CoronaVirus: Does data justify mass testing?

As the world faces the brunt of the Chinese Coronavirus caused Covid19 disease there are calls for indiscriminate testing to be carried out by each country to detect cases. While I am not a epidemiologist or statistician, I ventured an amateur look into whether indiscriminate testing makes sense. I had carried out these workings based on data of about 10 days back but believe that it may still hold true. Readers are of course encouraged to update the data and even the methodology to share newer insights into this question.

I plotted data for 55 countries on two parameters:

a) Number of tests carried out

b) Number of cases detected

and plotted the result on a regression line below:

The number of interest in the above chart is the R2 (aka R-squared) data point. R2 in simple terms tells us what % of the data is explained by the regression equation (y=0.1623x+377.4). Basically if rise in tests had a strong relation to the number of cases being detected, the the R2 should have been very high. However it is only 0.5736 or 57.36%. In other words the number of tests carried out by different countries is explaining only 57% of the infection cases being discovered across these 55 countries. 

This was very counter intuitive to me because logically it would make sense for more tests = more cases being discovered. But this does not seem to be the case, and I was surprised to say the least. As such I decided to take a closer look at the data. The data shows that there are a handful of countries that have a lion's share of the total tests carried out but discovered very few cases.

The countries are: S Korea, Canada, Russia, UAE, and Australia which accounted for 29% of the total global tests, but only 3.2% of the total cases. So I decided to run the regression after excluding these countries, and derived the below chart:

The R2 has improved markedly to 69.82% suggesting now there is a closer relation between the number of tests and number of cases discovered. The moot question remains whether removing countries which carried out tests indiscriminately is the right thing to do at all. But even assuming that its ok to do so, the R-squared is still below 70%.

Why is R2 still not high enough? I surmise it is because a large number of tests carried out turn up negative. Given the fear about the virus, and the similarity of its symptoms to the seasonal flu means non infected individuals are getting tested. 

Basically indiscriminate testing carries threefold costs:

a) the economic cost of testing and test kits

b) the cost the deprivation as tests that will be used on healthy individuals will deprive the infected

c) the risk of contagion as mass testing will increase people's exposure

So what R2 would I have liked to see? I would say at least 80% if not 90%+. We have to evaluate R2 from the point of view that many tests are carried out on people who are suspected to have been exposed. So the data is actually not random but inherently massaged to generate a high R2, hence the standard has to be that much higher. 

The question is does a R2 of 57% or even 69% justify taking on the direct economic costs, and the indirect costs of deprivation and contagion? Unless an alternative equation can be shown (with even multiple variables if needed) that generate an R2 of over 90%, it is my opinion that indiscriminate testing is not the obvious choice. 

Note: R2 is not the same as correlation. Correlation measures the direction of movement of two data streams, it does not measure how closely related the data streams are. R-squared does that. Two tightly bound data streams will have an R2 of close to 100%, as the independent variable explains more of the dependent variable.