74 MULTICENTER VALIDATION OF CORRECTION FACTORS FOR CEREBROSPINAL FLUID LABORATORY VALUES IN YOUNG INFANTS WITH A TRAUMATIC LUMBAR PUNCTURE

Abstract

Abstract Background The evaluation of fever among infants in the first months of life remains one of the most common problems in pediatric healthcare. Approximately 10% harbor potentially life-threatening infections including bacterial meningitis, frequently necessitating invasive cerebrospinal fluid (CSF) testing by lumbar puncture (LP). LPs are often traumatic leading to uninterpretable results and consequently, broad-spectrum antibiotic exposure and prolonged hospitalization. Several strategies have been proposed to identify low-risk infants with traumatic LPs, including recently-derived correction factors, however studies validating the safety and diagnostic utility of such approaches are lacking. Objectives To evaluate the test characteristics and misclassification rates of recently described ratio-based correction methods for the interpretation of CSF results among young infants with traumatic LPs. Design/Methods We undertook a multicenter cohort study of infants aged ≤60 days with a traumatic LP performed at two urban tertiary Pediatric hospitals from 2006 to 2018. Traumatic LPs were defined as CSF specimens with ≥10,000 RBCs/mm3, and for infants aged ≤28 days and 29-60 days, pleocytosis was defined as ≥20 and ≥10 WBC/mm3, respectively, and abnormal protein ≥1.15 and ≥0.89 g/L, respectively. CSF WBCs and protein were adjusted downward for traumatic LPs using RBC ratio-based correction methods (newly derived 877:1, commonly used 500 and 1000:1, peripheral RBC:WBC ratio, and newly derived 1000 RBCs:0.011g/L protein). Descriptive statistics are presented with sensitivity, specificity, and negative predictive values of unadjusted and adjusted CSF for predicting culture-proven bacterial meningitis. Results Of 4,912 LPs meeting inclusion criteria, 437 (8.9%) were traumatic, among which 4 (0.9%) were positive for bacterial meningitis. Compared to uncorrected CSF WBC counts, both 877 and 1000 correction factors classified fewer infants with pleocytosis (38.0% and 42.6% vs 81.7%). These correction factors both maintained 100% sensitivity and 100% negative predictive value, and performed with greater specificity for bacterial meningitis than the uncorrected WBC count (62.6% and 58.0% vs 18.5%). No infants with bacterial meningitis were misclassified using either 877 or 1000:1 correction factors. CSF 500:1 and peripheral RBC:WBC correction ratios performed with the lowest sensitivity and negative predictive values and both misclassified 1 infant with bacterial meningitis. Corrected CSF protein outperformed uncorrected protein in specificity (66.8% vs 33.9%), but did not add diagnostic value when used in combination with WBC correction ratios. Conclusion Correction of the CSF WBC count substantially reduced the number of infants classified with CSF pleocytosis. The newly-derived 877:1 correction factor performed with the best test characteristics, safely reclassifying nearly half of all infants with a traumatic LP. It may be appropriate to use a correction factor in the evaluation of CSF cell counts in traumatic LPs in order to more effectively risk-stratify febrile young infants, reduce antibiotic exposure and admission duration.