3 ms·

Why is "suggesting that nearly 2% of the population has been infected “under the radar” even remotely shocking ?

vkou 7y ago
Because when you claim that 2-3% of the population has been infected in an area where 1/30000 of the population died from the virus, it implies that the virus has a 1/600 infection/death rate. Since ~1/500 New Yorkers have died from this virus, that would imply that 120% of the population of NYC have been infected, and that new NYC cases will drop to zero in two weeks. Which is, obviously, nonsense. One of the numbers here doesn't fit the facts, and it's probably the one that claims that 2-3% of the population of Santa Clara was infected, with the overwhelming majority not having any symptoms. Data from the Diamond Princess points to something between 20% and 50% of infected people being asymptomatic - not 95%.
- irq11 7y ago
  Try 1 in 600. There are also huge error bars on that ratio, because the denominator (population size of New York), is a rough estimate, at best. In any case, this is nit-picking: the obvious source of error here is trying to cross-apply the IFR from a limited sample in one city to another, completely different city. If you’re only off by a factor of 20%, that’s a pretty strong indication that you’re on to something real. New York is testing a lot more than California, so I expect their ratio of undiscovered cases to be lower. But it’s becoming clearer and clearer that the true IFR for this virus is substantially lower than previously estimated.
  - mcbits 7y ago
    Nobody is estimating that 100% of New York has already been (or even will be) infected, so something is off by much more than 20%. With an optimistic, but at least plausible, estimate that 40% of NYC have been infected, then the IFR is around 0.4% which actually is within 20% of the 0.5% conservative guess that people have been making.
    - irq11 7y ago
      Who are “people”? Who is defining “optimistic”, and “plausible”? Where are you getting these numbers? The LA study implies a ratio of 30-50x the number of confirmed cases. Even at the high end of that range, given the current confirmed infection count in nyc (141.2k), then about 7M people would have been infected. That’s not 100% of the population, and it’s entirely plausible.
      - mcbits 7y ago
        e.g. > Comparing deaths onboard with expected deaths based on naive CFR estimates using China data, we estimate IFR and CFR in China to be 0.5% (95% CI: 0.2-1.2%) and 1.1% (95% CI: 0.3-2.4%) respectively. https://www.medrxiv.org/content/10.1101/2020.03.05.20031773v2 https://www.medrxiv.org/content/10.1101/2020.03.05.20031773v... 7 million is 80% of the population, which is at the high end of high estimates for the total portion of the population expected to be infected in the end. That would be more plausible if daily deaths were well into the long tail, but NYC appears to be just past the hump and hundreds/day are still dying. I would say 30x undercounting (48% infected) is highly optimistic but still plausible if you want to embrace that.
        irq11 7y ago
        Ok, so you’re taking one paper and assuming it is correct, and using that to dismiss more recent data. Keep in mind that not even two weeks ago, the best estimate of IFR in China was 0.66%. It keeps dropping. Also, the confidence interval on the paper you’re citing extends to 0.2%. I’m not saying that the factor is 50x, just that it’s plausible.
        mcbits 7y ago
        I don't assume that paper is correct. It's the first example I found of people estimating 0.5% since you apparently hadn't noticed that number being bandied around for the last month. It's a number I've been seeing since early March, along with 40-70% of the population ultimately being infected. I saw one guy suggesting up to 80%. If NYC is already at 80% and rising, that's pretty surprising. I would like to get an idea of what the IFR really is, not which extreme end of the range of uncertainty looks the best or worst. If it's a little less than 0.5%, great. But when some paper comes out of left field and suggests it's closer to 0.05%, sure, I'm skeptical, especially when mortality is already above that fraction of the entire population in several places.

vkou 7y ago

Because when you claim that 2-3% of the population has been infected in an area where 1/30000 of the population died from the virus, it implies that the virus has a 1/600 infection/death rate. Since ~1/500 New Yorkers have died from this virus, that would imply that 120% of the population of NYC have been infected, and that new NYC cases will drop to zero in two weeks. Which is, obviously, nonsense. One of the numbers here doesn't fit the facts, and it's probably the one that claims that 2-3% of the population of Santa Clara was infected, with the overwhelming majority not having any symptoms. Data from the Diamond Princess points to something between 20% and 50% of infected people being asymptomatic - not 95%.

irq11 7y ago
Try 1 in 600. There are also huge error bars on that ratio, because the denominator (population size of New York), is a rough estimate, at best. In any case, this is nit-picking: the obvious source of error here is trying to cross-apply the IFR from a limited sample in one city to another, completely different city. If you’re only off by a factor of 20%, that’s a pretty strong indication that you’re on to something real. New York is testing a lot more than California, so I expect their ratio of undiscovered cases to be lower. But it’s becoming clearer and clearer that the true IFR for this virus is substantially lower than previously estimated.
- mcbits 7y ago
  Nobody is estimating that 100% of New York has already been (or even will be) infected, so something is off by much more than 20%. With an optimistic, but at least plausible, estimate that 40% of NYC have been infected, then the IFR is around 0.4% which actually is within 20% of the 0.5% conservative guess that people have been making.
  - irq11 7y ago
    Who are “people”? Who is defining “optimistic”, and “plausible”? Where are you getting these numbers? The LA study implies a ratio of 30-50x the number of confirmed cases. Even at the high end of that range, given the current confirmed infection count in nyc (141.2k), then about 7M people would have been infected. That’s not 100% of the population, and it’s entirely plausible.
    - mcbits 7y ago
      e.g. > Comparing deaths onboard with expected deaths based on naive CFR estimates using China data, we estimate IFR and CFR in China to be 0.5% (95% CI: 0.2-1.2%) and 1.1% (95% CI: 0.3-2.4%) respectively. https://www.medrxiv.org/content/10.1101/2020.03.05.20031773v2 https://www.medrxiv.org/content/10.1101/2020.03.05.20031773v... 7 million is 80% of the population, which is at the high end of high estimates for the total portion of the population expected to be infected in the end. That would be more plausible if daily deaths were well into the long tail, but NYC appears to be just past the hump and hundreds/day are still dying. I would say 30x undercounting (48% infected) is highly optimistic but still plausible if you want to embrace that.
      - irq11 7y ago
        Ok, so you’re taking one paper and assuming it is correct, and using that to dismiss more recent data. Keep in mind that not even two weeks ago, the best estimate of IFR in China was 0.66%. It keeps dropping. Also, the confidence interval on the paper you’re citing extends to 0.2%. I’m not saying that the factor is 50x, just that it’s plausible.
        mcbits 7y ago
        I don't assume that paper is correct. It's the first example I found of people estimating 0.5% since you apparently hadn't noticed that number being bandied around for the last month. It's a number I've been seeing since early March, along with 40-70% of the population ultimately being infected. I saw one guy suggesting up to 80%. If NYC is already at 80% and rising, that's pretty surprising. I would like to get an idea of what the IFR really is, not which extreme end of the range of uncertainty looks the best or worst. If it's a little less than 0.5%, great. But when some paper comes out of left field and suggests it's closer to 0.05%, sure, I'm skeptical, especially when mortality is already above that fraction of the entire population in several places.