RESUMEN
OBJECTIVE: To identify the validity of surveillance cultures in predicting causative organism(s) of late onset neonatal sepsis. RESULTS: Prospective analytical study was conducted from January to April 2011 at the Neonatal Intensive Care Unit, Teaching Hospital, Karapitiya, Galle, Sri Lanka. Fifty neonates were screened on admission and weekly thereafter for colonization with potential pathogens. On suspicion of infection, relevant samples were cultured and tested for antibiotic sensitivity. There were 55 episodes of clinically suspected infections including 33 nosocomial infections. One-third (17/55) of all clinically suspected infections were culture positive. Out of 55, only 33 episodes were clinically suspected nosocomial infections. Clinically suspected nosocomial infection rate was 50/1000 patient-days. Culture proven nosocomial infection rate was 13.61/1000 patient-days. Coliforms were the commonest clinical isolate (76%) and 2/3 of them produced extended spectrum ß lactamase. More than 80% of the isolates causing late onset sepsis were sensitive to carbapenems and aminoglycosides. Sensitivity, specificity, positive predictive value and negative predictive value of surveillance cultures were 77.8, 37.5, 31.8 and 81.8%, respectively. Surveillance samples can be used to predict pathogens of late-onset sepsis. Broad-spectrum antibiotics (carbapenems, aminoglycosides) are recommended as empirical therapy for late-onset neonatal sepsis.