The utility of sepsis scores for predicting blood stream infections in extracorporeal membrane oxygenation

Abstract

Introduction Extracorporeal membrane oxygenation (ECMO) is an increasingly used modality of life support with high risk for nosocomial infections. The accuracy of sepsis prediction tools in identifying blood stream infections (BSI) in this population is unknown as measurement of multiple variables commonly associated with infection are altered by the circuit. Methods This study compares all blood stream infections for patients receiving ECMO between January 2012 and December 2020 to timepoints when blood cultures were negative using the Sequential Organ Failure Assessment (SOFA), Logistic Organ Dysfunction Score (LODS), American Burn Association Sepsis Criteria (ABA), Systemic Inflammatory Response Syndrome (SIRS) scores. Results Of the 220 patients who received ECMO during the study period, 40 (18%) had 51 blood stream infections and were included in this study. Gram-positive infections composed 57% ( n = 29) of infections with E. faecalis ( n = 12, 24%) being the most common organism isolated. There were no significant differences in sepsis prediction scores at the time of infection compared to infection-free time points for SOFA (median (IQR) 7 (5–9) vs. 6 (5–8), p = 0.22), LODS (median (IQR) 12 (10–14) vs. 12 (10–13), p = 0.28), ABA (median (IQR) 2 (1–3) vs. 2 (1–3) p = 0.75), or SIRS (median (IQR) 3 (2–3) vs. 3 (2–3), p = 0.20). Conclusions Our data shows that previously published sepsis scores are elevated throughout a patient’s ECMO course, and do not correlate with bacteremia. Better predictive tools are needed to determine the appropriate timing for blood cultures in this population.