11 ms·

When that happens, it's too late. Between the incubation period and the time to death it's about three weeks. At a 20% increase per day that means that even if

When that happens, it's too late. Between the incubation period and the time to death it's about three weeks. At a 20% increase per day that means that even if you stop going out at that moment deaths will be multiplied by 46. If the rate is 30%, the multiplier is 247.

btilly 7y ago
Not quite that bad. Current measured growth rates are closer to 15% per day (5 day doubling time). And people's behavior changes not when people die, but when people get sick. Which happens (depending on your sensitivity) only 1-2 weeks after exposure. But the basic principle still applies. We are trying to buy as much time as we can to ramp up emergency provisions. And are trying to avoid overwhelming emergency rooms. By the time people's behavior modifies through direct experience, it is too late. Making this concrete I live in Orange County, CA. Population, 3.3 million. Hospital beds, 6600 (source https://www.hasc.org/orange-county https://www.hasc.org/orange-county). I don't know how many ICU beds, but https://www.accjournal.org/journal/view.php?number=630 https://www.accjournal.org/journal/view.php?number=630 suggests it likely is about 6% of the total. So about 400ish. Now walk that backwards. 400 people in ICU beds probably means 4000 people with symptoms means less than 0.1% of the population with COVID. Except that it is worse than that. We have those beds for normal stuff that goes wrong. They aren't actually empty now. Long story short, we start piling up excess dead bodies due to lack of capacity long before most of us personally know anyone who actually shows symptoms.
- acqq 7y ago
  > Current measured growth rates are closer to 15% per day (5 day doubling time) Citation needed. In the US, via Wikipedia: https://en.wikipedia.org/wiki/2020_coronavirus_pandemic_in_the_United_States https://en.wikipedia.org/wiki/2020_coronavirus_pandemic_in_t... 2020-03-17 5656 infected 96 dead 2020-03-19 11980 infected 172 dead That's doubling time of 2 days, not 5. When was it 5 days the last time? If that 2 days continues, in two weeks: 1.5 million infected, 22000 dead. Also, the data across the world show only: the typical doubling time once it really starts going is 3 days, never 5, at least without the beneficial contribution of lockdowns: Italy dead: 2020-03-16 1809 2020-03-19 3405 It's 3 days doubling time there, and they started to quarantine municipalities since February 22, and locked down the whole country on March 10. https://en.wikipedia.org/wiki/2020_coronavirus_pandemic_in_Italy https://en.wikipedia.org/wiki/2020_coronavirus_pandemic_in_I... Netherlands dead: 2020-03-16 24 2020-03-19 76 Doubling time faster than 3 days there. Not much of measures. https://en.wikipedia.org/wiki/2020_coronavirus_pandemic_in_the_Netherlands https://en.wikipedia.org/wiki/2020_coronavirus_pandemic_in_t... So "the risky hack" this article is about buys one not more than 3 days under the circumstances. Speed of deaths is a good indicator because the number of those who die and those who need an intensive care unit are very close. A lot of countries aren't able to test as much as it would be needed, so the number of "verified cases" which have less serious symptoms tends to be lower than the actual numbers. But those who die and could have been exposed to the virus and have matching symptoms are more consistently tested. Main point: the US data can't be considered in vacuum, we know much more than US was able to test due to their botched procedures. But across the world, it's nowhere 5 days, as soon as the number of dead is more than a single digit.
  - rincebrain 7y ago
    At least some of that growth in the US is probably due to increased testing availability, not infection rate, so I'd suspect "actual" numbers of infected in the US on 3/17 were higher than the data showed. But that's dealing with lots of unknowns.
  - btilly 7y ago
    Epidemiological estimates and projections seem to be based on the 5 day estimate. See https://www.hpnonline.com/infection-prevention/screening-surveillance/article/21130206/covid19-predicted-to-infect-81-of-us-population-cause-22-million-deaths-in-us https://www.hpnonline.com/infection-prevention/screening-sur... for an example. I am honestly not sure of all the lines of evidence leading to that estimate, but I've seen it used a number of times. Measuring what is going on by confirmed cases measures ramping out of testing more than the actual growth of the virus. Death figures are likely to be more accurate. But note that we expect it to take 3 weeks before quarantines show up in fatality figures. (A week for exposed to get sick, 2 more weeks before they die.) Therefore Italy shouldn't yet see the benefits of quarantines.
    - acqq 7y ago
      What you cite is from: https://www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships/Imperial-College-COVID19-NPI-modelling-16-03-2020.pdf https://www.imperial.ac.uk/media/imperial-college/medicine/s... topic of which isn't the evaluation of the recorded growth across the world, but to show that lax measures are problematic even using the slower growth rates. They specifically write just: "Infection was assumed to be seeded in each country at an exponentially growing rate (with a doubling time of 5 days) from early January 2020, with the rate of seeding being calibrated to give local epidemics which reproduced the observed cumulative number of deaths in GB or the US seen by 14th March 2020." That's what they use in their model to demonstrate the point. They never claim it's accurate or quote any sources for that number, because they just use that very "optimistic" value for the demonstration purposes: to show that even with that number, it's not reasonable not to implement serious measures. That's how mathematical proofs are constructed: you construct the lowest bound which doesn't have to be accurate, just to be obviously lower than the actual numbers, and show that even then some assumption doesn't hold. I'm claiming here that that is not a "current measured rate." > But note that we expect it to take 3 weeks before quarantines show up in fatality figures. That's why I've stated about Italy: "they started to quarantine municipalities since February 22." It's not that they let it waiting for "herd immunity" like UK or NL. It was less than 4 weeks ago. 2020-02-22 2 dead 2020-03-19 3405 dead with 3 days doubling time they would have today 2048 dead. It doesn't appear to be possible to be slower than 3 days. With 5 days doubling time they would have now just 128 dead, with 4 days doubling time 256 dead. It's that easy to calculate. That's the Italian "current measured rate" for a month now.
- sp332 7y ago
  The growth rate is different in different countries. The USA is currently on a 35% slope. http://nrg.cs.ucl.ac.uk/mjh/covid19/ http://nrg.cs.ucl.ac.uk/mjh/covid19/
  - btilly 7y ago
    That is growth rate as measured by confirmed cases. However given various restrictions on availability of testing, confirmed cases and real cases are likely to tell very different stories. Given that most countries refuse to test, refuse to test, hit an "oh shit moment" and start ramping up, the confirmed rate that your link uses measures testing more than actual infection rates. References such as https://www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships/Imperial-College-COVID19-NPI-modelling-16-03-2020.pdf https://www.imperial.ac.uk/media/imperial-college/medicine/s... have used a doubling time of about 5 days. As https://www.statnews.com/2020/03/10/simple-math-alarming-answers-covid-19/ https://www.statnews.com/2020/03/10/simple-math-alarming-ans... points out, studies of what happened in Wuhan suggest a real doubling time of about 6 days. There are admittedly a lot of question marks about everyone's numbers.
  - nerdponx 7y ago
    New York's growth rate is going to look terrifyingly high as testing capability ramps up significantly.
    - ummonk 7y ago
      It is ramping up significantly, with a corresponding growth rate of confirmed cases. 4 days ago it was approaching 1000. It will probably be 10,000 by tomorrow.
    - gwright 7y ago
      There has been a steady drumbeat of warnings to expect that spike, but unfortunately the statistical nuances of epidemiology will be lost on many.

btilly 7y ago

Not quite that bad. Current measured growth rates are closer to 15% per day (5 day doubling time). And people's behavior changes not when people die, but when people get sick. Which happens (depending on your sensitivity) only 1-2 weeks after exposure. But the basic principle still applies. We are trying to buy as much time as we can to ramp up emergency provisions. And are trying to avoid overwhelming emergency rooms. By the time people's behavior modifies through direct experience, it is too late. Making this concrete I live in Orange County, CA. Population, 3.3 million. Hospital beds, 6600 (source https://www.hasc.org/orange-county https://www.hasc.org/orange-county). I don't know how many ICU beds, but https://www.accjournal.org/journal/view.php?number=630 https://www.accjournal.org/journal/view.php?number=630 suggests it likely is about 6% of the total. So about 400ish. Now walk that backwards. 400 people in ICU beds probably means 4000 people with symptoms means less than 0.1% of the population with COVID. Except that it is worse than that. We have those beds for normal stuff that goes wrong. They aren't actually empty now. Long story short, we start piling up excess dead bodies due to lack of capacity long before most of us personally know anyone who actually shows symptoms.

acqq 7y ago
> Current measured growth rates are closer to 15% per day (5 day doubling time) Citation needed. In the US, via Wikipedia: https://en.wikipedia.org/wiki/2020_coronavirus_pandemic_in_the_United_States https://en.wikipedia.org/wiki/2020_coronavirus_pandemic_in_t... 2020-03-17 5656 infected 96 dead 2020-03-19 11980 infected 172 dead That's doubling time of 2 days, not 5. When was it 5 days the last time? If that 2 days continues, in two weeks: 1.5 million infected, 22000 dead. Also, the data across the world show only: the typical doubling time once it really starts going is 3 days, never 5, at least without the beneficial contribution of lockdowns: Italy dead: 2020-03-16 1809 2020-03-19 3405 It's 3 days doubling time there, and they started to quarantine municipalities since February 22, and locked down the whole country on March 10. https://en.wikipedia.org/wiki/2020_coronavirus_pandemic_in_Italy https://en.wikipedia.org/wiki/2020_coronavirus_pandemic_in_I... Netherlands dead: 2020-03-16 24 2020-03-19 76 Doubling time faster than 3 days there. Not much of measures. https://en.wikipedia.org/wiki/2020_coronavirus_pandemic_in_the_Netherlands https://en.wikipedia.org/wiki/2020_coronavirus_pandemic_in_t... So "the risky hack" this article is about buys one not more than 3 days under the circumstances. Speed of deaths is a good indicator because the number of those who die and those who need an intensive care unit are very close. A lot of countries aren't able to test as much as it would be needed, so the number of "verified cases" which have less serious symptoms tends to be lower than the actual numbers. But those who die and could have been exposed to the virus and have matching symptoms are more consistently tested. Main point: the US data can't be considered in vacuum, we know much more than US was able to test due to their botched procedures. But across the world, it's nowhere 5 days, as soon as the number of dead is more than a single digit.
- rincebrain 7y ago
  At least some of that growth in the US is probably due to increased testing availability, not infection rate, so I'd suspect "actual" numbers of infected in the US on 3/17 were higher than the data showed. But that's dealing with lots of unknowns.
- btilly 7y ago
  Epidemiological estimates and projections seem to be based on the 5 day estimate. See https://www.hpnonline.com/infection-prevention/screening-surveillance/article/21130206/covid19-predicted-to-infect-81-of-us-population-cause-22-million-deaths-in-us https://www.hpnonline.com/infection-prevention/screening-sur... for an example. I am honestly not sure of all the lines of evidence leading to that estimate, but I've seen it used a number of times. Measuring what is going on by confirmed cases measures ramping out of testing more than the actual growth of the virus. Death figures are likely to be more accurate. But note that we expect it to take 3 weeks before quarantines show up in fatality figures. (A week for exposed to get sick, 2 more weeks before they die.) Therefore Italy shouldn't yet see the benefits of quarantines.
  - acqq 7y ago
    What you cite is from: https://www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships/Imperial-College-COVID19-NPI-modelling-16-03-2020.pdf https://www.imperial.ac.uk/media/imperial-college/medicine/s... topic of which isn't the evaluation of the recorded growth across the world, but to show that lax measures are problematic even using the slower growth rates. They specifically write just: "Infection was assumed to be seeded in each country at an exponentially growing rate (with a doubling time of 5 days) from early January 2020, with the rate of seeding being calibrated to give local epidemics which reproduced the observed cumulative number of deaths in GB or the US seen by 14th March 2020." That's what they use in their model to demonstrate the point. They never claim it's accurate or quote any sources for that number, because they just use that very "optimistic" value for the demonstration purposes: to show that even with that number, it's not reasonable not to implement serious measures. That's how mathematical proofs are constructed: you construct the lowest bound which doesn't have to be accurate, just to be obviously lower than the actual numbers, and show that even then some assumption doesn't hold. I'm claiming here that that is not a "current measured rate." > But note that we expect it to take 3 weeks before quarantines show up in fatality figures. That's why I've stated about Italy: "they started to quarantine municipalities since February 22." It's not that they let it waiting for "herd immunity" like UK or NL. It was less than 4 weeks ago. 2020-02-22 2 dead 2020-03-19 3405 dead with 3 days doubling time they would have today 2048 dead. It doesn't appear to be possible to be slower than 3 days. With 5 days doubling time they would have now just 128 dead, with 4 days doubling time 256 dead. It's that easy to calculate. That's the Italian "current measured rate" for a month now.
sp332 7y ago
The growth rate is different in different countries. The USA is currently on a 35% slope. http://nrg.cs.ucl.ac.uk/mjh/covid19/ http://nrg.cs.ucl.ac.uk/mjh/covid19/
- btilly 7y ago
  That is growth rate as measured by confirmed cases. However given various restrictions on availability of testing, confirmed cases and real cases are likely to tell very different stories. Given that most countries refuse to test, refuse to test, hit an "oh shit moment" and start ramping up, the confirmed rate that your link uses measures testing more than actual infection rates. References such as https://www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships/Imperial-College-COVID19-NPI-modelling-16-03-2020.pdf https://www.imperial.ac.uk/media/imperial-college/medicine/s... have used a doubling time of about 5 days. As https://www.statnews.com/2020/03/10/simple-math-alarming-answers-covid-19/ https://www.statnews.com/2020/03/10/simple-math-alarming-ans... points out, studies of what happened in Wuhan suggest a real doubling time of about 6 days. There are admittedly a lot of question marks about everyone's numbers.
- nerdponx 7y ago
  New York's growth rate is going to look terrifyingly high as testing capability ramps up significantly.
  - ummonk 7y ago
    It is ramping up significantly, with a corresponding growth rate of confirmed cases. 4 days ago it was approaching 1000. It will probably be 10,000 by tomorrow.
  - gwright 7y ago
    There has been a steady drumbeat of warnings to expect that spike, but unfortunately the statistical nuances of epidemiology will be lost on many.