Sensitivity analyses
We used the following statistical methods to conduct sensitivity
analyses to check the robustness of our findings. First, we performed
traditional multivariable regression analysis using a generalized linear
model. In this analysis, we created a generalized linear model using
outcomes as dependent variable and glucocorticoids on the day of
admission and all covariates as independent variables. Second, we
performed a propensity score adjustment analysis. In this analysis, we
created a generalized linear model using outcomes as dependent variables
and estimated propensity score in the main analysis as independent
variables. Third, we conducted an inverse probability of treatment
weighting analysis.12 For this, we used a weighted
generalized linear model with stabilized average treatment effect weight
calculated from the estimated propensity scores in the main analysis.
Continuous variables are presented as mean and standard deviation (SD),
and categorical variables as number and percentage. p<0.05 was
defined as denoting statistical significance. We used Stata version 16.0
(StataCorp, College Station, TX, USA) to perform all statistical
analyses.