neonates and ≥0.8 g/L in infants aged 1–3 month, or CSF neutrophils ≥2×106/L, independently increased the likelihood of bacterial meningitis (positive likelihood ratios ≥5 and ≥3, respectively). 3 of 39 cases of bacterial meningitis would have been missed using the commonly used WCC cut-offs alone. However, two would have been identified using CSF protein and neutrophil thresholds. All CSF parameters were poor at identifying viral meningitis.ConclusionA single CSF parameter cannot reliably diagnose bacterial meningitis. For identification of bacterial meningitis, elevation of CSF WCC, neutrophil count or protein levels above threshold values improves accuracy of diagnosis.