jeanfisch.bsky.social
Analysis, rationalism & objectivity are my sins
1,048 posts
1,489 followers
255 following
Getting Started
Active Commenter
comment in response to
post
Of course. There will be pockets of non infected
My question was a different one: is there a non trivial set of people whot get infx but later test negative on infx ABs?
comment in response to
post
Good point Marc even if the authors say that some did not have any infx ab
Is this possible on some scale?
(Sorry if question is naive - not a biologist / medic!)
comment in response to
post
Just for clarity: I took Musk here as an example of a political influencer with a huge following, please don't read anything beyond that in this example :-)
Here a review I posted on what I understand the findings to be (and not to be)
bsky.app/profile/jean...
comment in response to
post
@acritschristoph.bsky.social goes actually further and puts forward some indirect confounders around the time since the latest infection (which is shorter in those self-declaring post-vac syndrome and hence one would expect a higher residual response)
bsky.app/profile/acri...
comment in response to
post
Thanks very much, as said I sensed an honest attempt (at last!) but couldn't estimate what the conclusions were: You help here a lot!
Have a nice evening!
comment in response to
post
Here some insightful context on the study by @marcveld.bsky.social
If I understand correctly, it's a bit what I felt: the study is well done but, conclusion-wise, probably limited by the size of the sample and the confounders
bsky.app/profile/marc...
comment in response to
post
Thanks Marc, I was wondering what to do with the study
One the side they try to do things correctly (afaics without medic background) ie address the right questions no exagerated or spin wording, etc.
But then, somehow, they failed to put their conclusions into a formal (or informal) CI context
comment in response to
post
NB: My estimates for flu deaths per season for Germany are much higher than those calculated by RKI
One major reason is that I follow the UKHSA approach (ie any deaths beyond expected by a MOMO approach)
RKI only takes those deaths that are in excess of expected + 1z score
comment in response to
post
Quick reminder on covid: I use deaths with covid on the certificate as per death register (data shared by Destatis on a monthly) basis currently up to April 2024 and, a pro-rata estimate based on the weekly RKI deaths afterwards
comment in response to
post
Quick reminder: I estimated deaths triggered by flu using the UKHSA approach ie as
- excess deaths
- in weeks with high flu prevalence
- vs a MOMO like level of expected deaths (which is my estimate / Germany does not have a national MOMO-style mortality monitoring)
comment in response to
post
I write "could" because I read the paper three times and don't understand how they managed to extract a "long vax" signal from "long covid" (ie induced by an infection) but they say they did address it
Any insights welcome if other did understand!
Here the link: medrxiv.org/content/10.1...
2/2
comment in response to
post
Ah! Of course! I knew there was something I was missing
Thanks Stuart and have a nice evening!
comment in response to
post
in what sense do you see the rates of early 2025 as similar to 2023?
I see them as "being mid-way between 2024 and 2023" (which is also what I get) and that's a level which is roughly what one would expect from pre-pandemic trends so "not wild" but you make me curious!
comment in response to
post
Thanks for this Adam - Is the model you use here assuming a case to death distribution curve or a fixed ratio?
(I ask because I read somewhere that using a distri curve has an impact)
comment in response to
post
Merci Xavier pour cette mise au point factuelle
J'avais entendu l'affirmation de Thomas Dermine ce matin à la radio et je me disais qu'il devait y avoir un gros stud ou une extrapolation tangentielle du troisième degré :-)
comment in response to
post
Yes, I was sharing my thoughts as it might inspire some lines of questions to be asked
comment in response to
post
Aftertought: It's a bit bizarre that suddenly, a database with death notifications is no longer available ... so could it be that Stats NZ applied correction ratios to the data it shared on the covid website up to August 2024 ... and now only shares the raw data?
As said, just another thought
comment in response to
post
If this is confirmed, then one needs to make quite hefty corrections on the data by occurrence date (and it will take time to get a good feel of the IBNR ratios to apply as, I think NZ will share an updated set only every quarter right?)
Just a few thoughts
4/4
comment in response to
post
so deaths after COD has been assigned
One reason I think this is what happened is that the differential between registrations and occurrences is highest with <30 (where a high % go through coroners) and lower in the 80+ (where that % is low)
3/
comment in response to
post
That should only be a database query and either you have time to do the query or not
So my hunch is that, up to August 2020, the database somehow got access to deaths "notified" (so when a death is confirmed even if cause is not yet) ... and now it's only deaths registered
2/
comment in response to
post
Ah thanks, so it looks as the step change in the DIA happened after the covid website was terminated in August so that would fit!
However, the website says that the wls data shared yesterday is for any registration of death occurrences up to Jan 26
1/
comment in response to
post
The thing is that I checked each week the accumulation in 2024 and it happened quickly (100% in six weeks)
So did this austerity only kick in at the end of the year?
Were there maybe changes in death notifications?
comment in response to
post
Just a quick question: Do you understand why occurrences and registrations differ by 1.5% in 2024 while they never differed by more than 0.2% in any preceding year?
I had developed an IBNR which essentially saw changes to the death occurrences over 6 weeks up to August 2024
Strange
comment in response to
post
Hurray! Thanks for the heads up!
comment in response to
post
Thx David! I know the Wellkiwis data. Also another country provides co-infections (sorry don't know which) and it's rare indeed
There are two possible explanations for that rarity
- biological: it's medically difficult? (no idea)
- Physical: people down with one infx shield don't interact
comment in response to
post
Absolutely, I forgot about that option (c)!
comment in response to
post
Remember: In order to make a wave, a virus needs people:
a) getting susceptible
b) but remaining out of its reach for a while
If the virus can infect people as soon as they come out of their immunity zone, all you get is a flat "baseload" of infections
For a wave, you need a pool of susceptibles
comment in response to
post
So also in Germany, the RSV wave was earlier than usual in 2021
Usually, the RSV wave came in parallel with the flu wave, sometime around end of Jan
BUT
I still don't see how this could trigger 5% of deaths to "kick in earlier" unless everybody missed the fact that RSV is a massive killer :-)
comment in response to
post
My point is not about Trump nor about the mandate he believes he has and the actions he takes
My point is about democracies and how robust they are against leaders potentially believing that their cause / mandate gives them the right to bypass laws
I am not sure democracies are actually
comment in response to
post
Just to be clear: I am not into politics here beyond that I believe that democracy
- is the least bad system around
- lives from people electing their government which implements their wishes by
/ possibly changing the laws
/ if there is vast majority their constitution
comment in response to
post
Ah, I see what you want to do - I will have a look at Germany (but if ARI as a whole is not increasing earlier than usual how can one particular virus do it but indeed, let me check)
comment in response to
post
Exactly ... but then there is no raise in non-covid ARI earlier in the season
So back to square one ;-)
comment in response to
post
Indeed, that ILI / ARI kicked in earlier in 2021-2022-2023 because everybody skipped 1-2 years of respiratory infections was my main suspect ... but I could not confirm it based on the German data
The advantage of Germany was that all the required data was spread but available in open data :-)
comment in response to
post
Fair point on the fact that it hasn't to be the same cause the same year
comment in response to
post
One think I noted is that the "seeming shift in the winter peak" effect is bigger in countries in which the winter peak was late (so End Jan to Mar) ... I just mention it should it trigger a thought!
comment in response to
post
Thanks for the suggestions! I am unsure how adverse effects of NPIs could play three years in a row
Previous covid infections cannot play in 2021 (Germany hardly had any prevalence by end 2021)
Aaarrgghh, I wish I had an answer :-)
comment in response to
post
On vaccines:
Through NZE, one can prove that dose 1 and 2 generated no top-line mortality signal since they were administrated before covid AND in a respiratory calm period (ie there is no confounder)
But formally, I can not use this approach for boosters
Hence, I need to keep option (c)
comment in response to
post
So I am left with 3 theoretical options
a) under-ascertained covid deaths
b) deaths that were avoided in spring but bound to happen ie we kept a frail pop alive for longer ie more die in autumn from the same illness level / temperature
c) something around vaccines (and then it can only be boosters)
comment in response to
post
Yes, but the thing is that the start of the winter rise came earlier (so in sep and not oct)
- it's not the weather (the weather was normal)
- it can't be due to flu or RSV (they came later)
- there is no signal in general respiratory (AFAICS)
comment in response to
post
But my claim of having found evidence of non-covid respiratory shifts in the autumns of 2021-2022-2023 was not backed by the analysis so I erased it
Sorry for the inconvenience - I will look further to understand this "shift in the moment of peak"
END
comment in response to
post
However, what remains is a clear signal on deaths excl. covid
The winter growth started earlier in 2021-2022 and 2023 (irrespective of when there was the peak)
So the mystery remains: Is this under-reported covid or something else?
3/
comment in response to
post
RKI does provide ARI incidence with confirmed covid + test and if I use that figure "x2" (to mimic infection under-ascertainment), the signal of shift of ARI prevalence gets small - Take "2.5x" (fully possible) and the signal is gone
2/