The news lately includes talk of the USA reaching a “plateau” in the spread of COVID-19. It’s based on incomplete and imperfect data (as has been pointed out by many), but the daily death toll is probably the best indicator and it has definitely stopped rising like it was during late March and early April:
The reason: preventive measures like social distancing, masks, handwashing, lockdown, staying home, all the things people have been doing to protect themselves and others. Those things are working. Without those measures, the death toll would have kept rising and today, it would be horrific.
So: congratulate yourselves. YOU ARE SAVING LIVES.
We also see a “plateau” (if it can be called such) in the number of new cases per day:
The data are less reliable than for mortality, but they also give us a “heads up” because illness precedes death; there’s a lag between new cases and new fatalities. And there are more cases, which means more samples, although irregularities in case counts are definitely there.
Nonetheless, let’s look at the number of new cases each day — just as in the last graph — but this time plotted on a logarithmic scale (note the numbers on the left-hand axis):
We see that from about March 5th until around Mar 23rd, it closely followed a straight line on a logarithmic plot — which means it showed exponential growth during that time span. This was the time that COVID-19 exploded in the USA.
Do bear in mind that there is a lag between infection and detection, so there’s a delay between when the disease spreads and when we see it spread, even in the caseload (and a further delay in the death toll).
From about the 23rd until near the 3rd of April, it slowed (at least, on a logarithmic plot). My hypothesis: this is when the social distancing measures started to “kick in” and that’s when the spread of COVID-19 slowed. But measures had just begun, and it was still spreading.
From around April 3rd onward, the case count has shown a “plateau” (if you want to call it that). My hypothesis: this is the result of widespread social distancing and other safety measures.
Despite the “plateau,” the numbers are still very high. That’s because we were so late implementing the prevention measures we have; the disease had time to spread and reach large numbers. The lockdown measures we have in place have stopped their growth — but new cases (and fatalities) have declined only very slowly if at all. We’re still seeing about 30,000 new cases a day, and around 2,000 deaths.
The virus first hit the U.S. on the west coast (the first reported death in Washington state), then invaded the east coast via massive international traffic through New York. It’s interesting to compare the daily case counts for the U.S.A. per million population, to the same for the states of Washington and New York:
The crucial time for these states was the first three weeks of March. While the disease spread rapidly in New York, it grew slowly in Washington state so that by the time lock down really kicked in nationwide, Washington had kept the infection rate to 50 a day while New York struggled to stay below 500.
Once the disease gets going, it’s very hard to slow it down. Even though New York has done an excellent job, not just halting but actually slowing the disease, it has taken most of April to cut the new case rate in half, slow going indeed compared to early-to-mid March when it could double in only 3 or 4 days.
That’s why KEEPING LOCKDOWN is so important. Once the case load gets heavy, it takes a very long time to reduce it. If we relax our restrictions now and the disease takes off, we can’t put the genie back in the bottle. It’s a matter of life and death.
In case you’re wondering where the case load is growing right now, I estimate it’s where the upward-pointing black triangles are in this map:
The “corn belt” seems to be in the line of fire.
This blog is made possible by readers like you; join others by donating at My Wee Dragon.
Open Mind 
In Washington state, for some days now the number of new cases per day has varied between 125 and, say, 300. So the effective infection rate hangs right at R0=1. Mandatory stay-at-home plus so-called social distancing plus encouraging masks has not resulted in the end of the pandemic via R0 <1.
It is still way to early to lift the lockdown.
Hi, wondered if you’ve seen the misleadingly optimistic predictions from Rutgers Uni about the US fatality rate, claiming 95% certainty that it will be held between 60k and 63k? I link to it in my tweet here: https://mobile.twitter.com/bourbakiaxiom/status/1255824466141016064
But it is being increasingly promoted, with no analysis, e.g. here: https://www.sciencedaily.com/releases/2020/04/200430091256.htm and here https://healthitanalytics.com/news/data-driven-model-predicts-number-of-covid-19-deaths-in-us But it looks like an abuse of curve fitting to me.
[Response: The death toll is already over 63k, and still rising about 2k per day.]
mozols66,
The paper Pham (2020) ‘On Estimating the Number of Deaths Related to Covid-19’ is but an exercise in curve-fitting with zero real-world application.
There is one error in the paper in that the graphs showing cumulative mortality are wrongly scaled – showing 0-5,000 deaths when it should be 0-50,000 deaths.
Plotting out the “new model” shows a peak daily modelled death toll of 2,383 occuring 16th April which isn’t a million miles away from the actual reported 5-day average of 2,990, with cumulative numbers of 32,000 modelled & 31,000 actual. This no surprise as it is an exercise of curve-fitting.
However the model’s predictive ability is poor. The expected mortality as of today (so a week after the analysis period) is modelled at 776 per day when the reported mortality sits stubbornly above 2,000 per day.
As a previous OP set out, ‘lockdown’ works. It prevents the exponential growth of Covid-19.
Sitting here in UK, my worry is that the numpties in charge have no real understanding of how to exit ‘lockdown’. They seemed to be expecting some magic bullet, a vacine, a cure, even a test for anti-bodies although how they reckoned that would help I was never quite sure.
And the ability of a particular ‘lockdown’ to reduce levels of infection does seem to vary. The likes of Spain, Italy, France and to some extent, Germany, the rates do seem to be decling under ‘lockdown’ far quicker than in UK & USA. The OurWorldInData.org numbers show the following drop in daily reported cases & deaths between the periods 10th-19th April and those reported daily since 22nd April. Of course, there could be many reasons for the UK & US numbers being different but they do point to an extended ‘lockdown’ and ‘lockdown’ is not a good state to be trapped in.
… … … Cases … …Deaths
USA … …92% … 85%
UK. … …87% … 81%
Spain . …51% … 68%
Italy . … ..66% … 69%
France …51% … 50%
Germany …55% … 86%
Germany is down from 5000 cases per day to 1000 cases per day. Figure 4: https://www.rki.de/DE/Content/InfAZ/N/Neuartiges_Coronavirus/Situationsberichte/2020-04-30-en.pdf?__blob=publicationFile
With a testing capacity of 800k tests per week we are getting in the region where testing and quarantining people can replace of lot of the lockdown.
The mandatory lockdown was never very strict in Germany, but the population is well educated and well informed and did a lot (and did this early). It is relatively easy to stay at home as we have paid sick leave. We have universal health care, so that people who loose their job do not lose their insurance. Also just 0.6% people lost their jobs since February, most workers are on “short work”, they work less hours and the salary difference is compensated partially by the unemployment insurance and their employer. This way it is much easier to get back to work, no time consuming hiring and training.
Yes, America’s brittle and unequal social safety net is causing a lot of discontent. The right-wing Libertarian types, sponsored by the Prince-DeVos family (though not, reportedly, the sitting Secretary of Education, who has recused herself from agitation)–a covey of whom were shown parading their [semi?] automatic rifles around the chamber Michigan legislature on TV yesterday–are getting more press, and not IMO always sufficiently analytical press either, but there are also more authentically grass-roots actions calling for rent relief and and medicare for all–for instance, the “People’s Bailout” campaign. Things could get very, very ugly in this country. I hope not, but the possibility for real unrest exists. Most people are sensible and well-intentioned, but too many people live paycheck to paycheck–and a lot of paychecks have stopped coming in.
Victor,
The data you link from Germany and data from China, Australia , New Zealand and South Korea shows that it is possible to limit overall damage by following scientific recommendations. That is very hopeful.
Unfortunately for those of us in the USA, the people in charge do not seem able to follow scientific recommendations. We all hope that individual social distancing keeps R under 1.0. In the next month we will learn if that works for this virus.
Saying people should stay at home, but not giving them the means to do so means that too little people will do so, naturally leads to frustration because it takes so long, and the reproduction rate will not go down enough to be able to open up the economy again.
Extremists like Trujmp and DeVos are fine with dragging America down for the pleasure of seeing others suffer. That is not just a matter of being ignorant or dumb, they have different values.
My question: How has Florida managed to turn down the number of cases when they have consistently gotten everything wrong policy wise?
[Response: While its reported case load has declined, Florida’s death rate continues to rise. I have doubts about the caseload data from Florida, in fact about all the data from Florida.]
FWIW: https://covidtracking.com/data/state/florida
What is your projection for total deaths on current numbers, it looks very bad per resolved cases https://www.worldometers.info/coronavirus/country/us/, stay safe
The naive ratio of reported mortality-to-date/reported cases-to-date give very different numbers depending on the country you examine (as shown here -. they vary from 15% in UK Belgium France down to 0.1% in Singapore Qatar) The medical treatment available, the age of the exposed population & genetics are all factors that will surely be significant to the actual Infection Mortality Ratio and the mortality-yet-to-appear may be a significant booster in some countries with recently peaking outbreaks. But there are serious limitations to the accuracy of the reported infection levels which make a nonsense of most attempts to use national data.
There are surveys of infection rates (eg reported here which give a nIFR =0.5% although this is appears to be a simplistic calculation of the ratio while elsewhere what appears to be the same work is interpreted as nIFR=0.79%.)
Yet I am still drawn to the South Korea national numbers which, having tested their way out of their outbreak perhaps points to a IFR=2.5% with mortality occuring in the 5½ week period following the infection report. So if South Korea is testing its way out of the Covid-19 outbreak, how are they missing asymptomatic cases? Could there be some simple reason for their high IMR (& this not apparently the infection age profile which don’t appear significantly different from the popuation age profile)? Or are the anti-body tests used in the surveys finding asymptomatic/non-infectious cases that would not appear in South Korean testing?
I can’t find the data again, but for Queens in NYC, a few days ago there were 3000 deaths from a population of 2 million. That is a death rate of 0.15% assuming that every single person had the virus. So 0.15% is definitely a lower limit for mortality rate.
But I find the slow decline in cases/deaths in some areas (US, UK, Spain) compared to the rapid decline in other areas (Austria) puzzling. And maybe population density is a factor.
“Western Europe, which consists of 10 countries (Worldometer’s definition) if you include U.K. and exclude Scandinavia, has a population of 260mil. The 3 countries of North America has about 500 mil. Combined, they represent 9.6% of the global population. This 9.6% of the global population has 67% of total Covid cases and about 80% of global Covid deaths.
East Asia (China, Taiwan, The Korea’s and Japan) and SE Asia (excluding the Philippines and Indonesia) have 2 billion people (1.7 bil is population of what is called East Asia by Worldometer) or 25% of the global population. They together have 4.1% of all cases and 2.5% of all deaths.
On a per million basis, the “West” has 2930 cases per million. The ‘West” has 253 deaths per million. “Asia” has 68 cases per million (about 1/50th of “West” and 3 deaths per million (1/85th of West.)
Putting aside any one of the 5 million theories that can explain this, one that we can safely exclude is “randomness”. Having never looked at these numbers before, it had never occurred to me that this magnitude of difference was remotely plausible.”
A comment by Michael Rulle at:
https://www.themoneyillusion.com/priorities/#comments
Unbelievably? there were ‘only’ 9 reported deaths in Beijing and 7 in Shanghai?
https://gisanddata.maps.arcgis.com/apps/opsdashboard/index.html#/bda7594740fd40299423467b48e9ecf6
One of the puzzles I’ve had as I observe the data is, why is the Russian mortality rate so low? In North America, the ratio of deaths to diagnosed cases sits in the 5% area; in Italy and Spain, 10%+. But in Russia, only now is that number starting to flirt with 1%. It seems implausible that this is medically-based.
But I found this story, which proposes that there is systemic under-reporting of Covid mortality:
https://www.rferl.org/a/why-is-russia-coronavirus-death-toll-so-low/30585706.html
(BTW, I note that the source is RadioFreeEurope/RadioLiberty, which on the one hand has some reputation for journalistic professionalism, but on the other originated as an anti-Communist propaganda operation. In fact, it received covert CIA funding prior to 1972. So it’s unlikely that they are going to go out of their way to burnish the image of the Russian government.)
Anyway, the story says that standards for attributing cause of death are different in Russia than in some nations. Specifically, attribution tends to go to the underlying condition–heart disease, say, or diabetes–not the proximate cause, which may be Covid-19. Some Russian activists reportedly feel that the government is lying about the numbers, but the story says that this practice affected not just the current epidemic, but previous flu outbreaks as well. Of course, both things could be at work.
Additionally, the story mentions that in Italy and Spain there was pretty widespread postmortem Covid testing, which naturally served to increase the mortality rate. If true, that could well explain the characteristic geographic differences I mentioned above.