Abstract
BACKGROUND: Cerebrospinal fluid (CSF) analyses are recommended in patients with meningitis and/or encephalitis, but evidence regarding its diagnostic yield is low. We aimed to determine predictors of infectious pathogens in the CSF of adult patients presenting with meningitis, and/or encephalitis. METHODS: Consecutive patients with meningitis and/or encephalitis form 2011-17 at a Swiss academic medical care center were included in this cross-sectional study. Clinical, neuroradiologic, and laboratory data were collected as exposure variables. Infectious meningitis and/or encephalitis were defined as the composite outcome. For diagnosis of bacterial meningitis the recommendations of the European Society of Clinical Microbiology and Infectious Diseases were followed. Viral meningitis was diagnosed by detection of viral ribonucleic or deoxyribonucleic acid in the CSF. Infectious encephalitis was defined according to the International Encephalitis Consortium (IEC). Meningoencephalitis was diagnosed if the criteria for meningitis and encephalitis were fulfilled. Multinomial logistic regression was performed to identify predictors of the composite outcome. To quantify discriminative power, the c statistic analogous the area under the receiver-operating curve (AUROC) was calculated. An AUROC between 0.7-0.8 was defined as "good", 08-0.9 as "excellent", and > 0.9 as "outstanding". Calibration was defined as "good" if the goodness of fit tests revealed insignificant p-values. RESULTS: Among 372 patients, infections were diagnosed in 42.7% presenting as meningitis (51%), encephalitis (32%), and meningoencephalitis (17%). Most frequent infectious pathogens were Streptococcus pneumoniae, Varicella zoster, and Herpes simplex 1&2. While in multivariable analysis lactate concentrations and decreased glucose ratios were the only independent predictors of bacterial infection (AUROCs 0.780, 0.870, and 0.834 respectively), increased CSF mononuclear cells were the only predictors of viral infections (AUROC 0.669). All predictors revealed good calibration. CONCLUSIONS: Prior to microbiologic workup, CSF data may guide clinicians when infection is suspected while other laboratory and neuroradiologic characteristics seem less useful. While increased CSF lactate and decreased glucose ratio are is the most reliable predictors of bacterial infections in patients with meningitis and/or encephalitis, only mononuclear cell counts predicted viral infections. TRIAL REGISTRATION: ClinicalTrials.gov identifier NCT03856528. Registered on February 26th 2019.