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.
Day 66: COVID-19 Numidiocy
I don’t know whether the word coinage “numidiocy” is unique to me. It is a contraction of number idiocy. That is, whenever numbers get so numerous that it ceases to convey clear and useful information. I am not autistic. I lack the focus (and likely the faculties) to scan tables and graphs and inerrantly sense the important from the irrelevant. I hear the arguments that people make for how to order the significance of this datum versus that. In the screengrabs above I have ordered them by “active cases,” for instance, believing that “total cases” do not necessarily best reflect the present challenge.
And then there is the time element. In the screen grabs above I show Yesterday versus Now. This Worldometer chart is updated continuously as and when information is available. The current counting restarts daily at 0:00 UTC, which is 8 p.m. EDT and 5 p.m. PDT. Imagine each column that includes “new” in the header being a bucket into which water is poured and measured throughout the day with the water coming from lots of vessels of varying sizes being dumped at various hours of the day. Then the bucket is kicked over and emptied and the filling begins again.
So screengrabs of tables at random times tells of a moment, but it doesn’t provide context. And that is why graphs can be useful. @snirtler focused my attention on 91-DIVOC that provides some outstanding graphical displays of information. (91-DIVOC reflecting COVID-19; clever, right?) It lets you slice and dice data, time, place, numbers, and trends both in linear and log form. It wonderfully juxtaposes information about countries and US states in a seemingly useful manner until I realized what it wasn’t showing me and the data noise that it was.
In each chart, there is a dashed line — straight in the log presentation, curved in the linear presentation. That line represents 1.35 daily growth:
In nearly every country in the world, when the virus reaches 100 people the number of cases begins to increase by 35% daily. (Dashed black line.)
With that dashed line in place, you can clearly see countries or US states progress through time at either greater or less than 1.35 growth. The implications are clear: Countries and states with sustained growths above 1.35 are moving into greater difficulty; countries and states below 1.35 are moving into lesser difficulty. The charts comparing countries and states by population start to add noise because what does it mean that Vatican City and San Marino are so far above the 1.35 line on a per capita basis? (Somebody really needs to work on the dataset for these small places.)
As I stared at these charts I began to ask myself what is the marginal utility of the data? Yes, countries and states above the dashed line are having more cases, more deaths than they would have if they were below the line. But where is the “existential threat” line for that country or state? Where is the line where the health system fails? Where is the line for a set of cascading events that condemns them to a season of desolation and of indeterminate length? And the placement of those lines varies by locality, not just country or state.
I think it was back in the ’90s when the concept of a “dashboard” for management came into vogue. Like driving a car there would be a limited number of data outputs — speed, temperature, battery, RPM– that could be quickly scanned to determine that things are running just fine. Yes, there could be warning lights lit when the system said some combination of events were not in order. (A “check engine” light begs inquiry, it doesn’t render a verdict.) And senior managers would gather periodically to review the color-coded dashboards that reflected algorithms fashioned within the bowels of the various departments and operations. This was considered state of the art management.
But we all know the “decision makers” were far away from the point where things went wrong. The line mechanic, if properly trained, could see where welds were failing, where lubricants through addition of grit and incessant thermal assault had lost their ability to do the job. But the mechanic did not control supply chain for needed maintenance, the budget for supplies, the credit line that secured funds when revenues were unsteady. Somewhere in the dashboard the data all came together as green, yellow or red. The check engine light lit, or it didn’t.
Crash scene investigators exist because either our dashboards are faulty, misunderstood, or ignored. Sometimes all three.
[Note: Links to all my COVID-19 posts can be found here.]
Published in General
Rodin, good points.
I’ve been refining my analyses as time has passed. I initially did much the same thing as the first graph in the OP, which started each country’s trend line when the country passed 100 cases (in my case, I picked 200 cases). But 100 cases in the US is not the same thing as 100 cases in Italy (with about 18% of US population), much less Luxembourg. So I switched to cases per million.
But there are issues with case reporting, as it depends on testing, and testing has been rolled out at different times in different countries. So more recently, I’ve been switching to starting the trend lines based on reported deaths per million.
It’s hard to know where to start the trend lines, and it can make a big difference. For example, in the first graph in the OP, if you shift the US trend line just 5 days to the right, the US does not stand out; and if you adjust by population, the US does not stand out.
I would look at deaths either total or new deaths or both.
Corpses are like cash flow. Corpses don’t lie.
Doesn’t it simply mean that they are extremely small populations with high population density? Vatican City is a city and thus could be compared statistically to other cities. How does it compare to Rome, Milan, or New York City?
Rodin,
Thanks much for this interesting data. I have no comments on your interpretation of it, only on some general comments you made about the process of interpreting scientific data.
In my view
…without having a clear scientifically answerable question in mind are the CAUSE of the confusion.
Many of the problems I see us Ricos having in getting correct, clear and useful conclusions from data are not caused by getting more data. They are caused by errors in interpreting the data.
As you stated, it’s only a snapshot of reported data, but when I refreshed the Worldometers USA Coronavirus page just a few minutes ago, the number of reported recoveries jumped up to 1863, surpassing the number of reported deaths (currently at 1080) for the first time.
Even I didn’t realize just how small Vatican City’s population is. The CIA World Factbook estimates it at about 1,000 people. Even for a city, that’s tiny.
Also, the CIA doesn’t offer its usual statistic of median age, but Vatican City is mostly populated by old men (clergy).
That considered, I thank the Lord it hasn’t been hit harder by disease.
Like a great many people, I’m struggling to make sense of a situation for which a paucity of usable data exists — and making most of the mistakes of assumption and misinterpretation that are likely to spring from that. I try to liberally sprinkle “we really don’t know yet” into my comments, but occasionally fall short of even that modest standard.
I’m quite frustrated by the lack of actual reporting on things that seem relevant. Take testing for example. How has increased testing impacted actual case counts? What’s the testing methodology in NYC and other hotspots? Are we doing any testing aimed at determining the prevalence of infection within the population, or is it being done only for infection confirmation and treatment planning? What’s the state of antibody testing that will allow us to begin to count the number of infected and recovered individuals hidden in the general population? Are cases reported today the same as cases reported two weeks ago — that is, sampled from effectively the same population of patients presenting at hospitals with somewhat serious symptoms? Or has the average subject changed in a significant way, so that we’re sampling a different population?
How are “recovered cases” counted? Is it a meaningful figure? What proportion of all cases are likely to include some kind of follow-up to confirm that patients have in fact recovered?
What are hospital admission policies? New York City claims about 4,000 “ever hospitalized cases,” (whatever that means) as of last night. Rodin’s figures above report a U.S. total of 1,452 “serious, critical” cases in the entire country. So how many of NYC’s 4,000 are serious, and what are the rest of them doing in the hospital — or are the immediately discharged after a short evaluation?
I have a friend in Florida, a head nurse at one of the larger hospitals, who says that her hospital now has a COVID wing and that their policy is to check into the hospital anyone who tests positive, regardless of seriousness of symptoms. How many hospitals are doing that, and how does that impact our understanding of hospital capacity?
Many have written of the death of small-town newspapers and, with that, the effective crippling of local reporting. There are no longer journalists available to cover school board and town hall meetings, to report on the “small” issues that used to fill the local papers. And that means there are no longer people available to cover the local hospital, to report on admissions and critical needs. We count on web sites providing aggregated statistics, but know next to nothing of their processes, their gaps and overlaps and accuracy.
It’s nice to imagine that the quality and availability of data is profoundly different at the highest levels of state and national government. It certainly could be: they have the resources to remove a lot of the ambiguity and so understand the numbers better. But it’s frustrating to those of us sitting at home, trying to make sense of it all. And the sensationalist politicized garbage of the national news provides little illumination.
There is a problem with Worldometer calculation of Total Cases/1 Million population for places under 1 million in population. Vatican city has ~1,000 occupants and 4 cases. This should be .004 per million not “4,994”. And since 91-COVID is drawing on Worldometer tables, this error is injected into their graphs as well.
Can you show us your computation? How did you get .004 per million ?
I appreciate your work on the graphs, @arizonapatriot. They help put a lot of this in an easier form to understand. Making clear and understandable graphs out of raw data is an art form, especially when trying to show multiple or competing trendlines. A lot of stuff I see published does not do this so well. I haven’t seen all of your posts on this, but I am trying my best to keep up.
3 million filed for unemployment this week.
are we willing to accept this trade off?
I vote no
Any cost/benefit calculation must necessarily involve a comparison of a cost and a benefit.
So, are we willing to accept this trade off against… what? Until we feel reasonably confident about the trade we’re making, it’s hard to say.
If I was working an ER in New York City right now, I’d be wearin’ my poopin’ pants. But I don’t, I live in a small town in the north west corner of Washington State. I can go for a walk at lunch or in the evening and see 10 people in an hour. I’m not worried about me. I have the luxury of looking at the data and thinking about what it means on a macro level. But those folks working the ERs? The graphs don’t mean a damn thing. All they know is how many people they put in body bags today.
None of that means we should be freaking out….just means that depending on where you live and what your situation is has an impact on how you are reacting to all of this.
True, but also like cash flow, corpse quantity can be relative. Good cash flow for my accounting practice might be a rounding error in the cash flow for AT&T. If we want to compare the two then rates and trends are better than raw quantity. Although raw quantity also has its uses and implications.
I have been extremely skeptical of the “Recovered Cases” category on that web site, because for a long time they gave nearly every single case in China the status of “recovered” while there were hardly any recoveries in the rest of the World, even after two months data. I attribute this (without any particular knowledge) to the fact that 80 -85% of cases are so mild that they do not warrant hospitalization, so the patients are just sent home to recover on their own. I am guessing that hospitals and doctors don’t generally call those patients at the end of the week and ask “Are you better yet” and if they do, and the answer is yes, they probably don’t phone the Centers for Disease Control to tell them that Joe Fabitz is feeling fine now. I’m guessing that the only people put in the “recovered” category are those who were hospitalized and released. I could be wrong.
*Diagnosing* more cases.
We have already made the COST bet, regardless of the benefit. This is pretending the cautious path is shutdown, instead of realizing that every decision has costs and benefits, and we need to work from what we have.
– Decide what economic downturn/unemployment we are willing to tolerate for this.
– Show me a chart that says when we will get there.
– Match with corona curves. Prepare for that many cases and get on with the fight. Protect the vulnerable. Crank up productivity.
Maybe. Maybe not. It could be that some country has a very old population with history of respiratory issues and that country has a infection spread wildly within the community of people already being treated for health issues. That morbidity data will be telling a lie, when applied to other countries. Some countries use different definitions for cause of death. The US has a lower life expectancy, because we count babies that die soon after birth. Other countries don’t count those losses.
Cha-ching!
I can answer a little bit of this since my wife is doing Covid 19 testing at a major hospital in Cleveland.
In Ohio and I think almost all other places(?) they are only testing people who are showing symptoms and have been referred by a doctor. Along with these, they are testing hospital personnel who are with at risk patients or otherwise at risk. They are not testing the entire hospital staff …..yet. If they let just anybody show up, there would quickly be a million people in line and it would overwhelm the system and push the actual sick people to the back of the line.
Of these sick people, they are getting a positive test rate of 5% which is in line with a national figure I heard at 4%- 5%. Of the at risk health care workers it is higher but I don’t know the exact percent. In Ohio overall, 16% of those infected are health care workers, who presumably picked it up from patients. This number might be artificially high in comparison because many healthcare workers are being tested automatically, whether they exhibit symptoms or not.
The only random testing that I have heard of has been done in tiny Iceland. They are finding that something like half of all infectees show no symptoms at all.
https://www.dailysabah.com/world/europe/icelands-mass-testing-shows-half-of-covid-19-carriers-have-no-symptoms
Yes, I would agree. Particularly in regard to China, Iran, Venezuela, and other authoritarian regimes, the data they are providing can’t be trusted, it’s more like propaganda. The charts are only as good as the sources they’re collecting the data from.
In regard to the recovered category in the US, I too see it as only those who have been hospitalized and released. The grand total in the recovered category (mild cases, untested cases, no symptoms cases, etc.), which I’m guessing is a much larger number, may not be known for a while. We’re definitely lacking key data to be able to properly calculate the disease’s mortality rate. This WSJ oped discusses these issues much better than I can.
@misterbitcoin, I think you may have caught me out. Let’s see: There are 1,000 residents, .001 of a million. 4 cases X .001= .004. I may be missing something. Let’s try another way: If the Vatican had a million residents, how many cases would the have if they have if they have 4 cases for each 1,000 residents? 4,000. So the ratio Cases/1 Million population is intended to normalize for different populations. When all of the populations are in the millions that is not too bad a ratio. But when the populations are vastly different the significance is lost. 4 people is still 4 people, not 4,000 (or nearly 5,000 as shown on the graph). So these small populations add noise but not insight.
Bottom line? We still don’t know the denominator (total infections) so we therefore don’t know the rate of serious or mortal infections. We don’t know, but I suspect that the actual denominator is much bigger than the numbers in charts like these. How can it plausibly be otherwise with so few actually tested and assuming 1) a high spread rate and 2) that reported cases and serious/mortal cases tend to correlate?
That said, if 20 million in the US are infected and even if the actual mortality rate is 0.50% then that would be 100,000 people. If it’s 1% then that would be 200,000 people. The raw numbers would (could) be big.
Another unknown to ponder: of those who have died or will die from WuFlu, how many would have died this year anyway from something else? I don’t see how we can ever know something like that, but I bet given good granular statistics on actual cases we can come up with a reasonable range of results.
I hope and pray that we’ll be getting back to normal by Easter. Being out of work causes suffering too (I’m not out of work at the moment, but I know people who are).
I’m glad this Comment ended when it did.
My head was nodding in agreement at least once per sentence, and by the end I think that some of my vertebrae were starting to wear out.
Even getting back to normal, the new normal will not be the old normal, at least for a while.
Me too. I suppose as conservatives we really shouldn’t be surprised by poor reporting, although frustration is another matter. After all, that’s how Rush, talk radio, Fox, and now conservative (and other) podcasting came to be. Like the political establishment, much of the media establishment is either incompetent or duplicitous. Also like with the political establishment, I don’t much care anymore which (incompetence or duplicity) applies. It’s all crap, and often dangerous crap. Russia hoax and covid hysteria are of a piece. Be skeptical of everything and look for original sources when possible.
I watched one of the press conferences yesterday. It was astounding how so many were trying to get the president or his advisers to say that the president is ignoring the advice of the health experts in his response.
First of all it’s quite obvious that he has many advisers and he is listening to them and putting them out front. Secondly, it’s also quite obvious that health advice (presumably death tolls) is only one aspect to consider in a public policy response. The way the questions were being asked indicated either that the reporter was looking for a gotcha or that the reporter genuinely didn’t understand that law and economics were equally relevant assuming that at some point one or all of those would come into conflict (if there were no conflict then it literally wouldn’t matter).
Except to create that wing the hospital probably stopped all elective procedures. Hospitals are going to end up broke, if they weren’t already.
Some might go broke. I think the current feel for ambulatory outpatient surgery centers seems to be to begin ramping back up on elective cases in May-June and back to full strength by July (obviously this could change). There are some which can’t last that long without major intervention, while most will be some degree of ok.
The flipside is that cash inflow is roughly 30-60 days lagging performance of cases while cash outflow (except for labor and rent) are roughly 0-30 before a case is performed. So as long as insurance and government payors keep paying a two month disruption should be somewhat doable. The employees will suffer, though, as staff is cut to essential levels, and even with good faith efforts to get something for them by both employers and governments, it might not quite cut it for most people already on the edge or underwater. One we start ramping up then suppliers and vendors will need to step up a little to help get past that rampup where facilities will need to purchase supplies for current cases without much in the way of collections on prior cases anymore.