The ANOVA was based on hospitalizations, not infection, etc.
Nah. Hospitalization was simply an inclusion criteria not a variable.
The other two inclusion criteria were respiratory failure and a positive SARS-CoV-2 test.
Their variables were blood type of course and severity class, which was determined by the maximum level of respiratory assist given - supplemental oxygen therapy only or noninvasive ventilatory support or invasive ventilatory support or extracorporeal membrane oxygenation.
I think I see what they are saying now. Since all of their study group patients, excluding control patients, were hospitalized and already had severe disease defined as respiratory failure, their study design was not looking at odds of infection or odds of severe illness. All their patients were already infected and severely ill.
Bottom line - Once they are severely ill - Type A seems to get more severe severity and Type O seems to get less severe severity.
The article about the paper was just poorly written.