Re:Pfizer's oral Covid-19 antiviral cuts hospitalization, de...
by tzs on Friday November 05, 2021 @11:35AM
(#29120772)
When both the risks and vaccination rates are significantly different across demographic groups statistics for the whole population are often nearly useless due to Simpson's paradox.
Depending on how we define severe COVID you link is showing between 36% and 44% of the severe COVID patients at the hospital are vaccinated.
That's similar to what they have seen in Israel. As of about a month ago they were seeing about 60% of their severe cases were in vaccinated people.
Sounds pretty bad for vaccines, right? It does until you remember Simpson's paradox and take a finer look at the data [1]. It turns out that the Israel data showed in each age group efficacy against severe COVID ranging from 81.1% to 100%, with above 92% in all the 10 year age groups under 60 and still above 88% is the 10 years groups through 80.
It is very likely a similar thing is going on at the hospital whose data you linked to. That's been the case for every place I've come across in the US that published breakdowns of the stats by age group.
Depending on how we define severe COVID you link is showing between 36% and 44% of the severe COVID patients at the hospital are vaccinated.
That's similar to what they have seen in Israel. As of about a month ago they were seeing about 60% of their severe cases were in vaccinated people.
Sounds pretty bad for vaccines, right? It does until you remember Simpson's paradox and take a finer look at the data [1]. It turns out that the Israel data showed in each age group efficacy against severe COVID ranging from 81.1% to 100%, with above 92% in all the 10 year age groups under 60 and still above 88% is the 10 years groups through 80.
It is very likely a similar thing is going on at the hospital whose data you linked to. That's been the case for every place I've come across in the US that published breakdowns of the stats by age group.
[1] https://www.covid-datascience.com/post/israeli-data-how-can-...