Abstract
Background: Data from African neonatal units conducting bloodstream infection (BSI) surveillance is limited. Methods: Prospective clinical and laboratory surveillance of incident BSI episodes was conducted among in-patients at the 132-bed neonatal service at Tygerberg Hospital, Cape Town, South Africa (2017-2021), describing patient demographics, BSI rates, pathogen profiles, and empiric antibiotic concordance rates. Results: In total, 842 BSI episodes were identified in 740 neonates; most were preterm (661/740; 89.3%) and of low birth weight (640/740; 86.5%). The early onset BSI rate (<3 days of life) was 2.9/1000 live births, with S. agalactiae, K. pneumoniae, and E. coli predominating. Over time, ampicillin plus gentamicin coverage rates for early onset BSI pathogens declined from 93.8% to 63.6%. The healthcare-associated BSI rate (onset >3 days of life) was 3.4/1000 in-patient days, with K. pneumoniae, S. aureus, and S. marcescens predominating. Antibiotic coverage rates for healthcare-associated BSIs improved over time, from 72.2% to 89.2% (piperacillin plus amikacin) and from 68.1% to 84.6% (meropenem). Nearly one-third of BSI episodes were fatal (244/842; 29.0%), with two-thirds of these deaths considered BSI-attributable. Gram-negative BSIs increased mortality (OR 2.88; 95% CI 1.93-4.32) compared to Gram-positive BSIs (p < 0.001). Discordant empiric antibiotic therapy (OR 1.55; 95% CI 1.10-2.17) increased the risk of death compared to concordant therapy (p = 0.012). Conclusions: Neonatal BSI surveillance demonstrated that Gram-negative pathogens remain important causes of early onset and healthcare-associated BSIs in this resource-limited neonatal service. Declining coverage rates for empiric antibiotics prescribed for early onset BSI highlight the need for a change in treatment guidelines to minimise discordant therapy.