ABSTRACT
PURPOSE: The aim of this study was to identify independent predictors of acquiring a nosocomial bloodstream infection (BSI) during extracorporeal membrane oxygenation (ECMO). METHODS: This retrospective cohort consisted of 202 neonates treated with ECMO from 1989 to 1998 at the author' institution. Data collected included patient demographics, primary and secondary diagnoses, white blood cell counts, antibiotic usage, presence of central lines, operative procedures, and outcome. Surveillance blood cultures were drawn daily from the circuit using sterile technique to identify acquired pathogens. Statistical analyses included logistic regression, Cox proportional regression analysis, and discriminate analysis. RESULTS: There were 1,245 blood cultures drawn on 202 patients (6.2 cultures per patient), and a nosocomial BSI was identified in 7 patients (3.4%) during this 10-year span. These were infections that were neither present nor incubating on admission. Pre-ECMO diagnoses of patients who had a nosocomial BSI while on bypass included group B beta-hemolytic streptococcal sepsis (n = 2), herpes simplex viral sepsis (n = 1), congenital diaphragmatic hernia (n = 2), persistent pulmonary hypertension (n = 1), and congenital heart disease (n = 1). The median time on ECMO before obtaining a positive culture was 390 hours. The infectious agents responsible for these BSIs included Staphylococcus epidermidis (n = 5), Staphylococcus aureus (n = 1), and Escherichia coli (n = 1). The major factor associated with acquiring a nosocomial BSI on ECMO was the duration of bypass (391 v 141 hours, P =.002). Additionally, patients in the BSI group were more likely to have had an arterial catheter in place (16 v 7 days, P =.009) and to have received more screening blood cultures (16 v 6 cultures, P < 001). White blood cell counts, absolute neutrophil counts, and immature/total (I/T) ratios were not useful in predicting a nosocomial BSI. Of the 31 patients who required ECMO for more than 10 days, 7 (23%) had a positive blood culture, and 5 of these 7 infants (71%) died (P =.03). CONCLUSIONS: The only predictor of acquiring a nosocomial BSI on ECMO was the duration of support for greater than 10 days. Because classical predictors of infection are unreliable while the patient is on ECMO, the authors suggest that obtaining daily surveillance blood cultures beginning on the tenth day should be performed with prolonged ECMO courses. The authors confirmed previous reports of the association between a prolonged ECMO course and a high mortality rate. However, the authors speculate that, in actuality, the primary diagnosis leads to the prolonged course of support and is the major factor in the infant' demise.
