Citat:
Ursprungligen postat av
OUFCompulsive
Again, you have misunderstood me. It’s not that the estimate has risen from <0.01%, but the fact they have raised it several times — not to mention that despite their antibody results being skewed high, they still do not match a 0.6% IFR. Their latest would indicate a 0.96%~ IFR before even adjusting for test parameters and taking into account the skewed data set.
The weighted 5% comes from their antibody results release from four days ago. The FHM have extrapolated this result, not me. But you are correct in that the time to antibodies forming is faster than the time to death, and even this should be adjusted for when calculating the true-IFR, alongside Bayesian inference models to adjust prevalence rates to the test parameters, and keeping in mind that the data set will be skewed high as they are samples from vårdcentralen.
So, let me do this quickly for you. Assuming they only measured IgG antibodies (the slowest to form), we would have to measure results against the 31st of last month (majority of antibodies form after 14 days (95%)). Without digging into the excess deaths too greatly we can assume that they remain constant at 1,000 (although this is, in fact, an undercount).
Further adjusting for the test parameters, which have not been given in the report — but assuming the current-best use parameters previously given (99%~ sensitivity/specificity) we arrive at a death total of 5,400~ (rounded downwards) and a nationwide seroprevalence of 4%.
4% of 10,330,000 is 413,000 rounded.
5400 / 413000 = 1.3% IFR
Now, we can also look at these statistics and assume that the result has already been weighted for false-positives and the result is extrapolated as such. This would give us:
5400 / 517000 (rounded up) = 1.04% IFR
But this death count in this manner is an undercount due to excess mortality being higher than 1,000 deaths and difficult to predict exactly due to the all-cause mortality falling at a time where excess mortality spiked. The FHM admit that excess mortality is not factored into their IFR estimates, or the fact that deaths always trail antibodies forming, or the fact that their data sample sets are skewed high (although they have admitted the sample sets are biased).
Kul att du anstränger dig men det gäller ju att man stoppar in rätt siffror. Var får du siffran 5% ifrån?
"Insamlingsperioden 18 -24 maj (vecka 21)
Analys visar en förekomst av antikroppar på 6,3 % (95% KI 4,76-8,24) "
"Andelen antikroppspositiva var fortfarande högst i Stockholm (10,0%; 95% KI 6,71-14,20), medan andelen i Skåne var 4,5% (95% KI 1,75-9,29) och i Västra Götaland 2,7% (95% KI 0,88-6,18)."
https://www.folkhalsomyndigheten.se/publicerat-material/publikationsarkiv/p/pavisning-av-antikroppar-efter-genomgangen-covid-19-i-blodprov-fran-oppenvarden-delrapport-1/
Räknar jag lite snabbt på Sthlm enbart (2.344 miljoner innevånare, <1936 döda), så får jag IFR till <0.8% per 24 maj. Västra Götaland och Malmö kommer att ligga under det värdet.
Att betydligt fler varit smittade än de som visar positivt i antikroppstest kan man utgå ifrån så 0.8% blir den högsta skattningen av IFR som är aktuell.
Fast man kan ju inte bara hugga en siffra mitt i konfidensintervallet utan att redovisa osäkerheten...