Read on Twitter
New #COVID19 serosurvey in Italy. Report in Italian so I couldn't read it carefully. Nice thread from @stats_q with some details.
Some @nataliedean @AdamJKucharski @isabelrodbar are questioning results b/c they seem low, but deaths & IFR actually consistent w/ other data.
Thread https://twitter.com/stats_q/status/1290330189810081792
Some @nataliedean @AdamJKucharski @isabelrodbar are questioning results b/c they seem low, but deaths & IFR actually consistent w/ other data.
Thread https://twitter.com/stats_q/status/1290330189810081792
Background
Italy & Spain were worst early hit countries by COVID19 w/ reports of overloaded hospitals & at the time, huge daily deaths reaching almost 1000/day. Epidemics in US, Brazil & Mexico have sadly made these epidemics seem small in comparison.
Italy & Spain were worst early hit countries by COVID19 w/ reports of overloaded hospitals & at the time, huge daily deaths reaching almost 1000/day. Epidemics in US, Brazil & Mexico have sadly made these epidemics seem small in comparison.
An outstanding question is: what fraction of Italians were infected with the virus?
Some towns were known to be hit hard & the Lombardy region was thought to be heavily affected.
This new study suggests quite low: 2.5% for Italy & 7.5% of Lombardy had antibodies.
Some towns were known to be hit hard & the Lombardy region was thought to be heavily affected.
This new study suggests quite low: 2.5% for Italy & 7.5% of Lombardy had antibodies.
Some very smart people think these estimates are surprisingly low ( @nataliexdean @AdamJKucharski @isabelrodbar)
https://twitter.com/isabelrodbar/status/1290685761482711040
but I think this is just b/c we didn't have proper scale for epidemics when they occurred in Italy.
https://twitter.com/isabelrodbar/status/1290685761482711040
but I think this is just b/c we didn't have proper scale for epidemics when they occurred in Italy.
One way to assess whether seroprev is unexpectedly low is by comparing estimated fatality of COVID19 (infection fatality ratio: https://twitter.com/DiseaseEcology/status/1252844190070829056) by age in Italy to other estimates.
Here's all rigorous age-specific IFRs I know of + crude Italy values by @stats_q. Italy IFR only high for 1 age group (70+) & could be b/c many 80+ in group (other studies split 70-9, 80-9).
(fig needs CIs)
https://doi.org/10.1016/S1473-3099(20)30243-7
https://science.sciencemag.org/content/369/6500/208
https://thelancet.com/pdfs/journals/laninf/PIIS1473-3099(20)30584-3.pdf
(fig needs CIs)
https://doi.org/10.1016/S1473-3099(20)30243-7
https://science.sciencemag.org/content/369/6500/208
https://thelancet.com/pdfs/journals/laninf/PIIS1473-3099(20)30584-3.pdf
Thus, the low seroprevalence in this study is actually not greatly at odds with what we know from other studies based on recorded deaths in Italy, & other estimates of age-specific IFRs.
One other point of reference is Spain which did fantastic large national survey ( https://twitter.com/DiseaseEcology/status/1280303269269532672). There, prev was higher: 5% overall; 14.4% in some regions; 11% in Madrid. But Spain pop is younger that Italy (19.4% vs. 23% >65) & smaller (67M vs 47M), ...
so w/ slightly higher total deaths in Spain despite smaller & younger pop the serop data from Italy do not appear to be wildly low compared to Spain or IFRs.
There *may* be issues w/ Abbott serology assay missing some mild cases that @isabelrodbar & others mention in thread above but I don't think these Italy seroprev data imply that.
In conclusion, it'd be best (as always) to get to read study in detail before drawing strong conclusions. Hope they make full detailed report available very soon.
Specifically, detailed data on variation in space, b/w ages, etc. would be very valuable.
Specifically, detailed data on variation in space, b/w ages, etc. would be very valuable.
Small correction: There apparently was a correction to Salje posted which decreased their IFRs by a bit (pop IFR from 0.6 to 0.5%). Here's the figure re-done w/ new corrected values.
H/T @brian_gadd
H/T @brian_gadd
Here's fig w/ error bars (except crude Italy estimates) +additional estimate of older IFR from Russell ( https://www.eurosurveillance.org/content/10.2807/1560-7917.ES.2020.25.12.2000256)
