An approach to develop clinical prediction rule for candidemia in critically ill patients: A retrospective observational study.

Purpose: Early detection of candidemia in critically ill patients is important for preemptive antifungal treatment. Our study aimed to identify the independent risk factors for the development of a new candidemia prediction score.

Methods: This single-centre retrospective observational study evaluated 2479 intensive care unit (ICU) cases from January 2016 to December 2018. A total of 76 identified candidemia cases and 76 matched control cases were analyzed. The patients' demographic characteristics and illness severity were analyzed, and possible risk factors for candidemia were investigated.

Results: Multivariate logistic regression analysis identified renal replacement therapy (RRT) (odds ratio [OR]: 52.83; 95% confidence interval [CI]: 7.82-356.92; P < 0.0001), multifocal Candida colonization (OR: 23.55; 95% CI: 4.23-131.05; P < 0.0001), parenteral nutrition (PN) (OR: 63.67; 95% CI: 4.56-889.77; P = 0.002), and acute kidney injury (AKI) (OR: 7.67; 95% CI: 1.24-47.30; P = 0.028) as independent risk factors. A new prediction score with a cut-off value of 5.0 (80.3% sensitivity and 77.3% specificity) was formulated from the logit model equation.

Conclusions: Renal replacement therapy, AKI, PN, and multifocal Candida colonization were the independent risk factors for the new candidemia prediction score with high discriminatory performance and predictive accuracy.