a further look at the "partially vaxxed" population as a predictor of covid deaths

southeast asia: adding some confirmation and some questions.

yesterday we spoke about this remarkable alignment of the number of people who have one (but not two) doses of a covid vaccine and covid deaths following shortly thereafter.

we first looked at palestine: (pawlestine, according to the endlessly hilarious “islamicat”)

and this led me to seek similar signals using the same metric:

it shows up with near identical signature in israel during their initial vaccine push:

and, also in the world leading israeli push for boosters. (i could not synthesize the same metric here and so used “the number of people who had received a booster in the las 14 days” instead as it should amount to the same thing. (and provides both the best fit and the best alignment with what we, a priori, believe to be the underlying biology)

this is enough coincidence to really start wondering what’s going on here. note that both israel and palestine are using mRNA vaccines.

to assess further, i jumped to a new region in which vaccine programs are recent: southeast asia.

it’s always good to start from the simplest data and work into the more manipulated to you keep a sense of what you’re looking at:

we can see 2 different start times for the surge in deaths. provocatively, they align with different start times for vaccines. even the vaccine surge and then cessation in viet nam seems mirrored by the later deaths surge and drop and they had basically no deaths prior to vaccines though whether they were looking for them is an open question.

full region looks like this.

all in all, this looks quite fertile for inquiry.

thailand provided a near perfect replication of what we saw in the levant.

of interest, they largely did NOT use mRNA vaccines, choosing to use mostly astra zenca (AZ) instead. this is a modified adenovirus vaccine and works similarly to mRNA by inducing cells to produce covid spike proteins to elicit immune adaptation. this seemingly identical relationship may indicate that the AZ vaccine induces a similar “worry window” to the mRNA vaccines (i have not seen any specific clinical evidence here. if anyone has, please link in the comments. i’d love to see it.)

where the plot starts to thicken is in the rest of southeast asia and, frankly, the data is getting a little weird and hard to parse as we see a different sets of vaccine programs used.

cambodia may provide a really interesting test case of vaccine variance. back on march 2, they got 324k doses of the AZ vaccine with 1.1 million to follow by may. but then, they switched to sinovac and sinopharm and have received over 23 million doses, so they are nearly fully chinese vaccines at this point. this is interesting because these vaccines are meaningfully different from others and purport to be based on actual inactivated coronavirus. so they look little like pfizer, moderna, JnJ, or AZ.

what we see is sort of striking one you realize the switch that occurred.

the first 2 humps in the vaxxed population in april and may are from AZ. both are followed quite rapidly or align with by rises in deaths, just as in other places. it’s easier to see if we zoom in:

but after that, the series decouple and vaccination becomes a follower of death rate. it does not look like size of of single dosed population in the chinese vaccines predicts forward deaths.

(note that this says nothing about their efficacy and i have no idea what it actually is. sugar water would decouple too. so would the world’s most perfect vaccine with no side effects or immunosuppression. we simply cannot speak to that question from this data.)

but there is one more feature here that i find extremely interesting. let’s zoom in on only the VERY recent past:

note the sudden re-alignment where a rise in the single vaxx population leads a sudden (and unexpected as no one else in the region is showing this rise, all are in decline (see above)) increase in deaths.

this could be coincidence and the simple result of minds seeking to impose patterns on randomness. but it could be something else as well. it could be a VERY strong signal IF (and i need to stress “if” here, it turns out that this recent round of vaccinations were AZ or mRNA.

and there is reason to suspect it may be. cambodia has started using AZ boosters mixed with chinese vaccines to “aid resistance.” and it just started recently. in addition, the US sent about a million doses of JNJ. so the potential is certainly there for this recent vaccine surge to be boosters and first doses of modified adenovirus vaccines that we saw predict later deaths so well in thailand.

if we can find a way to pin this down beyond the circumstantial, it would provide quite an interesting data point.

from here, the data gets kind of messy. laos is simply unusable. their reporting of deaths is so non-existent and implausible as to render meaningful analysis impossible. i tried to work with their data but simply gave up. the quality is not there.

vietnam (using AZ and mRNA) is, frankly, just weird.

either their vaccination data reporting is off or they are using a highly differentiated model for vaccination that seems like “get everyone one dose before giving out second doses”. unlike everyone else i’ve graphed, their “one dose but not 2” cohort just keeps growing.

it’s also possible they are using something uncommon like a 6 week interval between doses. this data on the right makes that look possible.

this puts us into an odd weird paradigm very different from others and renders the “one dose but not 2” unfit as we have strong reason from clinical studies to believe in a 2 week risk window and not the 6ish weeks this one would be measuring.

while i dislike jumping around on metrics, it seems the only sensible thing to do here. it will make this a bit of an asterisk and i just have less overall confidence here.

we get this:

this shows some of the patterns we’ve seen elsewhere, but not others. there is linkage with vaccine dosing and later spikes, even in that first spike on 7/7 and the little deaths echo on 7/18-22. but it then becomes a following series which seems counter to or at least less supportive of the premise.

tantalizingly, it then leads on the peak, with cohort size peaking on 8/18-22 ahead of deaths peak on 8/28. however, this recent surge has led to nothing and has not disturbed the gompertz pattern at all.

so, all in all, i’m really not sure what to make of vietnam.

it’s possible that this current surge of vaccination is in a less vulnerable population and that we’ve been looking at a lot of “first surges” in the highly vulnerable and thus getting a very different deaths signal.

it’s also possible that this new surge has just now shown up in deaths yet and that we may see an upturn as in cambodia.

there is also an possibility that the fully vaxxed rate is having an effect, though this does not seem to be the case in cambodia (of course, they used mostly chinese vaxxes about which many have significant doubts, so perhaps that is telling too?)

we’re also using a different metric that i think should be similar but that may have injected new data issues, lags, and transpositions, so caution is warranted.

going to keep looking at more of these with a focus on mRNA and adenovirus vaxx countries that reside predominantly in one climate zone/seasonal pattern. first vaccination surges and surges of boosters look most interesting to me.

open to ideas on where to go next of you have some good ideas.

also very curious as to whether anyone knows what the current vaccination surge in cambodia is, who is getting vaccinated, and what vaccine is being used. if anyone has that info, please share.

this puzzle is getting pretty interesting.