News for hackers. Stuff that matters.
Re:CA governor projecting 25.5M COVID19 cases in CA in 8 wee... by grej on Thursday March 19, 2020 @11:05PM (#22634295)
There are 1030 known positive CA cases right now according to the JHU map.

The governor's letter is using a doubling time of 4 days, but there's good reason to believe this 4-day time is due to recently increased testing.

If the governor instead takes the best-guess mean doubling time from several epidemiological studies[1] of 6.4 days, then, you would have to assume there are are 38.5 actual cases for any one tested positive case right now in order to get to 25.5M.

(25.5 * (10 ^ 6)) / (1030 * (2 ^ ((7 days/wk * 8 wk) / 6.4 days_to_double))) = 38.46742327016

So to get there, the governor has to be assuming that the current measures being put in place will do absolutely nothing, and also assuming almost actual 40 cases per real case, which is quite possible.

If you assume the 4-day doubling time they quoted will continue for 8 straight weeks, and the number of current cases is accurate (again, probably not - probably an underestimate), you still just get to 16.9M cases.

Seems to me the more likely situation is that: (1) there are more cases than we think right now (Maybe not 40x, but maybe 10-20x?). (2) the doubling time will be affected quite A LOT by the measures being put in place. It might drop the R0 to close to 1.

In that case, I would expect the enacted measures to at least half the R0 starting from this week, beginning of a flattening of active case count starting at this point.

That new R0 would hugely affect doubling time. But even conservatively estimating that R0 only goes down by 25% vs 50%, in 8 weeks time the projection of cases in CA would still be reduced by almost 60x.

TLDR - Best guess 8 weeks from now, CA has between 400k-1M cases. But not anything close to 25.5M. You can downvote me in 8 weeks if I'm wrong. Hopefully, the real number is closer to mine than to Gov. Newsom's.

[1] https://www.thelancet.com/journals/lancet/article/PIIS0140-6...

Re:CA governor projecting 25.5M COVID19 cases in CA in 8 wee... by eanzenberg on Thursday March 19, 2020 @11:53PM (#22634694)
You can't have it both ways. Either

1) There are 10x (or 40x if you think) more cases out there and this is less dangerous than the flu or

2) There are 1x the cases out there and the transmission rate is what it is.

Bayesian stats is leaning towards 1) right now, and has been for a while given the evidence from SK and the cruise ship. However, when lives are at stake, it's incredibly important to err on the safe side until you're sure. In other words, be willing to lose $1 million dollars rather than lose 1 life (just throwing numbers out there).

But at some point, the greatest western economy crashing will lead to more loss of life than the flu. Let's just be sure 1st that it is just a bad flu and not potentially something worse.

Another critical reason for death rates to vary so much per country: Death rate is very dependent on quality and access to care.

Re:CA governor projecting 25.5M COVID19 cases in CA in 8 wee... by mid on Thursday March 19, 2020 @11:14PM (#22634342)
“Any statistics can be extrapolated to the point where they show disaster” — Thomas Sowell, 1996
Re:CA governor projecting 25.5M COVID19 cases in CA in 8 wee... by grej on Thursday March 19, 2020 @11:34PM (#22634534)
Also - “If you torture the data long enough, it will confess.”

― Ronald H. Coase

Re:CA governor projecting 25.5M COVID19 cases in CA in 8 wee... by TMWNN on Friday March 20, 2020 @02:47AM (#22635549)
Also "If something cannot go on forever, it will stop." —Herbert Stein
Re:CA governor projecting 25.5M COVID19 cases in CA in 8 wee... by gonzo41 on Thursday March 19, 2020 @11:55PM (#22634699)
I think you mean https://xkcd.com/605/
Re:CA governor projecting 25.5M COVID19 cases in CA in 8 wee... by btilly on Thursday March 19, 2020 @11:40PM (#22634591)
The governor's letter is using a doubling time of 4 days, but there's good reason to believe this 4-day time is due to recently increased testing.

I find this dubious given the widely publicized shortage in tests. Also I question why you are using research from a month and a half ago when there is lots of more recent data to refine those models.

For a back of the envelope, see https://www.worldometers.info/coronavirus/coronavirus-cases/... on a log scale. You'll find a doubling time worldwide very close to what the governor's letter says.

Seems to me the more likely situation is that: (1) there are more cases than we think right now (Maybe not 40x, but maybe 10-20x?). (2) the doubling time will be affected quite A LOT by the measures being put in place. It might drop the R0 to close to 1.

(1) is definitely true given that doctors still can't order a test simply because someone has symptoms of COVID, or even symptoms of COVID and a known exposure route from someone who is positive. (Source, my ex who is a doctor who couldn't get a test for someone today in Orange County.)

(2) is to be strongly hoped for. That is the reason why we are implementing the kinds of lockdowns that we are seeing.

Re:CA governor projecting 25.5M COVID19 cases in CA in 8 wee... by __s on Thursday March 19, 2020 @11:22PM (#22634407)
> there are more cases than we think right now (Maybe not 40x, but maybe 10-20x?)

We know there are more cases than confirmed cases. Order of magnitude at least

Re:CA governor projecting 25.5M COVID19 cases in CA in 8 wee... by vanniv on Thursday March 19, 2020 @11:48PM (#22634652)
I did some back-of-the-envelope maths on this topic using South Korea stats, Diamond Princess stats, and the CDC Influenza and All-cause-pneumonia death dashboard.

Assuming that South Korea catches most of their cases and Diamond Princess caught all of them, and assuming also that those two populations are of the same basic age-and-health distribution as rest-of-world, and assuming that the virus is roughly equally deadly everywhere.

Further assuming that all deaths caused by COVID are deaths involving pneumonia, and that all deaths of elderly, I'll people whose proximate cause of death was pneumonia or pneumonia as a complication of their other illness, and yet further assuming that none of those deaths would be recorded as death by cardiovascular disease or chronic respiratory disease (as those two causes of death are in the top 5, and could hide the entire population having COVID)

All of those assumptions led me to an upper bound for undetected COVID cases as of March 1st of 350,000-400,000 cases. (That is how many cases would have produced a death rate that could have hidden in the moderate spike of influenza and pneumonia deaths from Jan 15th to Feb 15th)

My assumptions are massive, and have wide error bars in both directions, but each assumption either seems entirely defensible or else was mandatory to get an upper bound other than "everyone, plus or minus everyone"

Re:CA governor projecting 25.5M COVID19 cases in CA in 8 wee... by mNovak on Friday March 20, 2020 @12:38AM (#22634939)
Yes, it seems deaths are the most reliable indicator we have (though even this is limited [0]). Looking at highly sampled cases (SK and Diamond Princess) seems to indicate a net fatality rate of 1%.

Thinking about the cohort of infections eventually leading to those deaths, though, means we're really getting a snapshot of the number of infections some 19 days ago (5 days incubation, 14 days from symptom to death [1]). With a conservative 6 day doubling period, and assuming still in growth phase, cumulative infections today is roughly 900x cumulative deaths today.

So ballpark for the US (217 deaths) is about 195k current infections.

Spain (831 deaths): 748k infections

Italy (3405 deaths): 3M infections

[0] https://www.reuters.com/article/us-health-coronavirus-italy-...

[1] https://www.worldometers.info/coronavirus/coronavirus-death-...

Re:CA governor projecting 25.5M COVID19 cases in CA in 8 wee... by vanniv on Friday March 20, 2020 @12:52AM (#22635018)
Your source is yet another avenue for deaths to occur without being counted.

My flu and pneumonia data came from https://gis.cdc.gov/grasp/fluview/mortality.html

Check out the disclaimer on that page, too (even routine things like flu and pneumonia, that happen every year, the statistics are really bad until a few years later when all the adjustments are made)

One interview I saw from a CDC modeller last week was using 5-day doubling, 15-day median time to death, and 1% CFR to estimate, and was saying thus that current cases ~= 800x current deaths.

So, yeah, it's clearly more widespread than reported, and clearly all numbers everywhere are totally wrong (including yours and mine)

Re:CA governor projecting 25.5M COVID19 cases in CA in 8 wee... by grej on Thursday March 19, 2020 @11:29PM (#22634482)
yes, that's a best guess based on Korea numbers and other countries that have done more extensive testing.

But even if it's 100x, you can't get near 25.5M cases in 8 weeks.

Re:CA governor projecting 25.5M COVID19 cases in CA in 8 wee... by btilly on Thursday March 19, 2020 @11:44PM (#22634616)
It all depends on what doubling time you assume.
Re:CA governor projecting 25.5M COVID19 cases in CA in 8 wee... by grej on Thursday March 19, 2020 @11:53PM (#22634689)
Fair enough. I stand by my guesstimate. I hope I'm more correct than the governor.

Also though, NYC is likely to need that hospital ship more than CA.