Objectives
A candidaemia risk prediction model by Keighley et al. stratified patients with candidaemia into <20% or ≥20% predicted 30-day all-cause mortality based on age >65 years, intensive care unit admission, chronic organ dysfunction, no recent surgery, haematological malignancy, source of candidaemia, and prolonged antibiotic therapy. We aimed to validate this model in a contemporary patient cohort.
Methods
This retrospective cohort study was conducted at a 671-bed tertiary hospital comprising both abdominal solid organ and allogeneic stem cell transplant services (Melbourne, Australia). All adult inpatients with Candida spp. isolated from blood cultures from 2018 to 2023 were included. Model performance was evaluated using logistic regression and area under the receiver operating characteristic curve.
Results
A total of 121 patients with candidaemia were identified, of whom 40 (33%) died within 30 days. The median mortality risk score was 3 (interquartile range, 2.5–4.5). The risk score demonstrated good discriminative ability in predicting mortality in our cohort (area under the receiver operating characteristic curve, 0.759), with a score of 2 providing a discriminative cut off (11% mortality score ≤2, 43% mortality score >2, and negative predictive value 89.2%). Patients aged >65 years, intensive care unit admission, and a gastrointestinal or unknown source of candidaemia were independently associated with mortality.
Conclusions
The candidaemia mortality risk predictive model by Keighley et al. performed well in our contemporary and complex patient population. This model can be utilized to risk-stratify patients into low and high mortality risk to assist with candidaemia management and antifungal stewardship.