Thread by @dperetti

Are reported deaths in VAERS coincidental?

⤵️A thread, with data visualisations. ✨

According to REUTERS fact checking:
« Claims of widespread death due to COVID-19 vaccines reported on VAERS are false. Entries in VAERS do not prove causality. »
reuters.com/article/factch…

To put it another way, the reported deaths are said to be purely coincidental, and totally unrelated to the injections.

☝️But if that statement is 100% true (ie the "death" events and the "injection" events are not related), an homogeneous distribution of deaths should be observed, regardless of the subset of data we look at.
No vaccine « fingerprint », or dose related pattern should appear.

In VAERS, we are lucky to have information about the vaccine manufacturer, but also the dose number and the number of days between injection and death (NUMDAYS).

That parameter is interesting because it’s intrinsic (as opposed to the number of injections of vaccine which would be difficult to retrieve and link reliably).

What’s more, if the injection is potentially lethal, a small NUMDAYS should be an indicator of its dangerousness.

Let's first plot the histograms of each vaccine / dose number using age and NUMDAYS < 28 days.

Can you spot the difference between Pfizer first dose and second dose ?

The difference between Moderna dose #1 and dose #2 is more striking and illustrates that most deaths related to the second dose of Moderna occur more quickly after the injection.

In order to make sure this is not an illusion due to the different number of cases (1166 vs 761 in the previous example), I have written a function that randomises the subsets, keeping their cardinality, but using data from any vaccine.

See how the different histograms in the bottom row all look the same.
At the bottom, this is what it would look like if the data was really unrelated to the injection.

Let’s do it again, but no longer taking age into account. See the difference between the real data (top) and the randomised data (bottom).

This made me feel like plotting the median of NUMDAYS.

Here is what we get for Pfizer dose 1 vs dose 2.

And here is what we get with *randomised* data (respecting the size of each age group):

Now Moderna vs its randomised counterpart:

Now all vaccines together.
Again, the bottom chart represents what it would look like if no relationship existed between the injection and the vaccine manufacturer / dose.

It’s interesting to note that:
1. Janssen turns out to be the less dangerous.
2. Lethality appears to be correlated with the total quantity of mRNA (= expressed spike protein) (Pfizer 30mcg, Moderna 100mcg) received.

Afterthought: maybe Janssen is not less dangerous and it just takes more time "by design".

https://twitter.com/dperetti/status/1453822857918681088?s=20