2. While I 've not spent enough time with the raw data to be willing to make absolute claims, I suspect this apparent decline may be an artifact; I think we can explain why; and I think this needs to be looked at seriously.
@jsm2334 laid out the argument:https://www.covid-datascience.com/post/israeli-data-how-can-efficacy-vs-severe-disease-be-strong-when-60-of-hospitalized-are-vaccinated …
-
Show this thread
-
3. Thanks also to
@dylanhmorris who brought it to my attention, and@jbakcoleman and@BillHanage for discussion. The basic story is this: Looking at active severe cases in Israel we have 16.4/100K among unvaxxed and 5.3/100K among vaxxed. That looks like a mere 67.5% VE. But…3 replies 47 retweets 486 likesShow this thread -
4. This results are confounded by age. Irrespective of disease status most unvaxxed are young; irrespective of vaccine status most severe disease cases are old. This creates counterintuitive statistical result. VE overall is far lower than VE in under 50 or in people over 50.pic.twitter.com/CAI0vVgIAk
36 replies 302 retweets 1,462 likesShow this thread -
5. You may recognize this as an instance of Simpson's paradox. In this case, both cohorts have a much higher VE than the VE of the two cohorts aggregated together. This seems paradoxical, but results from the confounding described above.
9 replies 103 retweets 936 likesShow this thread -
6. I believe it also explains the apparent drop in VE over time. When % vaxxed is low across the board, the association between being unvaxxed and being young is weak. As vaccine coverage increases to Israel levels, that association strengthens, driving the Simpson's paradox.
6 replies 53 retweets 574 likesShow this thread -
7. As I mentioned, I have not spent enough time with the data to be 100% confident that this explains everything happening in Israel. That said, it does seem to be a strong alternative hypothesis to vaccine escape / waning, and merits serious consideration.
40 replies 50 retweets 719 likesShow this thread -
I appreciate the lively discussion. While I linked to him it in the second post and quoted from his blog subsequently, I want to again stress that this is
@jsm2334’s observation. I came across it his blog. Please give him a follow for interesting and insightful COVID analysis.10 replies 15 retweets 337 likesShow this thread -
Replying to @CT_Bergstrom @jsm2334
Certainly this kind of bias merits consideration. I think the particular figure cited in that table is an example of Simpson's paradox, which is a special type of confounding.
5 replies 10 retweets 42 likes -
The main confounder here is not age, not sex, it is SES. The spread in Israel until mid-July was mainly in highly vaccinated areas with high SES. Time of vaccination is also correlated with SES. The adjustment to weeks actually confounds the results.
4 replies 4 retweets 15 likes -
Replying to @dvir_a @mlipsitch and
Dvir Aran Retweeted Yair Lewis
There are dozens of posts on this issue. You can find some info here.https://twitter.com/_lewisy/status/1427741690098552832 …
Dvir Aran added,
Yair Lewis @_lewisyReplying to @_lewisy @thehowie @segal_eranWaning immunity in Israel? Apropos debate about 3rd dose. This is based on the public data, and is similar to the one distributed today by the MoH; taken by face value, this does suggest those who got vax earlier are at higher risk for infection. BUT some important caveats >>> pic.twitter.com/zMfYe23VgJ1 reply 1 retweet 3 likes
Also assuming urban/rural differences between who got vaccinated earlier and where Delta is spreading now—assume both would be more urban first, and pull in the direction observed thus overstated effect. Today's MMWR on NY hospitalizations was interesting & fairly reassuring tbh.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.