Followed by Candida albicans, Candida glabrata ranks as the second major species contributing to invasive candidiasis. Given the higher medical burden and lower susceptibility to azoles in C. glabrata infections, identifying these infections is critical. From 2016 to 2021, patients with deep-seated candidiasis due to C. glabrata and non-glabrata Candida met the criteria to be enrolled in the study. Clinical data were randomly divided into training and validation cohorts. A predictive model and nomogram were constructed using R software based on the stepwise algorithm and logistic regression. The performance of the model was assessed by the area under the receiver operating characteristic curve and decision curve analysis (DCA). A total of 197 patients were included in the study, 134 of them infected with non-glabrata Candida and 63 with C. glabrata. The predictive model for C. glabrata infection consisted of gastrointestinal cancer, co-infected with bacteria, diabetes mellitus, and kidney dysfunction. The specificity was 84.1% and the sensitivity was 61.5% in the validation cohort when the cutoff value was set to the same as the training cohort. Based on the model, treatment for patients with a high-risk threshold was better than 'treatment for all' in DCA, while opting low-risk patients out of treatment was also better than 'treatment for none' in opt-out DCA. The predictive model provides a rapid method for judging the probability of infections due to C. glabrata and will be of benefit to clinicians making decisions about therapy strategies.
Keywords: Candida glabrata; candidiasis; nomogram; predictive model; risk factors.
© The Author(s) 2024. Published by Oxford University Press on behalf of The International Society for Human and Animal Mycology.