Wednesday, July 29, 2020

Covid numbers have huge error margins

It helps to show that the virus is really inconsequential.

Because lots of people have been obsessively tested for almost half a year, lots of numbers were produced in the world. The Worldometers statistical page contains a large amount of numbers that describe the "cases" (positively tested people), serious cases, deaths, number of tests in the world's countries (and their states) and their historical evolution.

The two most important official statistical pages in Czechia are Illness-currently and Koronavirus-Mzcr. Let me discuss what was happening to the numbers on the first page.

OK, the first six boxes on the Illness-currently page are: tests performed (daily increase), total confirmed cases, active cases, cured people, deaths, and currently hospitalized. A special page on hospitalizations gives you the graph of hospitalizations and serious cases, too (click at "tabular review" which is what the automatic translation calls it).

OK, the number of cases has unsurprisingly shown some new signs of life when Czechia threw away face masks and restrictions basically everywhere on July 1st. We're getting 100-300 new cases every day these days (the lows were as low as 18 cases per day although these were weekend figures when the testing is limited). The "number of cases" graph shows a clear "second wave" and the total number of cases is 15,799 now. This "second wave" means nothing, as I try to argue with these new observations.

The number of deaths is 373 right now, about 1 is added every day in average in July. Maybe the rate doubled from 0.6 in early July to 1.2 now. The numbers are so small that the statistical error margins are high. But let's look at the number of cured people and active cases. Those were changing slowly in July and suddenly, on Sunday, the number of cured people jumped from 9,000+ to 11,000+, almost by 2,000, and the number of active cases correspondingly dropped from 5,500 or so to 3,500 or so (it is 4,055 now).

Why did it drop almost by 2000? Well, I expected that drop in advance because the minister of health had ordered an audit of the cured people. And they found out that almost 2,000 people were already cured – meaning having had 2 negative tests with at least 9 days in between – but this fact hasn't been imprinted to the databases on which the website is built. You can see that a simple audit lowered the number of active cases by some 30%.

Yesterday, on Tuesday, another big change occurred in the "number of hospitalizations" and I didn't expect this one. After 0:30 am on Tuesday, the number of hospitalizations showed about 150, where it approximately was in recent days as well. However, at 5:30 pm, that number – which normally changes "at most by a few" each day – dropped to 108. It means that a whopping 42 people were suddenly "dehospitalized" within hours. How could that happen? Shouldn't the hospitals and the ministry know how many people are actually lying on hospital beds?

I couldn't figure out the correct answer by pure thought even though I believe that the answer is so obvious that you should be able to find the correct theory by pure thought! In the morning, the explanation appeared on the website. These 42 people were... people who are actually still lying on some hospital beds, and they were there a day earlier, too. But... the reason for their hospitalization just cannot be said to be Covid-19 simply because they have been Covid-negative for a long time and the infection was just an episode in their long-term hospitalization that had started before any Covid was talked about.

This subtlety may sound as a hardcore detail and as I described it, the situation almost sounds contrived. But this description actually applied to 30% of the hospitalized patients! If you know something about the relative contributions of the virus to the deaths (and severe health complications) of the people, you can't be surprised. The deaths reported as "Covid deaths" are actually just "deaths while Covid-positive" and most of them are actually not primarily due to Covid. The boss of the Czech health statisticians Mr Dušek has estimated that 40% of the "Covid deaths" are "deaths mainly due to Covid".

So it makes a huge difference whether you count the hospitalized people "who have been Covid-positive at some moment" or the hospitalized people "who are Covid-positive now". Indeed, the difference is 50% of the lower (second) number or 30% of the higher (first) number!

The people who are obsessively watching all these numbers as if they were important – people who think that every change of those numbers by 5% is a big deal – would have to be shocked by these "audits" of the Czech ministry of healthcare which have almost instantaneously lowered the number of active cases and then the number of hospitalizations by 30%. But it is completely wrong to criticize the Czech ministry of healthcare. They did update the numbers and the newer ones are clearly more accurate than the older ones. It makes sense to remove the people who are already negative (twice) but weren't "shouting" about this fact to imprint the news into all the bureaucracy. And it makes sense to remove the hospitalized people whose reason for hospitalization is currently "really very different" than Covid-19 again.

One may also justify that these corrections weren't applied much more quickly. The doctors – and even statisticians – surely have a higher priority to "deal with the patients who really need it" than to "quickly update the databases about the people who don't really need much now". So those 2,000 cured people who waited to be cured in the database may be considered just a justifiable "delay" (perhaps comparable to a month) and the average delay was "abruptly shortened" by the audit. Similarly, the 42 hospitalized people could have spent some time on the list of "Covid hospitalizations" just to be sure – but as the disease gradually ceases, it gradually becomes ludicrous to consider them "Covid patients". There was some flexibility about "how much time after the Covid infection" you may still be counted as a "Covid hospitalization" and this extension was getting longer and ludicrous. Another audit brought this timescale to reasonable proportions again – which resulted in the 30% drop in the database number of "Covid hospitalizations".

Everything makes complete sense and these changes were so high, 30%, simply because the natural error margin associated with the "health impact of the virus" is very high. And it is very high simply because the quantity itself, the health impact of the virus, is very small. It is really a very small fraction of the "reason" why the employees of hospitals have some work to do. And if a variable is small, while the fluctuations caused by competing variables are kept fixed, the relative error margin of the first variable is obviously high!

The crucial point is that Covid-19 is often presented (by the hysterical demagogues) as something binary and extremely important, something that creates the same difference as the difference between Heaven and Hell. So you should surely be able to divide people to yes-Covid and non-Covid, this set of assumptions implies. But it's not true at all. The actual impact of the virus is small, gradual, and when you count some people as "affected by Covid", it matters a great deal how you exactly define "affected by Covid". It greatly matters whether you consider people who are infected now (and it matters whether you demand the PCR test to be from today or if the 4-day-old test is enough) or were infected in the past, if you use the latter, it matters which antibody test you use and how high concentrations of the antibodies you demand (and/or whether the asymptomatic carriers are allowed), if you count deaths or hospitalizations, it matters whether you require Covid to be the most important reason or the first reason or something else. All these choices may change the numbers by 30% or even hundreds of percent. "Affected by Covid" is almost as fuzzy as "affected by America". Which people are affected by America? Almost all, almost no one, it depends on the details.

I don't have the newest number but the Czech hospitals had something like 15,000 hospital beds a few years ago. I do have the March 2020 numbers measuring some subsets: there were 823 ARO beds and 3,658 ICU (more precisely JIP) beds. ARO is "Anaesthesiology-resuscitation department" and it usually supports critically threatened patients, too (and maybe, people go through such a bed when they need to be "made sleep" before a surgery etc., I am not an expert). It's possible that the English term ICU includes both the Czech ARO and the Czech JIP – and we have almost 4,500 of such beds in total.

Just 108 of them are reserved for "Covid patients" now. And again, even at this very moment, a huge fraction of these 108 hospitalized patients are "patients primarily hospitalized for other reasons" who just "happen to be Covid-positive right now". The patients hospitalized "really mainly due to the Covid virus" represent at most two percent of the Czech ICU beds.

In some other countries and states, the actual impact of the virus was or is higher. But it's always just a temporary state of affairs, something that can't last for periods much longer than a month or 6 weeks. Most people simply recover (or die) within a month and when lots of people are involved in epidemics, the disease runs out of victims and unavoidably drops. Even at places where you have "overfilled ICUs", it's really crazy to get anxious or build new hospitals. Basically every new hospital (or hospital building) that was built because of Covid anywhere in the world was an utter waste of money.

While the health impact of the virus is very small, the economic impact of the hysteria is huge. The U.S. GDP numbers from Q2 2020 will be released tomorrow; some other countries should do the same this week. The year-on-year drops of the GDP are expected to be "dozens of percent" and the uncertainty is large, too. Lots of Coronazi liars in the media were writing that Sweden would benefit neither medically nor financially from avoiding the Coronazi policies.

What a bunch of lies. Sweden has clearly reached "piecewise" herd immunity and the daily deaths dropped nearly to zero. The graphs of the total number of cases or deaths look like "tanh" (the kink from the level 0 to level 1 which has the "exponentially shrinking deviation from the asymptote" in both extremes!). Sweden has the best reasons to be certain that it doesn't have to be afraid of the winter or any other opportunity for a "second wave". It just can't happen.

And what about the economy? The Financial Times (free Yahoo copy) reviewed the Q2 results of Swedish companies in Q2 2020 and basically all of them hugely beat the expectations. The Swedish economy hasn't really been significantly affected by the coronavirus.

The fact that Sweden did the right thing when it continued "business-as-usual" while most other nations did the most damaging policy mistake in decades (or centuries) will be increasingly clear – both because of the medical situation and because of the economy.

No comments:

Post a Comment