Prediction of candidemia in pediatric patients based on central venous catheter: Real world evidence to guide antifungal empirical prescription

Abstract

Background: Candida is one of the main etiological agents of bloodstream nosocomial infections, and is associated with high morbidity/mortality in children. High use of empirical antifungals incurs in higher health care costs and concerns on emergent resistant Candida spp. Therefore, a prediction model should optimize prescribing empirical antifungals in pediatrics. Objective: Based on real-world evidence data, we aimed at assessing how accurate is a regression model for predicting central venous catheter (CVC) candidemia in pediatric patients. We also explored how this prediction model might be useful for prescribing antifungals empirically. Methods: A case-control study was conducted based on 144 patients with positive and negative blood cultures for Candida spp..  A model for predicting Candida spp. was developed through univariate and multivariate analyses (logistic regression). Data were reported as odds ratio (OR) and p-value less than 0.05 was considered statistically significant. Results: Candidemia was predicted by the presence of CVC (OR 2.561, p=0.0042). For this model, an area under the curve (AUC) of 0.550 (p=0.314) was estimated, representing the Receiver Operating Characteristic curve of the present study, with a sensitivity of 86% and a specificity of 24%. Conclusion: Our study demonstrated that CVC is a fragile predictor for candidemia, however, its clinical significance and low specificity obtained in the ROC curve suggests that other covariates rather than those investigated should be considered to assess its impact on a prediction model for pediatric patients.