Development, evaluation and validation of a screening tool for late onset bacteremia in neonates – a pilot study

• Main Authors: Sandra A. N. Walker, Melanie Cormier, Marion Elligsen, Julie Choudhury, Asaph Rolnitsky, Carla Findlater, Dolores Iaboni
• Format: Article
• Language: English
• Published: BMC 2019-07-01
• Series: BMC Pediatrics
• Subjects: Neonates; Late onset bacteremia; Screening tool
• Online Access: http://link.springer.com/article/10.1186/s12887-019-1633-1

Abstract Background Clinical and laboratory parameters can aid in the early identification of neonates at risk for bacteremia before clinical deterioration occurs. However, current prediction models have poor diagnostic capabilities. The objective of this study was to develop, evaluate and validate a screening tool for late onset (> 72 h post admission) neonatal bacteremia using common laboratory and clinical parameters; and determine its predictive value in the identification of bacteremia. Methods A retrospective chart review of neonates admitted to a neonatal intensive care unit (NICU) between March 1, 2012 and January 14, 2015 and a prospective evaluation of all neonates admitted between January 15, 2015 and March 30, 2015 were completed. Neonates with late-onset bacteremia (> 72 h after NICU admission) were eligible for inclusion in the bacteremic cohort. Bacteremic patients were matched to non-infected controls on several demographic parameters. A Pearson’s Correlation matrix was completed to identify independent variables significantly associated with infection (p < 0.05, univariate analysis). Significant parameters were analyzed using iterative binary logistic regression to identify the simplest significant model (p < 0.05). The predictive value of the model was assessed and the optimal probability cut-off for bacteremia was determined using a Receiver Operating Characteristic curve. Results Maximum blood glucose, heart rate, neutrophils and bands were identified as the best predictors of bacteremia in a significant binary logistic regression model. The model’s sensitivity, specificity and accuracy were 90, 80 and 85%, respectively, with a false positive rate of 20% and a false negative rate of 9.7%. At the study bacteremia prevalence rate of 51%, the positive predictive value, negative predictive value and negative post-test probability were 82, 89 and 11%, respectively. Conclusion The model developed in the current study is superior to currently published neonatal bacteremia screening tools. Validation of the tool in a historic data set of neonates from our institution will be completed.