By Willis Eschenbach – Re-Blogged From WUWT
I’ve been saying for some time now that the number of confirmed cases is a very poor way to measure the spread of the coronavirus infection. This, I’ve said, is because the number of new cases you’ll find depends on how much testing is being done. I’ve claimed that if you double your tests, you’ll get twice the confirmed cases.
However, that position was based on logic alone. I did not have one scrap of data to support or confirm it.
Max Roser is the data display genius behind the website Our World In Data. He has recently finished his coronavirus testing dataset, covering the patchwork quilt of testing in various countries. The data is available here.
Being a ‘Murican myself, I first looked at the US daily new testing versus number of US daily new confirmed cases. I have to confess, when I saw it, I did laugh …
Figure 1. Scatterplot, daily new tests versus daily new cases, United States. Yellow/black line is linear trend.
Just as I have been saying, in the US, new cases is a function of new tests. For every one hundred additional tests that we do, we find an additional nineteen confirmed cases of coronavirus.
Of course, when I looked further there were other countries which were nowhere near as linear as the US. Here’s Australia, for example:
Figure 2. Scatterplot, daily new tests versus daily new cases, Australia. Yellow/black line is linear trend.
However, there are also plenty of countries that are just as linear as the US.
Figure 3. Scatterplot, daily new tests versus daily new cases, Turkey. Yellow/black line is linear trend.
Poland shows the same type of mostly linear relationship.
Figure 4. Scatterplot, daily new tests versus daily new cases, Turkey. Yellow/black line is linear trend.
So … how about for the whole world? Glad you asked. Here’s that chart.
Figure 5. Scatterplot, world total daily new tests versus total daily new cases. Units are thousands of tests and thousands of cases. Yellow/black line is the linear trend. Black “whiskers” show the uncertainty (one sigma) of the individual mean values for the various days.
One item of interest is the difference in the rate of discovery of new cases in various countries. In the US there are nineteen new confirmed cases per hundred new tests; Turkey is 13/100; Poland is 4/100; Australia is 1/100; and globally, there are eleven new cases for every one hundred new tests.
I suspect that this variation depends directly on at least a couple of things — the underlying number of cases in any given country, and exactly which subgroup is being tested.
For example, in the US we’re still short of tests. So the tests are being reserved for people who are showing obvious symptoms … and as a result, the US tests would be expected to come up with more new cases than the global average.
This leads to a curious situation. In addition to being a function of the number of tests, confirmed cases can also be a function of the scarcity of tests …
Conclusion? Don’t use confirmed cases as a metric of the spread of the virus—the number of cases is indeed a function of the numbers of tests.
PS—When you comment, please quote the exact words you are discussing. It avoids endless misunderstandings.
PPS—While I’m here, let me shamelessly recommend the Watts Up With That Daily Coronavirus Data Graph Page. I create the daily graphs and maintain the page. I’ll also recommend my own blog, Skating Under The Ice. I note that it’s been one full month since I publicly called at my blog for an end to the American Lockdown. Finally, I’m on Twitter here. Enjoy.