Development and Validation of a Risk Score for Predicting Invasive Candidiasis in Intensive Care Unit Patients by Incorporating Clinical Risk Factors and Lymphocyte Subtyping

OBJECTIVE: To develop and validate a rapid invasive candidiasis (IC)-predictive risk score in intensive care unit (ICU) patients by incorporating clinical risk factors and parameters of lymphocyte subtyping. METHODS: A prospective cohort study of 1054 consecutive patients admitted to ICU was performed. We assessed the clinical characteristics and parameters of lymphocyte subtyping at the onset of clinical signs of infection and their potential influence on IC diagnosis. A risk score for early diagnosis of IC was developed and validated based on a logistic regression model. RESULTS: Sixty-nine patients (6.5%) had IC. Patients in the cohort (N=1054) were randomly divided into a development (n=703) or validation (n=351) cohorts. Multivariate logistic regression identified that CD8+ T-cell count ≤143 cells/mm(3), receipt of high-dose corticosteroids (dose ≥50 mg prednisolone equivalent), receipt of carbapenem/tigecycline, APACHE II score≥15, (1,3)-β-D-glucan (BDG) positivity and emergency gastrointestinal/hepatobiliary (GIT/HPB) surgery were significantly related with IC. IC risk score was calculated using the following formula: CD8+ T-cell count ≤143 cells/mm(3) + receipt of high-dose corticosteroids + receipt of carbapenem/tigecycline + APACHE II score ≥15 + BDG positivity + emergency GIT/HPB surgery ×2. The risk scoring system had good discrimination and calibration with area under the receiver operating characteristic (AUROC) curve of 0.820 and 0.807, and a non-significant Hosmer-Lemeshow test P=0.356 and P=0.531 in the development and validation cohorts, respectively. We categorized patients into three groups according to risk score: low risk (0-2 points), moderate risk (3-4 points) and high risk (5-7 points). IC risk was highly and positively associated with risk score (Pearson contingency coefficient=0.852, P for trend=0.007). Candida score had a moderate predicting efficacy for early IC diagnosis. The AUROC of the risk score was significantly larger than that of Candida score (0.820 versus 0.711, Z=2.013, P=0.044). CONCLUSIONS: The predictive scoring system, which used both clinical factors and CD8+ T cell count, served as a clinically useful predictive model for rapid IC diagnosis in this cohort of ICU patients. CLINICAL TRIAL REGISTRATION: chictr.org.cn, identifier ChiCTR-ROC-17010750.