Fungemia remains a major threat in intensive care units (ICUs), with high mortality rates despite advances in diagnostics and treatment. Colonisation by yeasts is an independent risk factor for fungemia; however, its predictive utility requires further research. In this 8-year study, we analysed 38,017 samples from 3206 patients and 171 fungemia episodes as part of a weekly fungal surveillance programme. We evaluated species-specific colonisation patterns, the predictive value of the Colonisation Index (CI) and Corrected Colonisation Index (CCI), and candidemia risks associated with different yeast species and anatomical site colonisation. Our results showed that C. auris, N. glabratus, and C. parapsilosis colonisation increased with longer hospital stays (0.8% to 11.55%, 8.13% to 16.8%, and 1.93% to 5.14%, respectively). The CI and CCI had low discriminatory power (AUROC 67% and 66%). Colonisation by any yeast genera demonstrated high sensitivity (98.32%) and negative predictive value (NPV) (95.90%) but low specificity and positive predictive value (PPV) (23.90% and 6.64%). Tracheal and urine cultures had the highest PPV (15.64% and 12.91%), while inguinal cultures had the highest NPV (98.60%). C. auris (12.32%) and C. parapsilosis (5.5%) were associated with a higher fungemia risk (log-rank < 0.001). These findings support the use of weekly surveillance to better stratify the fungemia risk and optimise antifungal use in ICUs.
Keywords: Candida auris; candidemia; candidemia predictors; colonisation; intensive care unit; surveillance.