Work

1449Clinical Prediction Model for Candidemia in Pediatric ICU Patients: Failure to Validate

Year: 2014

Type: conference-abstract

Abstract: Background. Candidemia is associated with significant morbidity and mortality and pediatric intensive care unit (PICU) patients with candidemia are at particular high risk for death. A clinical prediction rule for early detection of candidemia in the PICU was previously derived at a single center. Multi-center validation is necessary prior to routine application. Methods. Patients <19 years of age with candidemia were identified prospectively from six pediatric institutions (January 1, 2005-December 31, 2009). Cases were matched to two controls in the same PICU by incidence density sampling. Conditional logistic regression was performed to evaluate the association of candidemia with the previously derived predictor variables. Results. Ninety-six patients with candidemia were identified with a median age of 4 years (range 43 days to 18 years) and matched to similar controls. The proportion of predictors of candidemia for cases and controls varied significantly by study site (table) and only the presence of a central venous catheter was associated with candidemia (OR: 9.6; 95% CI: 3.4-27.3). Conclusion. Our multi-center study did not validate the previously derived prediction rule due to significant variability in predictors across hospitals and time. Consideration should be given to including biomarkers in future prediction rules. Disclosures. T. Zaoutis, Merck: Investigator, Research grant, Consultant, Consulting fee; Pfizer: Consultant, Consulting fee; Astellas: Consultant, Consulting fee

Source: Open Forum Infectious Diseases

DOI: 10.1093/ofid/ofu052.995