1/ Italy, South Korea and China all had different testing #COVID19 strategies. We can see this in how the breakdown of their confirmed pools by age age cohort differs from their respective general population pyramids.
2/ China's #COVID19 confirmed population (as of Feb. 11th) skewed towards an older population. This reflected the emphasis early on focusing on the most symptomatic patients, which tended to be the elder. Median age in the early 50s. 
3/ South Korea's confirmed case (as of Mar. 3rd) pool skewed younger than China's, because they focused their early efforts on the Shincheonji mega-cluster, which consisted disproportionately of younger women in their 20s. Median age of overall pool in the early 40s.
4/ Italy's confirmed case pool (as of Mar. 8th) was the oldest. Median age: 65.
5/ In terms of stage of crisis, these three are pretty close, with all three at or near peak intensity. But while China/Korea looked like they had peaked at the time of these numbers were recorded, Italy looks like it still has not bottomed out yet. Its numbers will get worse.
6/ Here are the confirmed pools side-by-side.
7/ In terms of case fatality rates, Italy is by far the highest at 6.6% (as of 3/11). This compares with China (2.3% as of 2/11) and South Korea (0.8% as of 3/11). This sounds like a huge difference, but part of this is explained by the differences in confirmed pools.
8/ If we apply by the CFRs to the general population pyramid for each of the three countries (i.e. a SUMPRODUCT of the two columns), we get to 1.4% for China, 0.6% for Korea and 2.0% for Italy.
9/ Italy is still lagging in this metric, but the age-normalized mortality rate looks a lot better. Elderly are 1 to 3 orders-of-magnitude more likely to die from #COVID19 so the CFR% skews heavily depending on the makeup of the confirmed pool.
10/ This analysis is just another reason why we really need to understand the context behind the CFR figures.
A lot of people are taking high-level estimates like "infection of up to 40-70% of the adult population" and simply multiplying it by "2-4%" crude fatality rates ...
A lot of people are taking high-level estimates like "infection of up to 40-70% of the adult population" and simply multiplying it by "2-4%" crude fatality rates ...
11/ ... besides the fact that only a fraction of the 40-70% are going to be confirmable/detectable via lab tests, you also need to break down the 40-70% into age groups and apply the age-specific CFRs ...
12/ As I have been emphasizing for much of the last 8 weeks since I started tracking #COVID19 closely, the point is that we need to be careful and understand the context behind the numbers before we try to extrapolate and draw conclusions from it.
quora.com/How-dangerous-…
quora.com/How-dangerous-…
13/ Now if we apply CFRs by age to the U.S. population, we would get 1.30%.
Note: the CFRs by age are based on an average of China, South Korea and Italy.
Note: the CFRs by age are based on an average of China, South Korea and Italy.
14/ OTOH if we look at the average NBA player, adj. CFR rate is a relatively minuscule 0.08% (<seasonal flu). It would be even lower if not for the ageless @mrvincecarter15
16/ The average @realDonaldTrump voter: adj. CFR of 1.47%
Note to Trump voters: you should care a lot more about #COVID19 than Hillary voters.
Note to Trump voters: you should care a lot more about #COVID19 than Hillary voters.
17/ I can go on ... or you can plug in your own numbers to see how an age-normalized CFR would change here: docs.google.com/spreadsheets/d…
18/ The next level of analysis should be to incorporate the concept of how CFRs differ based on the binary "Wuhan" (chaotic/stressed medical system) vs. "Non-Wuhan" (normal) that I started exploring a couple weeks ago ...
19/ And then layering on other specific country/regional factors (besides the population pyramid) like population density, quality of healthcare system, efficiency of response, economic indicators etc.
