Methods: A retrospective cohort of all CBSI from January 2002 to January of 2015 was assembled by collecting all first blood culture episodes positive for Candida spp. No repeat cultures were included, regardless of time between positivity. Cultures positive for Torulopsis glabrata were combined with those of CG. Demographics and laboratory data were recovered from the electronic medical record, and comorbidities were assessed using ICD-9-CM data. Multivariate logistic regression models were built using a forward stepwise entry method using p<0.05 as a cut off predicting either CK or CG. CG cases were censored from the CK model, and vice versa.
Results: A total of 1,951 unique CBSI episodes were identified. The two most common causes of CBSI were C. albicans and CG, totaling 964 and 400 cases (49% and 21%) respectively. CK represented 60 episodes (3%) of CBSI in this cohort. Age (OR 1.011, p<0.0001) and female gender (OR 1.712, p<0.0001) were significantly associated with CG infections. However, these factors were not predictive of CK infections, where the significant predictors were identified as the presence of neutropenia (OR 5.3, p<0.0001) and renal transplant (OR 33.0, p<0.014).
Conclusion: The predictors of CBSI with Candida spp. known to have decreased susceptibility to fluconazole vary by the species. Age and gender are the strongest predictors of CG CBSI, but renal transplant and neutropenia predict CK CBSI. This could have implications for the empiric therapy of CBSI in certain sub-populations.
W. Powderly, Merck: Scientific Advisor , Consulting fee