173 New categorization of central line associated bloodstream infection rates in neonatal intensive care units

Friday, March 19, 2010
Grand Hall (Hyatt Regency Atlanta)
Carlos Lapresta-Moros, MD , Miguel Servet Teaching Hospital, Zaragoza, Spain
Gonzalo Santana-López , Miguel Servet Teaching Hospital, Zaragoza, Spain
Souad Belkebir, MD , Miguel Servet Teaching Hospital, Zaragoza, Spain
María-Jesús Hernández-Navarrete, MD , Miguel Servet Teaching Hospital. . Spain, Zaragoza, Spain

Bloodstream infection (BSI) is the most frequent nosocomial infection reported in neonatal intensive care units (NICU). Although it is demonstrated that birth weight is an important risk factor for BSI, its usefulness to define risk groups is not sufficiently investigated.


To analyze the best way of categorizing BSI rates in NICU in order to reach the closiest rates to the real risk from both points of approach: the variables to include and the points of cut for categorizing


An analytical-observational study was carried out in the NICU located in Miguel Servet Universitary Hospital, based on the data obtained from the ICU-HELICS Nosocomial Infection Surveillance System. All neonates admitted in the unit for more than 24 hours, between April 2003 and July 2009, were eligible for the study.

The diagnosis of BSI was based on CDC definitions.

Three principal variables were used for the categorization: birth weight (BW), gestational age (GA) and birth weight for gestational age (Lubchenko´s percentiles). All other factors that may be confounders were also included in the analysis.

For the statistical analysis, bivariate and multivariate logistic regression models were performed.


During the study period, 1090 neonates were enrolled. Central line-associated BSI rate per 1000 central line-days by BW (NHSN categorization) was: ≤750 g 12.51; 750-1000 g 9.98; 1001-1500 g 8.82; 1501-2500 g 3.42; >2500 g 4.51 and total 7.82.

Gestational age, birth weight and birth weight for gestational age were significantly associated with BSI. In the bivariate and multivariate regression models, both BW and GA showed the strongest association (BW: p<0.001, Nagelkerke R2 0.171, area under curve (AUC) 0,775; GE: p<0.001, Nagelkerke R2 0.199, AUC 0.777). When categorizing both of them, the best combinations founded were: GA in weeks categorized as <= 27, 27.1-29, 29.1-32 and >32 (p<0.001, Nagelkerke R2 0.212, AUC 0.777) and BW in grams categorized as <=1000, 1001-1250, 1251-1750 and >1750 (p<0.001, Nagelkerke R2 0.225, AUC 0.777).


The use of the standardization of the birth weight in five categories suggested by NHSN is accepted internationally. Nevertheless, our results suggest that a more efficient option could be the use of GA categorized as above. The advantages would be to obtain a better risk adjustment and a sample size for each category that would allow making strong comparisons.