Nosocomial bacteremia (NB) occurs in ≅25% of VLBW infants and is associated with increased length of hospital stay, costs and mortality. Birth weight (BW) and gestational age (GA) are independent risk factors for NB; neither, however, is very discriminating in predicting NB. It has been proposed that CRIB (CS), a severity of illness measure which includes minimum/ maximum oxygen need and maximum base deficit data during the first 12 hours of age, as well as BW and GA, more accurately predicts NB (Arch Dis Child 1996;75:F49). Objective: To compare the utility of CS to that of BW and GA in predicting NB. Design: CS data were collected prospectively for Network VLBW infants (401-1500 gms) born 7/94-6/96 who survived >72 hours. NB was defined as clinically suspected sepsis at >72 hrs of age associated with bacterial growth on blood culture. GA was determined by best obstetrical estimate. Univariate models to compare the ability of CS, and BW and GA alone to predict risk of NB, and a multivariate model of CS components were developed. Results: CS was obtained on 3340 /3840 infants (87%). Incidence of NB was 25%. CS distribution was 0-1=38% with 13% NB, 2-4=24% with 23% NB, 5-9=24% with 36% NB and 10-20=13% with 46% NB. Odds ratio (OR), 95% confidence intervals (CI) and chi-square were:Table CS was related to risk of NB; both GA and BW alone, however, were better predictors of NB than CS. On multivariate analysis the addition of CS components other than BW and GA did little to improve the prediction of NB. Conclusion: In our large VLBW infant cohort there was no benefit to using a proposed measure of severity of illness (CS) compared to BW or GA alone in assessing NB risk.

Table 1
Full size table