Monday, April 6, 2020

KEPT WAKE BY NUMBERS


Another troubled night, this time agitated by arithmetic.  Yesterday evening, I heard Wolf Blitzer say that exactly one month ago [i.e. March 5th] there were 11 confirmed COVID deaths in the U. S. and now [last night] there are 9600.  Since a logarithmic graph of this increase, as I found it with a little Googling, is essentially a straight line, it follows that the number of deaths is doubling every 3+ days.  [That is, in one month it doubles between 9 and 10 times, or goes up more or less by a factor of 900.  9600 is roughly 900 x 11.]  If we are two weeks from the peak, that means total deaths to that point will go up by a factor of 16 [ = 2 x 2 x 2 x 2] to ~150,000.

Now, in the bell-shaped curve with which we have all become familiar, the x axis measures days, the y axis measures numbers of deaths on that day, and total deaths are represented by the area under the curve from the origin to that point.

Which means that if the curve really has the shape portrayed, total deaths when the pandemic has died down will be TWICE the total at the peak, or, in our estimation, 300,000.  And that is just over the period March-May!  By comparison,  2.8 million Americans die each year from all causes.  That is about 700,000 in the March-May timespan.

I am not sure we have yet comprehended the social impact of COVID deaths on that scale in that compressed time frame. 

8 comments:

  1. Sometimes you just don't want to do math- even if you can

    ReplyDelete
  2. The integral of the bell-shaped curve which is predicted (by who?) to peak in two weeks, is not an exponential, so the doubling rate should not be constant at the point it reaches the peak, but rather should be noticeably slowing by that point. Instead of an exponential you probably want to be trying to regress against something more like a logistic curve (L/(1+e^(-kx)), with a derivative L e^x / (1 + e^x). (The derivative is bell shaped, but not a Gaussian).
    I think your logic about number of deaths being twice the number at the peak still stands, as does your conclusion about the stress on the health care system.

    ReplyDelete
  3. https://twitter.com/alex_adamou/status/1243898659785375745

    ReplyDelete
  4. David Auerbach: Thanks for that. A clear description of the benefits and shortcomings of epidemiological models.

    ReplyDelete
  5. There will be many deaths not counted in the official statistics. People who need treatment for other serious diseases where the treatment is delayed while doctors and hospitals concentrate on covid; people who die in the recession after the pandemic, or who die in a new wave of austerity; people whose health is weakened by the disease and die of something else a year from now that they would otherwise have survived; people who die of loneliness, and just give up the ghost for the isolation.

    ReplyDelete
  6. https://www.nytimes.com/2020/04/05/us/coronavirus-deaths-undercount.html?algo=clicks_decay_6&fellback=true&imp_id=269181333&action=click&module=moreIn&pgtype=Article&region=Footer

    How are they going to hide the bodies? Simple, don't count them.

    ReplyDelete
  7. Cartoon rendering by Matt Bors:


    https://thenib.com/world-war-c/

    ReplyDelete