I've written several twitter threads on Prof Ioannidis' papers on IFR, recently herehttps://twitter.com/GidMK/status/1316511734115385344?s=20 …
You can add location information to your Tweets, such as your city or precise location, from the web and via third-party applications. You always have the option to delete your Tweet location history. Learn more
I've written several twitter threads on Prof Ioannidis' papers on IFR, recently herehttps://twitter.com/GidMK/status/1316511734115385344?s=20 …
While there are many areas in which we disagree, I think the biggest error remains including studies that are clearly inappropriate to determine population estimates of infection rates
One example of this is the IFR for Wuhan. There are 4 estimates of seroprevalence that you could use to calculated an IFR for Wuhan in early 2020 In the paper, 3 of them are usedpic.twitter.com/02a4XQVtTA
The samples are: 1. Blood donor study 2. Single hospital study 3. Single hospital study 4. Massive, random, province/city-wide studypic.twitter.com/Y11cAz0cJV
Now, leaving aside some of the other issues with using these studies, let's look at the simple fact that only one was even remotely representative of the population of Wuhan
The blood donor study is enormously unrepresentative. The population is (as with most blood donors) MUCH younger than the general poppic.twitter.com/Kk4E5y2JCM
The larger single-site study literally says at the start that a major limitation is the biased samplepic.twitter.com/d834C8m5gT
The smaller single-site study includes hospitalized patients as well as people returning to work at a single location in a city of 11 million. Not even remotely representative of the populationpic.twitter.com/spWByih9lW
The final study was a large, carefully designed serosurvey that was representative not just of the city of Wuhan but also surrounding regionspic.twitter.com/g9bsNGFGWH
For the four serosurveys, using Professor Ioannidis' methodology, the inferred IFRs are: 1. 0.45% 2. 0.35% 3. 0.42% 4. 0.82% So the representative sample implies an IFR double that of the biased samples
This actually accords with published data. There is some fairly strong evidence that selection bias can double your estimate of seroprevalence (thus halving the estimate of IFR)https://twitter.com/GidMK/status/1381391856773197824?s=20 …
Now, the fourth seroprevalence sample was not published until after Prof Ioannidis' paper came out, but the point here is that the three samples included in the paper are not sufficient to infer infections in the population
All 3 estimates of the IFR that use biased sampling and survey methodology are half the more rigorous data. This inclusion of inappropriate estimates is repeated numerous times in the IFR review @AtomsksSanakan covered this in detailhttps://twitter.com/AtomsksSanakan/status/1341183815176364038?s=20 …
Anyway, I always think it's quite telling when people choose to attack the qualifications of their critics rather than discussing the critique itself
Apologies! Another sample has been recently published that I was not aware of. This is also a random citywide estimate that implies an IFR of 0.5% So a reasonable range might be 0.5-0.8% for the IFR of Wuhanhttps://www.thelancet.com/journals/lancet/article/PIIS0140-6736(21)00238-5/fulltext …
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.