Predictors and empiric anti-microbial therapy of late-onset sepsis in the neonatal intensive care unit

Background: Whenever suspicion of late-onset sepsis (LOS) is raised, sepsis workup is performed and empiric anti-microbial therapy (AMT) is initiated. However, the prescribed AMT may often be inappropriate for the eradication of the causative pathogen. Study aim: To evaluate the clinical, interventional and laboratory predictors of LOS, and to evaluate the appropriateness of empiric AMT in late-onset neonatal sepsis. Patients and Methods: The medical records of all neonates admitted to a tertiary NICU during a 1-year period were retrospectively reviewed. Out of 352 infants admitted, 84 neonates developed 96 suspected septic events beyond 3 days of age and comprised the study population. LOS was defined as clinical signs of sepsis accompanied with positive blood and/or urine and/or CSF cultures obtained at the onset of the septic event. For each infant, we collected demographic and perinatal characteristics, neonatal diagnoses and interventions, clinical signs and laboratory abnormalities at onset of sepsis, the AMT instituted, and microbiological data. Results: Of the 96 events of suspected LOS, 26 (27.0%) had positive blood cultures, with coagulase-negative Staphylococcus (CoNS), Klebsiella and Candida accounting for 22/26 (84.7%) of these events. In four out of 75 events (5.3%), urine culture was positive. Logistic regression multivariate analysis showed that birth weight (OR=0.9, 95% CI: 0.82-0.99 p=0.031), apnea/ bradycardia (OR=3.16, 95% CI: 1.08-9.25 p=0.036), and platelet count<100,000/mm3 (OR=7.04, 95% CI: 1.21-40.9 p=0.03) were significantly associated with LOS. Out of 29 proven septic events, 14 (48.3%) had received AMT within 3 days prior to onset of LOS. In 26/29 (89.7%) of proven septic events, empirical AMT was started after sepsis work-up, mostly vancomycin 18/29 (62.0%), imipenem 14/29 (48.3%), and amphotericin B 9/29 (31.0%). Out of the 26 proven septic events treated with empiric AMT, the causative pathogen was sensitive to the prescribed AMT in 24 (92.3%) of the cases. Conclusions: Significant predictors of LOS include lower birth weight, apnea/bradycardia and thrombocytopenia. Empirical AMT was initiated in 90% of proven septic events and was appropriate in 92.3% of these cases. When confronted with suspected LOS in the NICU setting, empiric coverage of CoNS, Klebsiella and fungi should be considered, based on the infant's condition and on local microbiological data.