Corticosteroid Duration and CAPA Risk

All statistical analyses were conducted with JMP 16 software (SAS Corporation, Cary, North Carolina). Baseline characteristics and health outcomes were compared between corticosteroid duration groups using the χ², Fisher exact, or Wilcoxon rank-sum tests as appropriate. Steroid duration was assessed as an independent predictor of CAPA using logistic regression. First, univariate logistic regression was used to identify which baseline characteristics were associated with CAPA at a threshold of P < .2. Significant univariate covariates were then entered as covariates in a multivariate logistic regression model to generate adjusted odds ratios (ORs) and 95% confidence intervals (CIs). Covariates that achieved P < .05 were considered significant independent predictors of CAPA in the final model. CAPA was also assessed as an independent predictor of mortality using a multivariate logistic regression model that included the following variates: age, steroid duration, salvage therapies for ARDS, PaO₂/FiO₂ ratio, SOFA score, secondary infection, and duration of ventilation. Additional analyses were conducted to examine the relationship between corticosteroid duration as a continuous variable and CAPA risk. Median total steroid duration was calculated for those who did and did not develop CAPA and was compared between CAPA groups using the Wilcoxon rank-sum test. The duration of steroids that most accurately predicted CAPA was assessed using a logistic regression model with CAPA as the dependent variable and steroid duration as the independent variable to generate the area under the receiver operating characteristic (ROC) curve (AUC). The cutoff was determined by the duration with the highest positive likelihood ratio (ie, sensitivity – [1-specificity]).
The study was approved by the University of Texas Health Science Center at San Antonio institutional review board and the University Health research department (HSC20200207EX).