External validation of bedside prediction score for diagnosis of late-onset neonatal sepsis

Abstract

To prospectively validate performance of a prediction score for diagnosis of late-onset neonatal sepsis (LNS) in a new patient population. Data were prospectively collected from March 2003 to May 2004. Newborns were enrolled if they were in the neonatal intensive care unit (NICU) between 2 and 90 days, and during the first episode of clinical sepsis suspected. LNS was defined as a positive blood or cerebrospinal fluid (CSF) culture, which became the criterion standard. A total of 105 neonates were evaluated for sepsis. Demographic characteristics were as follows: (mean (s.d.)) were gestational age (GA) 29 (3) weeks; birth weight (BW) 1232 (620) g and postnatal age 17.5 day (12). Thirty-five (33%) neonates had LNS (35 positive blood cultures; 2 positive CSF). No significant differences in GA, BW, gender, age and central line utilization were found between LNS positive and LNS negative groups. Using a cut-off score of < or = 3, the score predicted positive culture with sensitivity of 0.97 (95% confidence interval 0.85, 0.99) and a negative likelihood ratio of 0.07. The discrimination and calibration ability of LNS score was acceptable. A simple clinical decision rule previously developed to predict LNS performs equally in an independent population and NICU.