Background: Evaluation of the role on patient mortality exerted by biofilm forming (BF) Candida strains, by using predictive clinical data.
Methods: Eighty-nine strains isolated from Candida bloodstream infection, occurring in two Italian University Hospitals, were employed in this study. A random forest (RF) model was built with a procedure of iterative selection of the risk factors potentially able to predict the probability of death. The similarity between patient conditions and Bayesian clustering was calculated in order to evaluate the role of predictors in the stratification of the death risk.
Results: Three different groups of patients with different probability of death were obtained with a RF approach: Group 1 (mortality in 33.3% of cases), Group 2 (death in 50% of cases), and Group 3 (mortality in 76.9% of cases). The comparison between these three groups showed that BF correlated well with increased mortality in patients, admitted for medical diagnosis, with high APACHE II score and treated with azoles. Early treatment within 24 h between candidemia diagnosis and the beginning of antifungal therapy was associated with the lowest of BF rate and mortality.
Conclusions: BF by Candida spp. seems to be clinically associated with increased mortality especially in medical patients with higher Apache II score or treated with azoles.
Keywords: Antifungal therapy; Biofilm; Candidemia.