Hello, Hive world!
I'm (the) Mikky, a data analyst from Israel. I was inspired to make this post because of 
As someone who has swum in the sea of Israeli data over and over again, I feel the need to share my 0.02 HIVE.
Here I analyze a strange anomaly in the vaccine efficacy over time, specifically those whose vaccines expired.
If I find that this post peaks interest, I'll post more analyses. Comments, suggestions and criticism always welcome.
Foreword
To try to defeat as many biases as possible,I took the already normalized data from the MoH (Ministry of Health) dashboard (link below) by:
- Vaccination status (unvaxxed, vaxxed + expired (6 months), vaxxed + not expired)
No data on recoveries. These are data limitations, unfortunately - Only 60+ to avoid age-dependent bias
I know that 60 is not 80 is not 100, but these are the data I was given - Normalized per population size of each vaccination group
This is a huge pain in the *** to do on your own as the population sizes constantly change, so I'm glad that the MoH did this for me
Side notes:
- We started vaccinating boosters around 1st of August.
- I look at 2-week averages to get a cleaner signal, because any individual day doesn't tell us a general trend.
Back to the anomaly
Look here - notice that those whose vaccine expired reach a peak around end of August and then drop rapidly, while the unvaccinated keep getting severely ill?
This means that per capita, the expired-vaccinated start to get severely ill at lower rates, while the unvaccinated keep getting severely ill more and more.
One can't say that this has to do with the vaccine making their sick stays shorter while the unvaccinated linger - because this graph shows how many people were new admissions to severe status.
Vaccine Efficacy
OK, that's odd - but let's see how this reflects the status of VE (Vaccine Efficacy, sold to us as 95% effective) over time for expired vaccines.
For those who don't know, a 95% VE works something like:
If out of 100 unvaccinated people, 20 will get severely ill, then at 95% VE we expect that 100 vaccinated people, only 1 will get severely ill. At 90% VE, 2 severely ill, 50% VE, 10, etc.
Notice the extreme drops in VE (blue/orange lines), followed by an uptick?
Vaccine efficacy shouldn't be such a time-dependent variable - or at least, it should follow a logical pattern over time.
Drop in non-expired VE (blue)
I am somewhat willing to forgive the drop of non-expired vaccinated when there were near-to-no patients in the wards, making statistical analyses flimsy at best.
Also, these "not expired" vaccinated patients may have been 5.5 months after their dose, so the VE will have dropped quite a bit by then.
Drop in expired VE (orange)
This makes no sense. Israel started giving boosters, so in raw figures, sure - there are less people who have expired vaccines, are better protected, and therefore less likely to get severely ill.
But at rates normalized for population size?
The (absurd) explanation would be that people getting boosters somehow make the expired vaccinated less likely to get ill. One could say "herd immunity!", but then we would see this drop in the unvaccinated as well, but we see the opposite.
Conclusion
My most optimistic explanation is that everyone who's at risk got boosters, whereas people who are healthier decided to forego the booster, meaning that instead of a pool of "everyone", the expired vaccinated became a pool of "healthy people who don't need a booster" (or say, recovered). I find this highly unlikely but can't prove or disprove it due to lack of data.
My pessimistic explanation is that the numbers are being toyed with. Perhaps criteria of severely ill differ between vaccination groups, perhaps they started pooling unvaccinated car accidents that tested positive for Covid as "severely ill". I really don't know - but I find these findings to be really bizarre and inexplicable.
Appendix
I invite anyone who wishes to, to use the following MoH resources if they choose, though the language barrier could be a problem. PM or reply if you want translations of terms.
- https://datadashboard.health.gov.il/COVID-19/general
MoH dashboard. You can download CSV versions of most graphs here - https://data.gov.il/dataset/covid-19
MoH corona data. Has some additional interesting CSVs including my favourite "event-among-vaccinated.csv" which divides hospitalisations + deaths by age, vaccination status (0-1-2-3 doses), per week.
Note: No data on recoveries in that file.