LAUSR

Background Patients harboring an external ventricular drain (EVD) who develop signs of infection require screening for infection in the central nervous system (CNS). The cerebrospinal fluid (CSF) can be collected by the EVD or by lumbar puncture (LP). If only one sample is analyzed, the diagnosis might be dubious or false-negative. The objective of this study was to compare the diagnosis accuracy of CNS infection of CSF samples collected from EVD and LP. Methods We conducted a transversal study where data were prospectively collected from 2016 to 2019. Patients harboring EVD with signs of infection were submitted to the CSF analysis collected by LP and EVD. Diagnosis sensibility and results correlation were analyzed using the kappa index. Results The 141 samples from LP and 141 samples from EVD were collected from 108 patients. Among the 282 samples, a total of 77 had infection. Seventy CSF samples from LP fulfilled infection criteria. However, only 32 EVD samples demonstrated infection. Among the 70 cases of infection based on the LP sample, 25 CSF samples collected from the EVD were also suggestive of infection; but in 45 patients only the CSF samples from LP met infection criteria. Seven patients had diagnosis of infection only in the EVD sample. The kappa correlation index of the results obtained from LP and EVD was 0.260 and the McNemar χ2 test was Conclusions The CSF analysis exclusive from the EVD has a low sensibility and negative predictive value. CSF collected from LP has a sensibility 2.18 times higher than EVD.