Thread by @MatanHolzer

Last week's update

#VaccineEfficiency
#DeltaVariant
#Israel

Methodology and sources:

upper figure - cases for vax and unvax. The two rightmost columns here
data.gov.il/dataset/covid-…

Then, vax status.
Go here (data pulled from Israeli MoH "dashboard").
github.com/dancarmoz/isra…
Sum accumulated # 2nd vax number 20 days ago over 20-59 age groups. let's call it A.
Sum pop over 20-59 age groups, let's call it B.
A/B is %vax 2nd dose >= days ago.

From all pop (B), subtract the number of those who got at least 1 dose. That's the number of people who didn't take any dose. Call it C.
C/B is unvax percent.

You'll get the middle figure here:

As you can see, it doesn't sum up to 100%. So there are about 5% of population who got 1 dose only. It's quite constant over time, so it's not people in middle of the process.

Supposedly, VE calculations are valid, as compare unvax (at all) to vax also in cases.
BUT: >>

The problem is that there can be people who are unvax (at all) because they have been recovered. Of course those people are immune. Hence we shouldn't include them in unvax group.
So we should remove recovered&unvax from unvax group.
How many of them are they?

I don't have formal, exact, per date and age-stratified numbers (even asked the manager of database, they replied they don't have those data).

But from media publications it turns out that 30%-50% of recovered people took 1 dose till late April (when vaccination rate saturated for this age group).
ynet.co.il/health/article…
So subtract 70%-50% (respectively) of those recovered from unvax group. That % of recovered&unvax.

To make it simple:
About 10% recovered.

If half of them took one dose, so 50% haven't, then 5% of population are unvax&recovered. That's the drop from 17% (in the naive figure that doesn't take into account recovery status) to 12%.

And if 30% of recovered take 1 dose, and 70% haven't, then 10% recovered includes 7% of population who are unvax&recovered. That's a drop from 17% to 10%

But we can, and should, take something very important from the naive figure, where I didn't take recovery status into account:
VE <= 0
That is, compared to a group with large portion of recovered, they are same / slightly more immune >>

And this is "average" immunity between recovered and not recovered.
What if all of them were recovered?
It will be much more negative.

Conclusion:

Recovered people are much more immune for June "wave" than vaccinated people!

Facing the virus, the immunity is much more long-lasting and robust (against new variants)

Of course there many possible biases, I discussed in details here

https://twitter.com/MatanHolzer/status/1410312813260718081?s=19

Question is, weren't those possible biases applied for the conclusions about VE on February through May?
I mean, it wasn't RCT as well... They were based on the same raw numbers of cases. Regardless of testing rate, intrinsic differences between groups..

https://twitter.com/MatanHolzer/status/1412143469561327618?s=19

Even the experts from @HebrewU noticed that decline in efficiency.

They throw the number "64% efficiency" - they aggregated data over June, when the efficiency is continuously declining, so they got some number in the middle
ynetnews.com/health_science…

The number they got in the middle

Israeli MoH telegram channel [1] published
bottom left figure (all ages).
%vax in cases very similar to %vax in population! while %recovered in cases is 9 times lower than their % in population (835K out of 9.2M, 9%) >>

t.me/s/MOHreport
Look for the file תמצ קורונה

So

* immunity of recovered is much more stronger, long lasting and robust, than that of vaccination.

* should reduce few cases in unvax group (those unvax & recovered, b/c I compare to unvax & not recovered), thus slightly decrease base rate and VE, but I think it's negligible.

Look at the top right graph.
Does the reason that % of ultra-orthodox people in cases (1%) is much lower than their % in population (about 12%), is the high percent of recovered there?

(officially 30% of them are confirmed cases. The true number can be easily 50%)

Thanks @RanIsraeli for translating the titles and labels

Of course those numbers have many biases. e.g. if the ultra-orthodox stopped collaborating with mass testing. We'll be smarter on the next month, to see if it's really not spreading in their communities (no hospitalizations and deaths)