2/n If you look at the cases/deaths data from the US, it certainly seems true. Far MORE confirmed cases of COVID-19 in the second wave, but far FEWER deaths At first glance, looks like COVID-19 is getting less lethal!pic.twitter.com/M9IwnCFids
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
2/n If you look at the cases/deaths data from the US, it certainly seems true. Far MORE confirmed cases of COVID-19 in the second wave, but far FEWER deaths At first glance, looks like COVID-19 is getting less lethal!pic.twitter.com/M9IwnCFids
3/n Problem is, we know that testing has changed enormously over the year, especially in the US WAY more people are being tested now than in the earlier months of the year, which makes the cases/deaths equation a bit uselesspic.twitter.com/A2kzf7bEeN
4/n Think about it this way - even if there had been 3 million cases/week in the US in the first wave, there's no way we'd know because they weren't doing that many tests back in March/April!
5/n But something that is quite well reported and recorded at all times is deaths Here's the breakdown of deaths by age until mid-September in the US Immediately, something springs outpic.twitter.com/5BIzPMu3p8
6/n Yes, the waves are numerically different, but the PROPORTION of deaths in each age band is...remarkably stable. Lots of deaths in older people, far fewer in the younger groups
7/n But what does that say about cases? Well, we can actually use these death figures to (VERY crudely) back-calculate an estimate of the number of cases each week in the US by age
8/n Using our age-stratified IFR paper, we can (again VERY CRUDELY) estimate cases by saying: IFR = deaths/true cases => true cases = deaths/IFR https://www.medrxiv.org/content/10.1101/2020.07.23.20160895v6 …pic.twitter.com/Wxpf2RPHhS
9/n If we plug these deaths and age-stratified IFRs into our equation, and assume that there is a median ~30 day lag between infection and reported death, we end up with a graph that looks like this for cases Not as different as before!pic.twitter.com/u2GPqrbnT6
10/n Now, as I've repeatedly said, this is VERY CRUDE For example, deaths are so rare in the under-5s that it's hard to estimate any realistic number of cases in this age group. It's likely that they are underrepresented to an extent
11/n But we can go further even than this. What if we graphed the PROPORTION of infections in each age band, inferred from deaths? It looks something like this Now the waves barely look different at allpic.twitter.com/IozFDdIxpu
12/n In particular, if I put the two graphs side-by-side, you can see how a TINY change in the proportion in older people being infected can lead to a HUGE numerical increase in deathspic.twitter.com/QUyfQQh9z7
13/n But broadly speaking, using this (again, VERY CRUDE) method, there appears to be little difference between waves 1 & 2 in the US It's simply that our testing changed, not that the disease itself was different
14/n Indeed, as you can see, the FIRST wave in the US PROBABLY HAD MORE CASES THAN THE SECOND despite having fewer confirmed casespic.twitter.com/20lUU4KuOR
15/n Interestingly, this also gives us a very crude number of total cases in the US roughly in line with @youyanggu's modelled estimates, with about 12-15% of the country infected by late September
16/n Another point - people have said that this is flawed because the IFR in the second wave is less than that of the first If you reduce the IFRs in the second wave by 35%, this is what the graph looks like. Still not very different!pic.twitter.com/cuE89EVxps
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.