Development of a new clinical prediction rule for candidaemia

Candidaemia is a leading cause of hospital-acquired bloodstream infections, especially in intensive care unit (ICU) patients, and is associated with high mortality and healthcare costs. Prediction rules aid in the identification of patients at risk of developing candidaemia. This study aimed to develop a prediction rule for candidaemia within our population and evaluate its performance against two established rules: the Ostrosky-2 and the Nebraska Medical Center (NMC). We conducted a case-control study at a tertiary care center in Mexico City. Data were collected from candidaemia patients and controls matched by sex, age, admission area, and prior hospitalization at a 1:2 ratio, covering the period from 2018 to 2024. Significant risk factors for candidaemia were identified via multivariate logistic regression. By calculating the logistic regression coefficients and the final intercept (B0), we developed the “CandID” rule. The sensitivity, specificity, and predictive values of this rule were compared with those of existing rules. We included 146 cases and 292 controls. Abdominal surgery, broad-spectrum antibiotics, candiduria, cirrhosis, central venous catheterization, and solid tumors were identified as risk factors for candidemia in the multivariate analysis and were incorporated into the CandID rule. A cutoff value of ≥ 1.92 demonstrated 80% sensitivity, 63% specificity, a positive predictive value of 52%, and a negative predictive value of 86%, with an area under the curve (AUC) of 0.78. When applied to this population, the Ostrosky-2 rule had a sensitivity of 55%, a specificity of 69%, and an AUC of 0.5, whereas the NMC rule had values of 51%, 72%, and 0.61, respectively. The CandID rule showed the best performance in predicting candidaemia in our population. However, the single-center retrospective design and internal validation may limit generalizability; prospective multicenter validation is needed.