Read on Twitter
Results of massive (N=64000) nation-wide serosurvey in Italy are out. Available in Italian only, for the moment, so I& #39;ll translate a few highlights if you are interested.
https://www.istat.it/it/archivio/246156">https://www.istat.it/it/archiv...
https://www.istat.it/it/archivio/246156">https://www.istat.it/it/archiv...
As little as 2.5% [2.2%,2.7%] of people surveyed had IGg antibodies.
This does not change by sex, and very little by age (with the 50-59 category the most hit).
![As little as 2.5% [2.2%,2.7%] of people surveyed had IGg antibodies. This does not change by sex, and very little by age (with the 50-59 category the most hit).](https://pbs.twimg.com/media/EegggiRXkAAEHMP.png "As little as 2.5% [2.2%,2.7%] of people surveyed had IGg antibodies. This does not change by sex, and very little by age (with the 50-59 category the most hit).")
This does not change by sex, and very little by age (with the 50-59 category the most hit).
Implied age-specific IFR:
0-17: 0.002%
18-34: 0.02%
35-49: 0.11%
50-59: 0.40%
60-69: 1.74%
>70: 8.30%
0-17: 0.002%
18-34: 0.02%
35-49: 0.11%
50-59: 0.40%
60-69: 1.74%
>70: 8.30%
Considering the overall age structure of the country, the implied overall IFR is:
0.002*0.17 (17% of Italians are 0-17) +
0.02*0.17 +
0.11*0.21 +
0.4*0.16+
1.7*0.12+
8.3*0.17 =
1.7% (!!)
0.002*0.17 (17% of Italians are 0-17) +
0.02*0.17 +
0.11*0.21 +
0.4*0.16+
1.7*0.12+
8.3*0.17 =
1.7% (!!)
Huge regional variation, in line with expectation given most cases/deaths were from the northern regions.
Not much variation in terms of occupation, apart from the obvious exception of health workers. Interestingly, no difference at all between workers affected/not affected by lockdown.
As little as 40% of people with someone positive living with them developed antibodies. 16% of people with a known positive contact outside the family, 12% for health workers with a positive patient.
Proportion of asymptomatics was 27%.
About 50% of positives developed fever. 37% lost sense of taste/smell. (But only ~25% of people losing sense of smell/taste were positive!)
Caveats:
1) the original sample size was far larger (150 000) but lots of people did not respond. There is a suspicion most of the people who refused did so because of fear of being quarantined. This would lower the estimated proportion of infected people.
1) the original sample size was far larger (150 000) but lots of people did not respond. There is a suspicion most of the people who refused did so because of fear of being quarantined. This would lower the estimated proportion of infected people.
At the same time, I guess people who had symptoms might be more likely to accept if interested in knowing whether they had it, so the two effects *might* even cancel out.
2) People living in care homes were excluded from the survey, while they made up probably a good proportion of the overall deaths. I could not find any data on deaths by settings.
3) IFR estimates I provided above are based on reported deaths, excess deaths are ~40% higher.
3) IFR estimates I provided above are based on reported deaths, excess deaths are ~40% higher.
4) I did not estimate variation around IFR,which might be high.
5) It took quite some time to take the samples, almost two months. But prevalence was supposed to be low at that time (end of May to mid July).
5) It took quite some time to take the samples, almost two months. But prevalence was supposed to be low at that time (end of May to mid July).
Methodological aspects:
Two-stage stratified sampling. Startification by region, sex, age, profession.
Target N=150k
Achieved N=64k
IPW used to address missing data (I think, not overly clear to me)
Two-stage stratified sampling. Startification by region, sex, age, profession.
Target N=150k
Achieved N=64k
IPW used to address missing data (I think, not overly clear to me)
Oh, and of course, congrats to people at @istat_en and @crocerossa for completing such an important study in such difficult circumstances.
A further clarification given that this has attracted far more attention than I thought: the report does not mention IFR, these were just my back-of-the-envelope calculations. Plenty of possible reasons why my calculations might make little sense (antibodies waning, etc etc).
Oh, and an obvious thing I missed and everybody is rightly asking for: https://twitter.com/FedeViganego/status/1290643609965015043">https://twitter.com/FedeVigan...
