This is a great idea to skeptically analyse the studies promoted in Ärzteblatt. As you've pointed out, many doctors rely on this "mainstream" newsletter to inform them about developments which they don't have time (or motivation?) to investigate more thoroughly themselves. (it's a bit like one of those Testsieger magazines which we all know are so impartial ;)
Cox regression assumes that the influences of the independent variables are constant over time.
They write, "A Cox proportional hazards regression model with time-dependent covariates was used to estimate the association between the bivalent vaccine and hospitalization due to COVID-19 while adjusting for demographic factors and coexisting illnesses."
Fine, the model handles time-dependent variables, but does it handle their time-dependent influences also? My guess is no, and I bet they didn't check.
This is a great idea to skeptically analyse the studies promoted in Ärzteblatt. As you've pointed out, many doctors rely on this "mainstream" newsletter to inform them about developments which they don't have time (or motivation?) to investigate more thoroughly themselves. (it's a bit like one of those Testsieger magazines which we all know are so impartial ;)
Cox regression assumes that the influences of the independent variables are constant over time.
They write, "A Cox proportional hazards regression model with time-dependent covariates was used to estimate the association between the bivalent vaccine and hospitalization due to COVID-19 while adjusting for demographic factors and coexisting illnesses."
Fine, the model handles time-dependent variables, but does it handle their time-dependent influences also? My guess is no, and I bet they didn't check.
Of course they didn't. I mean, it is kind of ridiculous to use "days since the most recent vaccination" as time-invariant influence. :)
I absolutely agree with this paper.