Last, the author considered diabetes, alcohol use, and chronic heart failure, among others, as potential confounders in the multivariable models. However, several criteria should be met for a variable to be identified as a potential confounder. One of the criteria is that the potential variable should not be in the pathway between exposure (ABO blood type) and outcome (ARDS). If it is in the pathway, this variable should be considered as a mediator, which cannot be adjusted in the models when total effect (direct and indirect effect) of ABO blood type on ARDS is the main research question. In this study, some variables, such as diabetes and chronic heart failure,4 may lie in the pathway between ABO blood type and ARDS, so they should not be adjusted in the final analysis. The authors can choose to draw a directed acyclic graph5 to select which variables should be included in the models. I anticipate that the effect of ABO blood type on ARDS would be stronger than reported in the article.