Ricochet is the best place on the internet to discuss the issues of the day, either through commenting on posts or writing your own for our active and dynamic community in a fully moderated environment. In addition, the Ricochet Audio Network offers over 50 original podcasts with new episodes released every day.
Coronavirus Update for April 12, 2020
Happy Easter, everyone, and happy Passover for our Jewish friends. Though I’m not quite sure if “happy” is right for Passover, any more than for Good Friday. They are certainly momentous days, demonstrating the mighty power of El Shaddai to deliver His people from bondage, to Pharoah or to sin, and to save us from the Angel of Death. I happen to believe them both, with all my heart.
It’s been almost a week since my last report on Death’s latest weapon. I’ve been monitoring the reports daily, and there have been few changes in the trends, which are favorable in the sense that the worst of the COVID-19 epidemic seems to be over in many places, and close to over in our own country.
Thanks to a helpful hint from The Reticulator on Mark Hamilton’s post yesterday, you should now be able to click my graphs to enlarge them, if you wish. Who says you can’t teach an old dog?
I. Reported Deaths
The US reported 12,401 COVID-19 deaths this week, up from 7,152 last week. In Western Europe, it was 26,736 this week, also up from 24,466 last week.
The US reported it’s worst day thus far on Friday, April 10, with 2,108 deaths reported. Western Europe’s worst day was Tuesday, April 7, with 4,649 deaths reported. (Remember that these are days on which the deaths were reported, not necessarily on which the deaths occurred.)
The daily figures are down somewhat in the last couple of days, to 1,557 in the US today. We have seen minor dips before that preceded new highs, but at least there is a bit of encouraging news.
On to the graphs, starting with total reported deaths per 100,000 population:
You can see that the general trends remain the same, with Spain and Italy at the top, France rising rapidly, the US quite low compared to the others, and Germany the most fortunate thus far.
Here is the graph of daily reported deaths per 100,000 population, which allows visualization of a possible “hump” for each country:
I took France out of this graph, because it has had very wide variations in daily reports which make the graph difficult to read. There is an anomaly in Germany, which actually reported fewer deaths yesterday than on Friday and therefore had a negative daily figure. I assume that this is a reporting issue. It does look like Italy’s peak was around March 27, and Spain’s was around April 2.
You can see that the US reached a peak on Friday, and has declined since. It is possible that Friday was the peak of the epidemic, but I’d want a few more days of data before I’d be confident about such a conclusion.
Here is the 3-day moving average in the percentage increase in total reported deaths:
You can see that the trend lines continue down in all countries, which is quite good news. The US has actually dipped to 10%, from about 30% just 2 weeks ago. France is in this graph, so you can see that the large daily anomalies that caused me to leave France out of the last graph to not have a significant effect on its downward trend line.
Here are the daily percentage increase in reported deaths, for just the US, Italy, and Spain (inclusion of all countries makes the graph too hard to read):
Here you can see the very low percentage increase in the US over the last 2 days, down to just 7.6% today. As previously noted, daily reports fluctuate somewhat, so don’t draw too many conclusions from one or two days of reporting. But again, good news as far as it goes.
II. Reported Cases
Here are the graphs for reported “confirmed” cases, starting with total reported cases per 100,000:
Here is the same data in logarithmic scale, which helps you visualize the declining rate of increase. Remember that in logarithmic scale, exponential growth will look like a line, so a downward curving trend line is what we want to see.
As before, I don’t see any significant change in any of these trend lines that would correspond to the implementation of lockdown measures. It may be difficult to identify such a change, because of the variable number of days between infection and the emergence of symptoms. The jagged line in France indicates some reporting issues, with unusually wide variations on a few days.
Here is daily reported new cases per 100,000 population:
You can see a downturn in the US over the past two days.
Here is the 3-day moving average in the percentage increase in total reported cases:
As with the reported deaths, you can see a general downward trend line for all countries. All countries and areas are now below 10% daily growth.
III. My Comments
I am very puzzled by the progression of COVID-19, in two respects. Well, more than two, but two that come to mind from the graphs above.
First, everything that we have been told — and my own common sense — suggests that reported cases and reported deaths should not be declining in tandem. There is supposed to be a lag between cases and deaths, probably on the order of 2-3 weeks. Yet we see no such lag. The peaks of reported cases is occurring at about the same time as the peaks of reported deaths. The growth rates in total reported cases and total reported deaths are closer than I would expect.
Second, there are very wide variations in the trend lines for a few countries, with some countries being strangely and inexplicably fortunate in having a low number of deaths compared to the number of cases. This is most notable for Germany, and also noticeable for the US.
The US has about 168 reported cases per 100,000, which is about the same as Spain on March 29 and Italy on March 30. Yet Italy and Spain had about 19 and 15 deaths per 100,000 on those dates, respectively, while the US currently has less than 7. It is possible that this is a result of demographics (an older population in Italy and Spain), but Germany has slightly fewer reported cases per 100,000 than the US, and an even lower death figure (less than 4 per 100,000). As I said, puzzling.
ChiCom delenda est.
Published in Healthcare
The official case count is under the actual number of infections.
Like the census
Great suggestion, Reticulator! After clicking on them, the graphs are really easy to read now with my declining eyesight.
Great stuff Jerry, as always.
Regarding the discrepancies between deaths/100K across various countries, I wonder how much it has to do with laxity in making the died of/died with distinction of which we’ve heard so much. That’s something that we won’t have a handle on until we can compare the deaths by other causes for the past few weeks to similar figures for prior years, to decide how many — if any — of the death-by-COVID deaths are truly that.
It has been suggested that Italy in particular may be notorious in this regard. In contrast, I would expect the Germans to be ticking the appropriate boxes.
The Our World In Data site runs a daily doubling-time report for various countries, tracking both doubling times for deaths and doubling times for new cases. I have no idea how accurate they are, how synchronized is the reporting (a special challenge in a country as big as ours), or exactly how they perform their calculations. But I notice that, as of right now, they have our deaths doubling every six days. I think that’s the longest interval I’ve seen since I started monitoring their site a few weeks ago.
You sure are treading on thin ice with the racial stereotype crowd, Henry! But I concur.
Well, at least I didn’t mention the Swiss. I’m sure their tallies are perfect.
It’s politically acceptable to mention that the Swiss are more reliable than anyone else, as long as you don’t imply that anyone else is less reliable than the Swiss.
Official statements and media reports in Germany suggest otherwise: most German communities are also recording Covid-19-related deaths in the exact same manner as Italy. In fact, the RKI (the German CDC) advises hospitals not to perform autopsies or any other post-mortem examinations due to the supposed infection risk, so most hospitals are simply reporting any deaths in virus-positive patients as Covid-19 deaths. (NB: I think this is very cowardly of the RKI; many German hospitals/authorities are very capable of safely autopsying infectious corpses)
One factor that may genuinely be lowering Germany’s death rates is the fact that hospitals here are far from full, so everyone is getting the best treatment. As I mentioned elsewhere, all patients here who require a ventilator get one and the majority of them end up surviving (at least among patients with a known outcome). That doesn’t seem to be the case in Italy or even eastern France, where their ICU capacities were overwhelmed.
However, as unsatisfying as it might be, I think the most plausible explanation still comes back to Germany halting the spread of the virus (through contact tracing and through early lockdowns) before a major spreading event could occur. The fact that Germany has tested so much thoroughly than neighboring countries aids the illusion of lower death rates.
Jerry, if you’re up for another post like this later on in the week I think it would be a good time to revisit Sweden.
Everyone seemed to want to talk about Sweden a week and a half or so ago, but at that time the data was still quite inconclusive. Now enough time has elapsed that it seems like we can talk about an emerging trend.
Unfortunately, Sweden’s authorities seem to take weekends off so there’s always a dip in their statistics every week, but by mid-week I imagine the numbers will be reliable enough again for a 7-day average curve to provide a meaningful comparison to other European countries. My sense from eyeballing the curves on Worldometers is that Sweden has actually done the best job of genuinely “flattening” the curve, i.e. not having a sudden surge in exponential growth but also not experiencing the clear plateau in cases seen in neighboring countries that adopted lockdowns.
But I’d be interested in a more robust charting of their progress, and at the moment everyone else seems to have forgotten about this obvious case study.
I’ve been following the CDC’s influenza page since this all started. Although I didn’t record them all, it seems for the last 3 weeks or so, the number of influenza deaths this season has been stuck at 24,000. That’s kind of interesting, even as the number of flu infections has risen from 34 million to 39 million. As I said, this is from my memory, but I have been looking at it each week.
I look at the Sweden data every day, but am not sure what to make of it.
I recall looking at the CDC’s estimation of flu deaths in the U.S. flu season as of March 15th, and they gave a “range” of 22,000 to 59,000 deaths, rather than an exact number I can’t find the link right now. I know that in Northeast Ohio (the most populous part of the State) flu cases have been drastically lowered since we started physical distancing.
For example, the deaths reported on the ourworldindata site have been way down the past two days, but the same thing happened a week ago. Maybe the people responsible for reporting those numbers honor the full two-day weekend. Their doubling time for deaths is still 7 days, which is a little better than ours, but a lot of countries are at 7 days now.
I wouldn’t be surprised if Easter Holiday had something to do with under-reporting this weekend. They take their holiday time seriously in Europe even if they are not religious.
Mendel, do you have information about the effectiveness of contact tracing? It sure didn’t seem to work in Gangelt. Dr. Wittkowski claimed that contact tracing is not feasible for a respiratory virus, especially one with a fairly long incubation period during which the infected person is contagious. I don’t know the facts, but it never seemed feasible to me to ask someone: “Tell me everyone you have been close to for the past week.”
Your hypothesis about widespread testing in Germany may be right, though again, the evidence is mixed. Switzerland was hit much harder than Germany, thus far, and yet has tested more. Germany’s per capita testing is about the same as Italy, though I realize that Italy has progressed further.
I think a lot of the variation in severity has to do with the luck of nature. The disease is only severe in about a dozen countries or so. In general, if you count up deaths in the top ten countries on the Wordometers website, you will account for 90% of the deaths world-wide. With the exception of China and Iran, those are all medically advanced Western countries, not places where you would expect disease to run rampant.
Yeah, though I’m quite doubtful about the accuracy of reporting from Third World countries. I don’t have any specific information about this. I just have the general impression that they are not particularly well organized, and they are quite poor, of course.
I was just comparing Switzerland with Sweden. And it’s not just an arbitrary pairing. The two are close together in any alphabetical listing, even in Russian!!
Their total populations are similar in size, with Sweden being a little larger. I don’t know how they compare in density of population, but both countries include regions that are hostile to year-round human habitation. Their death totals due to coronavirus are similar, with Sweden currently at 899 and Switzerland at 858. Sweden is currently at a 7 day doubling time for deaths, and Switzerland is a little better, at 8 days.
Switzerland hasn’t locked down completely, but has closed “non-essential” businesses. Sweden hasn’t even closed its schools.
Based on this comparison, Switzerland’s relatively moderate shutdown has gained it nothing over Sweden.
Conclusion: Workers, disperse! You have nothing to lose but your chains.
More Chinese visitors? More accurate diagnoses?
I’ve thought about that too, and there are many third-world countries where the testing and reporting surely must be sub-par. But there are also many first-world industrialized countries that have been little affected by the virus, too. Taiwan, Indonesia, Japan, Ukraine, Russia, Hong Kong, South Korea, Australia, New Zealand, Finland, the Baltic countries along with almost all of Eastern Europe, just about all of South and Central America, and nearly all of the Middle East have had little trouble with the virus.
You could almost make the case that this is a Caucasian Western-Industrialized World Virus. Give you any ideas for a new graph??
Good point, though I don’t think that some of those countries are actually first-world. Japan, Australia, New Zealand, and Finland definitely, HK and S. Korea probably, Taiwan maybe.
In looking at it now, I see that Sweden is at a 7 day doubling time and Switzerland is at 11 days. I’m guessing I made a mistake rather than that the numbers have changed since then.
I also see that IMHE projects a lot more deaths for Sweden than for Switzerland, even though the two countries are now neck and neck. We’ll see.
I think that the doubling time on Worldometer is an unstable measure. I haven’t looked at how they perform the calculation, but I assume that it’s based on daily percentage growth, which is a figure that bounces around a bit.
Looking at the long-term graphs — which I provide, lucky for you! — is a more reliable indicator. As demonstrated in the OP, the trends are down everywhere.
No, it appears to be just what it says it is, the number of days it took to double to the current number. Nothing unstable about it. If anything, it’s more stable than some might like. I’ve taken to doing it myself by eyeball when calculating the doubling times in Michigan or New York from the data at covidtracking.com. I take the current number of deaths, divide by two, and see how many days back the deaths were at the half-way number. Simple and easy. There is some built-in curve-smoothing, it doesn’t require the labor of computing a running average, and gives me the information I want as to how things are trending.
The only question in my mind (and I haven’t bothered to figure it out) is how they round up or down to whole days. When doing it in my head I round to the nearest half-day. Anything further is probably a false precision.
I don’t see how your graphs are more reliable, but they do show more information at a glance than what I see when looking at covidtracking.com. And that’s worth something, too.
To your first point: I’m not sure (and remember, I’m not an epidemiologist, I just sat through lots of epidemiology talks at conferences during the bird flu and swine flu scares).
Dr. Wittkowski has lost some credibility with me lately (more later), but a recent paper in Science by other quite credible researchers also suggested that normal contact tracing is too slow to stop a respiratory virus that may potentially shed before the infected person notices any symptoms. However, these researchers propose that a smartphone app that contains an individual’s movements over the previous four days may be able to stop the spread since it can react instantaneously.
I think there are two arguments suggesting that contact tracing and quarantine might indeed be effective. One is South Korea, which had such a comprehensive (and intrusive) contact tracing/forced isolation system that they essentially pulled off the feat described in the Science paper mentioned above. That might be a reason why the South Korean curve looks quite different than those of many other countries.
The second argument is that contact tracing/isolation doesn’t have to be perfect to make a big difference. One of the under-appreciated aspects of this pandemic, in my opinion, is the fact that it is not a global pandemic but a series of major local outbreaks together with background-levels of infection in the rest of the world. In other words, there seems to be some sort of threshold (based on infection rate? rate in increases in cases? Viral load? Geographic density of cases? Who knows) that needs to be passed – above this threshold, a region is screwed no matter what they do; below this threshold, measures like contact tracing and lockdowns are very effective.
If this (highly speculative) hypothesis were correct, it would mean that all contact tracing (or other social distancing) measures need to do is keep the local spread of the virus below the critical threshold, not to completely block the spread.
If you listen to German experts, they attribute much of the success at stemming the outbreak here to early action.
Part of this is helped by the fact that the German “Anthony Fauci” (Christian Drosten, who isn’t Fauci’s exact equivalent but plays a very similar role) has focused his career on SARS and coronaviruses in general for the past 20 years. So when the rest of the world was looking at the initial outbreak in Wuhan and thinking, “huh, that’s interesting”, he and his colleagues were already assuming it would hit Germany and the rest of the world and started planning.
It’s no coincidence that the most widely used PCR test worldwide – including in the US – comes straight from his lab.
Although Germany is actually much more decentralized than most people realized (in many ways power is as devolved here as in the US), researchers and top medical officials have close-knit informal networks. So Drosten and friends alerted major medical centers around the country to keep their eyes open for suspect patients and to start establishing their PCR tests well before the first cases were detected. Once those cases indeed showed up (around the same time as the ones in Italy), local doctors and health authorities did indeed quickly quarantine all infected patients AND all of their known contacts.
And it makes some sense that a robust yet imperfect policy of contact tracing early on (when it’s still possible to conduct thorough interviews with all patients) can be much more effective than a nearly-perfect contact tracing system implemented once an outbreak has already reached a critical mass.
We can take this one step further: even within the hardest-hit countries, the epidemic is still a non-issue for the vast majority of their populations.
If you look at maps of how the outbreaks spread in countries like Italy, Spain, or France (or the US), it’s very clearly focused on fairly discrete regions of the country, while the rest of the country has near-background levels of infection and death. That’s why I’ve been railing against grouping statistics into countries for weeks now – countries are too arbitrary a delineation at the present time.
Moreover, even within countries there’s often no logical explanation for the uneven distribution. For example, the majority of cases and deaths in Italy have been restricted to northern Italy, particularly Lombardy, whereas southern Italy remains fairly untouched. But as anyone who’s traveled the country knows, southern Italy is much poorer and less-developed than northern Italy. Indeed, in terms of wealth, development, and lifestyle, northern Italy is more akin to Germany or Switzerland than to southern Italy.
Going back to Germany, it’s worth comparing the distribution of cases geographically across Germany vs. Italy, Spain, France, or the US: Germany is one of the few larger countries where the cases are distributed much more evenly, instead of having one or two hotspots and near-zero activity elsewhere. Much of this is due to the fact that the virus was brought to Germany by tourists returning from ski vacations in the Alps, which meant it was spread at very low levels to many regions simultaneously. Since our public health authorities were keyed into the threat so early, managing those diluted outbreaks was like extinguishing lots of brush fires vs. trying to put out a hot, dense forest fire.
TL;DR: my take-home message is always: looking at static statistics, like country-wide figures, total numbers of tests/cases/etc., or taking single snapshots in time misses so much important context as to be worthless.
Looking at dynamic maps of how the virus spreads over time and on a regional (county- or state-level) is only a small step up in complexity but provides a much different perspective on the pandemic. Similarly, trying to imagine the transmission of the virus as a non-homogeneous, discontinuous phenomenon instead of a phenomenon that always occurs at a constant equilibrium also makes the apparent contradictions in data seem much more palatable.
I’ve already spammed poor Jerry’s thread too much here (as is my wont), but I do want to return to this for the data nerds still reading:
I scan the current literature on Covid-19 as part of my work and I came across a manuscript yesterday by Dr. Wittkowski that enshrines his “two strains” hypothesis (the good bug/bad bug theory) in writing. Unfortunately, it does not bolster either his theory nor his general credibility in my view.
Aside from a general sloppiness (on both a scientific and readability level) that is very sub-standard for somebody with his purported experience, he relies on very simple datasets to make a very complex case but doesn’t provide any further evidence to support his two-strain theory, despite the fact that much richer datasets actually exist. For just one example, he consistently hypothesizes the presence of two major, distinct strains, but never actually relies on the huge body of publicly-available sequence data to test this hypothesis. Instead he merely says “Only sequencing samples from these countries can help to answer these questions,” which is fairly ridiculous since thousands of such sequencing samples already exist.
Furthermore, his manuscript (using a data cut-off on March 31 and presumably written during the first days of April) makes two predictions of the total death count in America: if the mild strain were to predominate in the US, he predicted a maximum of 6,000 deaths over the course of the epidemic. If the more virulent strain were to predominate, he predicted a maximum of 12,000 deaths. Yes, that’s his absolute worst-case scenario for the US, and with no distinction for “dying with SARS-CoV-2 vs. of SARS-CoV-2”. Obviously, we’re currently at about twice that figure.
Based on this, I poked around Wittkowski’s publication history for a few minutes. He has indeed worked at several renowned departments, including with a famous HIV epidemiologist back in the 80s/90s. However, his research specialties (apart from HIV in the 80s/90s) have focused on psoriasis and several technical biostatistics issues related to new genetic technologies, not viral pandemics and never respiratory viruses. I wouldn’t label him as discredited by any means, but there are clear indications that his work and his theory don’t merit any special attention.
For examples of “contrarian” positions that are much more deserved of our attention, I would point to:
This article that uses a well-established reporting network for influenza-like illnesses (i.e. patients who go to the doctor with symptoms of a respiratory virus) which shows increases in patients with symptoms of respiratory viruses but who tested negative for influenza itself – specifically, increases during March 2020 in states whose Covid-19 cases were rising during that period. They use this correlation to assume that Covid-19 accounts for these increases, and extrapolate the numbers of ILI reports to suggest the number of actual Covid-19 cases may dwarf the number of PCR-confirmed cases in these states by factors of 100-fold or more. In contrast to Wittkowski, this paper relies on thorough and traceable data and transparent equations to draw its conclusions while still acknowledging the limitations of their modeling approach.
This article by John Ioannidis uses the highest-quality data available from regions with statistically meaningful numbers of recorded patients to compare the risk of Covid-19 to healthy people below the age of 65 with other daily risks. While I think it oversimplifies somewhat (as is Ioannidis’ wont), it still uses solid yet differentiated datasets and transparent calculations to reach its conclusions.
If you actually look at it by state. The areas were this flu season was already bad before COVID-19 out break is also were COVID-19 has been the worse. New York was already at the second worse year in the last 10 years for the flu before this. Both the city and state. LA was also already bad.
Where as states who were having a more mild season. Also seems to have a more mild COVID-19 outbreak like Florida. I have not looked at all states but there is a trend there. I still say it looks like we are going to be at or a bit above 2018’s numbers related to deaths. With places like NY being worse but other places like Florida being a lot better.