Abstract

Objective: Neonatal bloodstream infection (BSI) is a major contributor to mortality, health service costs, and the population burden of lifelong neurodisability. BSI surveillance, an essential component of infection control, requires an unambiguous standardised case definition as variability would invalidate any comparative analyses. In neonates a high proportion of blood cultures yield a mixed growth or skin commensals, principally coagulase-negative staphylococci (CoNS). As this might represent either genuine BSI or contamination, clinical correlates are necessary, but this adds to the difficulty of agreeing an objective, standardised case definition.

Design: Utilising data from 26 UK neonatal units, the population prevalence of 12 predefined clinical signs of infection captured daily for 28 days was evaluated. The sensitivity, specificity, odds ratio and positive predictive value of each sign and sequential numbers of grouped signs were determined to develop a predictive model for a positive blood culture. Sandwich estimates of the standard errors of the logistic regression coefficients were used to take account of the correlations between these repeated measures. The model was tested in an independent data set.

Results: ⩾3 clinical signs had the best predictive accuracy for a positive blood culture (76.2% specificity; 61.5%, 46.9% and 78.2% sensitivity for all positive cultures, cultures yielding CoNS, or a recognised pathogen, respectively).

Conclusion: This study suggests that a simple case definition for national and international neonatal BSI surveillance is provided by a blood culture yielding a recognised pathogen in pure culture, or a mixed growth or skin commensal plus ⩾3 predefined clinical signs.