15/n But by far and away, the biggest error in the text is simply to do with using clearly inappropriate samples to estimate population prevalence This is a fundamental flaw in the paper, and really something of a basic epidemiological mistake
-
Show this thread
-
16/n Some of these studies are just so clearly inappropriate to infer a population estimate that it doesn't really require explaining. Samples of a single business in a city, or inpatient dialysis unitspic.twitter.com/CTWnw0Fbjf
1 reply 9 retweets 102 likesShow this thread -
18/n Then we have blood donors, who again may give an erroneous result. These are people who, DURING A PANDEMIC are happy to go out and about and give blood. It is quite possible that they are MORE likely to have been infected than the general population!pic.twitter.com/0ME5x1br1B
4 replies 5 retweets 101 likesShow this thread -
18.5/n There are also a lot of included studies from places in which there is almost certainly an enormous undercount of deaths For example, India, where the official death counts may represent a substantial underestimatehttps://www.bmj.com/content/370/bmj.m2859 …
2 replies 11 retweets 91 likesShow this thread -
19/n A very basic, reasonable thing to do would be to conduct a sensitivity analysis excluding these biased estimates, to see what happens when you only use representative population estimates Which we can do
1 reply 1 retweet 80 likesShow this thread -
20/n If we take the median of only these somewhat good-quality studies (some of them still aren't great, but at least they're not clearly inappropriate), we get a value of 0.5% Double the estimate of 0.27%pic.twitter.com/tQSICfngT9
2 replies 13 retweets 111 likesShow this thread -
Replying to @GidMK
Actually it's 0.44% You ignored 65% of the studies & 94% of those under 0.2% While ppl can disagree on blood donors being included vs. not; ignoring studies w/ lower prev also bias's IFR towards heavier hit regions. Higher prev may indicate a more susceptible community.pic.twitter.com/t4SSxvzprQ
1 reply 0 retweets 0 likes -
Replying to @sangfroyd @GidMK
Moreover the best estimate from WHO is 10% infected WW. at 1.1M deaths plus a little lag, / 770M ... = ~0.2% IFR Much closer to Ioannidis meta than yours I suspect your aversion to blood donor & low prev, biases you towards the worse performing regions that are most studied
2 replies 0 retweets 0 likes -
Replying to @sangfroyd
That is incorrect. The WHO said the UPPER ESTIMATE for those infected is 10%, a more plausible reading is less than that I have no aversion to low prevalence studies, and indeed included many of them in my own meta-analysis
2 replies 0 retweets 3 likes -
Replying to @GidMK @sangfroyd
In our AGE STRATIFIED analysis, we excluded studies in which the confidence interval included 0% for age bands, because this produces a meaningless result (essentially, you get an upper bound of 100% IFR which is problematic), but that's not the same as disliking them
1 reply 0 retweets 2 likes
As for blood donors, I have laid out in detail why they are inappropriate to use as an estimate of population seroprevalence. This is not some kind of crazy, out-there point - we would not use blood donors to estimate the population prevalence of ANY disease precisely
-
-
Replying to @GidMK @sangfroyd
Important to note - this does not make blood donor studies USELESS. They are great for sentinel surveillance to monitor and track trends. But we can actually demonstrate numerically that they are inadequate to estimate population prevalence of COVID-19
1 reply 0 retweets 1 like -
Replying to @GidMK
Didn’t say crazy. Just somewhat subjective & biases / weights you towards larger studies that tend to gravitate toward heavier hit / high IFR regions Several of your peers disagree, including Ioannidis. You may be right, you may be wrong. But its extreme to call it an error
1 reply 0 retweets 0 likes - Show replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.