1/
Peer review (unsolicited)
Another analysis of Israel data on vaccine effectiveness. Fully vaccinated vs. unvaccinated
Observation period: Jan 24-Apr 3
Bottom line: again, likely over-estimation of effectiveness.
thelancet.com/journals/lance…
Peer review (unsolicited)
Another analysis of Israel data on vaccine effectiveness. Fully vaccinated vs. unvaccinated
Observation period: Jan 24-Apr 3
Bottom line: again, likely over-estimation of effectiveness.
thelancet.com/journals/lance…
2/
Over the study period:
1)People shifted from unvaccinated to fully vaccinated.
2)Rate of infection was declining (peak Jan 20)
Over the study period:
1)People shifted from unvaccinated to fully vaccinated.
2)Rate of infection was declining (peak Jan 20)
3/
Distribution of unvaccinated on the calendar is shifted to the left (time of HIGHER infection rate)
Distribution of vaccinated is shifted to the right (time of LOWER infection rate)
So, part of the difference between vaccinated & unvaccinated is due natural wave decline.
Distribution of unvaccinated on the calendar is shifted to the left (time of HIGHER infection rate)
Distribution of vaccinated is shifted to the right (time of LOWER infection rate)
So, part of the difference between vaccinated & unvaccinated is due natural wave decline.
4/
Authors try to attribute the decline to vaccination.
Unconvincing.
All waves (and previous wave In Israel) follow that trajectory. Peak & decline.
Certainly, cannot attribute the beginning of the decline to vaccination rates.
Bar graph shows weekly cases.
Authors try to attribute the decline to vaccination.
Unconvincing.
All waves (and previous wave In Israel) follow that trajectory. Peak & decline.
Certainly, cannot attribute the beginning of the decline to vaccination rates.
Bar graph shows weekly cases.
5/
Of course, authors were aware of time trends. Model included calendar week as covariate.
BUT, see table from article. Estimates of effectiveness are nearly identical (unadjusted vs. adjustment), implies no confounding by the wave trend (or anything else). That’s unexpected.
Of course, authors were aware of time trends. Model included calendar week as covariate.
BUT, see table from article. Estimates of effectiveness are nearly identical (unadjusted vs. adjustment), implies no confounding by the wave trend (or anything else). That’s unexpected.
6/
(Note that vaccinated were older, and older vaccinated earlier, but unadjusted/adjusted are similar in each age group).
(Note that vaccinated were older, and older vaccinated earlier, but unadjusted/adjusted are similar in each age group).
7/
What is the explanation?
Either no confounding, somehow
or
Modeling calendar week as a covariate did not take care of confounding by the time trend.
Most likely the second.
What is the explanation?
Either no confounding, somehow
or
Modeling calendar week as a covariate did not take care of confounding by the time trend.
Most likely the second.
8/
Here is someone’s better analysis, accounting for time trend.
lockdownsceptics.org/2021/05/07/stu…
Lastly,
1)What happened in days 1-14 of the 1st dose?
2)Again, no estimates for all-cause deaths.
Here is someone’s better analysis, accounting for time trend.
lockdownsceptics.org/2021/05/07/stu…
Lastly,
1)What happened in days 1-14 of the 1st dose?
2)Again, no estimates for all-cause deaths.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
