Before I delve into a little more number-crunching, a question (actually two) for those of you with a background in statistics or epidemiology. As has been repeatedly pointed out, one problem with the case fatality rate measure is the changing denominator as more people get tested. So, first question, given that we are unlikely to test 100% of the population, at what point do you consider the amount of testing to be statistically significant and thus close to giving us reliable numbers? And second but related to the first, given that the testing currently being done is generally not random, is this useless for drawing reliable conclusions?
For reference, as of yesterday, California with a population of 39.5 million, had conducted 4.8 million tests, so about 12% of the population (give or take, since some people have likely been tested more than once, and some were from out of state). A random sample of tests as a rough percentage of population for other states include Connecticut - 14.6%, Florida - 10.4%, Georgia - 10.6%, Illinois - 14.0%, and Virginia - 9.2%. I would guess that most other states are also around 10% or so.
And now, back to the number-crunching. As a few articles or internet posts have noted, for all the panic and locking down of the entire economy and society over the past few months, Covid-19 primarily only seriously affects the elderly, especially those with co-morbidities. For most others, it may not be much worse that the seasonal flu. In general, about 80-85% of deaths have been of people over 70 and about 90-95% have been of those over 60.
Here are a few numbers to give this context. Note that different states record data differently.
California:
Age 0-17: no deaths; CFR = 0
Age 18-49: 384 deaths in 178767 cases; CFR = 0.24%
Age 50-64: 1044 deaths in 54836 cases; CFR = 1.90%
Age 65 + : 4872 deaths in 35441 cases; CFR = 13.75%
Connecticut:
Age 0-19: 2 deaths in 1068 cases; CFR= 0.19%
Age 20-39: 25 deaths in 12396 cases; CFR= 0.20%
Age 40-59: 232 deaths in 14535 cases; CFR= 1.60%
Age 60-69: 533 deaths in 6551 cases; CFR= 8.14%
Age 70-79: 945 deaths in 4416 cases; CFR= 21.40%
Age 80 +: 2601 deaths in 6952 cases; CFR= 34.71%
Recall that Connecticut's overall fatality rate is worst among the 50 states.
Florida:
Age 0-14: 1 death in 9847 cases; CFR = 0.01%
Age 15-34: 28 deaths in 75301 cases; CFR = 0.03%
Age 35-54: 218 deaths in 63889 cases; CFR = 0.34%
Age 55-64: 342 deaths in 24774 cases; CFR = 1.38%
Age 65-74: 784 deaths in 14852 cases; CFR = 5.28%
Age 75-84: 1056 deaths in 8829 cases; CFR = 11.96%
Age 85 + : 1349 deaths in 5561 cases; CFR = 24.26%
Illinois:
Age 0-19: 4 deaths in 11519 cases; CFR= 0.03%
Age 20-39: 135 deaths in 49388 cases; CFR= 0.27%
Age 40-59: 829 deaths in 50321 cases; CFR= 1.65%
Age 60-69: 1253 deaths in 17151 cases; CFR= 7.31%
Age 70-79: 1678 deaths in 9534 cases; CFR= 17.60%
Age 80 +: 3127 deaths in 9895 cases; CFR= 31.60%
More children have been killed by gunfire in the last 2 weeks in Chicago alone than have died of Covid-19 in the entire state this year.
Virginia:
Age 0-19: 0 deaths in 7008 cases; CFR= 0%
Age 20-39: 17 deaths in 23510 cases; CFR= 0.07%
Age 40-59: 158 deaths in 21610 cases; CFR= 0.73%
Age 60-69: 267 deaths in 6486 cases; CFR= 4.12%
Age 70-79: 461 deaths in 3378 cases; CFR= 13.65%
Age 80 +: 948 deaths in 3473 cases; CFR= 27.30%
Federal Republic of Germany:
Age 0-19: 3 deaths in 14560 cases; CFR= 0.02%
Age 20-49: 107 deaths in 86421 cases; CFR= 0.12%
Age 50-69: 1181 deaths in 59519 cases; CFR= 1.98%
Age 70-89: 6029 deaths in 30279 cases; CFR= 19.91%
Age 90 +: 1691 deaths in 5359 cases; CFR= 31.55%
In Germany, people over 70 account for 86% of all Covid deaths. People over 50 account for 98.8%. That may be one of the reasons why Germany was among the first places to reopen schools, even as other large events, such as Oktoberfest, which draw an older population, have remained cancelled.
Also looking at German data, I note that I gave a CFR of 4.6% above for the country as a whole. Four states were higher than this average: Saarland at 6.2%, Bavaria at 5.33%, Baden-Württemberg at 5.13% and Hamburg at 5.00%. Hamburg is a densely-populated city-state, so a higher rate stands to reason. The main thing the other three states seem to have in common is they border Austria, Switzerland and France, so they were likely the first places in Germany affected as the pandemic spread from Italy north. Germany on the whole has a somewhat older population than the US, but I am not sure how these states compare to other states. Despite their higher population densities, Berlin and North Rhine-Westphalia were below the German average.
I can't get a good number for LTC facility deaths in Germany, because German law categorizes LTCs along with a number of other high-occupancy facilities, so the German number includes not just nursing homes, but homeless shelters, refugee/asylum-seeker facilities and prison.
Regarding France, there does seem to be some defensiveness from French sources about their high death rate, as well as a bit of playing the blame game, especially with regard to countries like Germany allegedly hoarding PPE. I also looked at some data from the
Institut national de la statistique et des études économiques. As high as they is, the overall numbers for France are relatively unchanged over recent weeks as the crisis there is essentially over, not that politicians and bureaucrats seem willing to admit it and let a crisis go to waste. Since the beginning of May, the daily mortality rate for all deaths in France has returned to the average of previous years (chart
here). In fact, deaths are in many cases lower than in previous years, suggesting either that the lockdowns reduced mortality in other areas such as traffic accidents, or that Covid-19 killed many people in March and April who would have died in May, June or later anyway. There is a breakdown by age, but it only covers those who died in hospitals.