What will it take for me to change my recommendation for a total lockdown?
I had a good conversation with Dale
Peterson of Digital Bond on LinkedIn yesterday. He commented on one of my posts
there that he’d like me to post numbers on my “forecasting” accuracy. He noted –
rightly – that there have been big swings in the numbers in my tables from week
to week (and he was being nice – there have been huge swings from day to day!).
I replied that I’m not forecasting at
all. I define forecasting as taking into account relevant current information
to estimate the future value of some variable. All I’m doing is projecting out
the total number of deaths over the next 2-3 months, based on the most recent
three-day growth rate in that number. Day-to-day variations in deaths lead to
day-to-day-variations in forecasts.
Why aren’t I even trying to forecast?
I was originally doing forecasts of total deaths over the pandemic, based on reported
cases multiplied by an estimated case mortality rate. But the problem with that
method is reported cases are a small fraction of actual cases, and that
fraction is clearly getting smaller all the time; I could verify this by
comparing my forecasts with the numbers for actual deaths up to that day. In order
to get more realistic forecasts, I would have had to keep increasing the
estimated case mortality rate, to the point where I would probably now be using
a rate of 100% or more, which is clearly nonsense.
This is because reported cases are
probably not much more than 1/50 of actual cases now. That fraction was much
greater a month ago, but there has been a huge number of asymptomatic people
who have been unknowingly spreading the virus since the very beginning (in
fact, since before there were any known cases at all). A big percentage of the people
who caught the coronavirus from them are also asymptomatic, and they have kept spreading
it, etc. etc. This illustrates the huge tragedy resulting from the fact that we
didn’t have adequate testing available from the start, and what we have now is
still woefully inadequate. Even now, we can’t “waste” tests on trying to find
asymptomatic people, since there are so many symptomatic people who aren’t
being tested, or are still waiting a week or two for results. The only good thing
is those people are hopefully sheltering in place now. If so, they’re likely to
infect their family members, but probably not any others.
The bottom line: Since I don’t know a
good way to really forecast total deaths without having a good idea on total
cases, I have to mindlessly project the current growth rate in deaths; that
inevitably leads to a lot of swings and “inaccurate” forecasts.
When Dale saw this reply, he replied “I'm
not sure there is much daylight between 'projecting' and 'forecasting'. The
reason I think it is important is I'm assuming these projected numbers have a
large impact on your recommendations. So I would expect that significant
variances in actual / expected fatalities would change your recommendations. For
example, May projected fatalities dipped by 50% in one week. Maybe one week
could be a reporting error or cycle issue. If this decline held for a second
week or dipped more I'd imagine you would consider changing your
recommendations or your projection method.”
In other words, Dale was saying that
his real concern wasn’t that my “forecasts” were “inaccurate”, but that I wasn’t
reconsidering my “recommendations” when they swung so much. And I’d further
rephrase his question: “You seem to keep recommending the same thing – i.e. a near-total
lockdown of the US, with a few other conditions – despite huge swings in the
forecast numbers (including changes in total March-June Covid-19 deaths of a
million or more from day to day). What would it take for you to change your
recommendations?”
This was an excellent point, and it
did force me to decide what would make me change my recommendation that we need
a total lockdown. Here’s my answer to that question:
First, in the short run there’s nothing that
would change my recommendation for a lockdown. If we don’t implement a lot more
social distancing than we have now, it’s just about inevitable we’ll have at
least a million US deaths by the end of June – and the epidemic won’t end
there, either. Just look at the numbers below, which reflect the fact that the
rate of increase in deaths has been slowing the past two days, and is now at
its lowest level ever (although the same level was reached about five days ago,
before going up again). We still have about a million deaths through June, and
there’s no reason to believe the coronavirus will call it a day on July 1, or
give us a break so that we can have massive crowds at fireworks displays on
July 4.
Once we’ve gone through a lockdown of
at least a month, we may then be in a position to know how many people are
infected with Covid-19 (either symptomatically or asymptomatically). At that point,
we will need to isolate those people (apart from their families, if at all
possible), trace all of their contacts and isolate them as well. Even more
importantly, the federal government (this simply isn’t something that’s going
to work on a state-by-state basis) will need to convince the public that it has
really identified close to all cases and isolated close to all infected
persons. Then businesses will will hopefully feel safe
to reopen (with a lot of additional social distancing measures, to be
sure), and workers will feel safe to return to work.
Any proposal to fully reopen before
that time (sure, reopening state parks is fine. And remember, 30-40% of the
economy is deemed essential and never closed) is pure fantasy (criminal
fantasy, I might add) and is 100% guaranteed to fail. That’s when we will see a
real depression, both economic and mental. People will realize they have no
choice now – after watching the first total lockdown fail because the other
prerequisites to reopening weren’t met - but to go through a much longer and
much more stringent lockdown, and that many more of us will die long before a
vaccine becomes available (if it ever
becomes available. Remember, 40 years after AIDS appeared, there’s still no
vaccine for the HIV virus).
The lockdown won’t do any good if we
still don’t have massive testing and contract tracing available when it ends, and
there would obviously have to be a complete change at the top of the federal
level for this to happen. The problem isn’t a technical one, but it will
require the federal government to a) take responsibility for the problem,
rather than pretending the states can handle it themselves; b) ride roughshod
over a lot of people who believe that lesser measures are adequate; and c) most
importantly, pay attention to and articulate actual facts, no matter how
uncomfortable (no more circumlocutions like “Well, yeah. I guess we’re still a
little short on actual testing capacity.” Saint Anthony Fauci said something
like this yesterday – proving that his only function now is to prop up the
public’s illusion that experts are really being listened to by the people who
make the decisions in the White House). The team currently in charge isn’t
doing any of these things.
Second, there’s a big milestone we need to
pass in order for me to change my recommendation for a total lockdown, although
I don’t see any way we’ll pass it until we do a total lockdown and meet the
other conditions just discussed. This milestone is that new deaths need to fall to zero (i.e. the percent change in total deaths is zero). This number
has fallen from a high of 159% on March 24 to 4% today, but that’s hardly been
a steady decline.
More importantly, consider that a 159%
change on March 24 (just a month and two days ago!) was only 225 deaths, since the total number of deaths on
that day was 780. Yesterday, the total number of deaths was 54,265. The 4%
change in that number (actually, in the previous day’s number) was 2,048. So
even though the percentage change yesterday is less than 1/50 the percentage
change on March 24, the total number of deaths is close to ten times higher! My
guess is the families of those 2,048 people aren’t taking any comfort from knowing
that their loved ones were a much smaller percentage of total deaths than they
would have been a month ago.
So anyone who says we will turn the
corner soon (let alone that we’ve already turned the corner, as president Trump
seemed to be saying a few days ago, and as David Leonhardt of the Times wrote
a few days ago) has to explain how we’re going to start having a decline in
the actual number of new deaths (the first derivative), not just in the
percentage increase in new deaths (the second derivative). Sure, there’s still
a lot of social distancing in place, but it’s been in place for a month and isn’t
increasing now. It’s literally impossible that this alone will suddenly push
new deaths to zero.
Without a lockdown, yeah, the rate of
increase in new deaths might continue to fall some, but this effect will continue
to be overwhelmed by the fact that the total deaths number continues to
increase by large amounts – and is getting larger in absolute terms most days.
And by the fact that the current trend is toward loosening social distancing rules, not tightening them.
If anyone knows what other event might
intervene to stop the new deaths number in its tracks, I’d like to hear it. But
don’t tell me about herd immunity. That wouldn’t kick in (if it does at all,
which is still debatable) until 40-70% of the population was infected, meaning
there will be multiple millions of deaths by that point. Whether it’s ten
million or six million or two million doesn’t interest me
very much. I’m not writing this blog because I think it will be a great victory
to avoid millions of deaths, if at least a million or two will happen regardless.
The
numbers
These
numbers are updated every day, based on reported US Covid-19 deaths the day
before (taken from the Worldometers.info site, where I’ve been getting my
numbers all along). No other variables go into these numbers – they are all
projections based on yesterday’s 3-day rate of increase in total Covid-19
deaths, which was 14%.
BTW,
don’t get too excited about the fact that new deaths in the week ended
yesterday were 17% lower than the week before. During the week before last, NYC
added 3,800 deaths to the total in one day, on top of their “normal” Covid-19
deaths. These were deaths that occurred at home since early March, because the
EMTs either arrived too late or they decided the person was too far gone to
take to the hospital and make the overcrowding worse than it already was. These
people clearly had Covid-19, but they were never tested.
Week ending
|
Deaths reported during week/month
|
Avg. deaths per day during week/month
|
Pct. Change from previous week/month
|
March 7
|
18
|
3
|
|
March 14
|
38
|
5
|
111%
|
March 21
|
244
|
35
|
542%
|
March 28
|
1,928
|
275
|
690%
|
Month of March
|
4,058
|
131
|
|
April 4
|
6,225
|
889
|
223%
|
April 11
|
12,126
|
1,732
|
95%
|
April 18
|
18,434
|
2,633
|
52%
|
April 25
|
15,251
|
2,179
|
-17%
|
Month of April
|
63,574
|
2,119
|
1566%
|
May 2
|
19,798
|
2,828
|
30%
|
May 9
|
25,635
|
3,662
|
29%
|
May 16
|
34,499
|
4,928
|
35%
|
May 23
|
48,959
|
6,994
|
42%
|
May 30
|
63,395
|
9,056
|
29%
|
Month of May
|
188,587
|
6,083
|
297%
|
June 6
|
85,315
|
12,188
|
35%
|
June 13
|
121,075
|
17,296
|
42%
|
June 20
|
156,774
|
22,396
|
29%
|
June 27
|
210,983
|
30,140
|
435%
|
Month of June
|
677,803
|
22,593
|
359%
|
Total March - June
|
934,022
|
Red = projected numbers
I. Total
deaths
Total US deaths as of yesterday: 54,265
Increase in deaths since previous day: 2,048 (vs. 1,974 yesterday)
Percent increase in deaths since previous day: 4% (vs. 4%
yesterday)
Yesterday’s 3-day rate of increase in total deaths: 14% (This number
is used to project deaths in the table above. It was 15% yesterday)
II. Total
reported cases
Total US confirmed cases: 960,896
Increase in cases since previous day: 35,138
Percent increase in cases since yesterday: 4%
Percent increase in cases since 3 days previous: 13%
III. Reported case mortality rate so far in
the pandemic in the US:
Total Recoveries in US as of yesterday: 118,162
Total Deaths as of yesterday: 54,265
Deaths so far as percentage of closed cases
(=deaths + recoveries): 31% (vs. 32%
yesterday) Let’s
be clear. This means that, of all the coronavirus cases that have been closed
so far in the US, 31% of them have resulted in death. Of course, this number
will come down as time goes on and more cases are closed in which the victim
recovered. But the number has gone down and up since Worldometers started
publishing the recovery rate on March 26 (when it was 41%). It hasn’t showed a
clear downward trend until the last three days. If that continues it will be
welcome, but this is still a very high number. It doesn’t have any impact on my
projections, but the longer it stays high, people need to start questioning the
idea that the final case mortality rate from the pandemic will be 1-2%, which
is what it has been in countries that got the virus under control early. BTW,
other countries like Italy and France also have high values of this number.
Comments
Post a Comment