COVID World Review, Part 7: COVID Testing

This is the seventh of my review reports on COVID-19 worldwide as at August 20^th 2021. This time, I’m going to look at testing for the presence of the COVID virus.

Testing, these days, is an oh-so-retro subject. Gone are the days, in the UK at least, when politicians would mislead the public by over-reporting the number of tests, or under-reporting the proportion of positives. And yet, success (or otherwise) at testing for COVID is, as we will see, quite well correlated with success (or otherwise) at minimizing deaths per million, which I looked at in the fifth essay of this set.

Europe

I’ll start, as always, in Europe. I’ll show two graphs for each group of countries. First, the cumulative numbers of tests per hundred thousand population. And second, the ratios of cumulative cases per test, as measured over the course of the epidemic.

The first graph will give an idea of how thorough each country has been in approaching its testing for COVID. The second will tell how well the country’s testing system has worked. A high percentage of cumulative cases per test probably means shortages of test kits, and thus likely a more or less serious under-statement of the number of cases. In contrast, a low percentage of cumulative cases per test will mean that tests have been available to people who suspect, but aren’t sure, that they have the disease. This will have helped to reduce transmission of the virus if there is an adequate track and trace system in place.

These countries divide into two groups. Austria, Denmark, Luxembourg and the UK lead the way, with from 3 up to 8 tests administered over the course of the epidemic per head of population. The rest are in the range of 0.7 (Netherlands) up to 1.6 (Portugal) tests per head.

These figures are more widely spread than the cumulative deaths per case in my last report. But they show a similar pattern of an initial peak, a steep fall, a rise to a second peak and a gradual decline. The Netherlands is the high outlier; Austria and Denmark the low ones.

In tests, Slovakia is the high outlier; the rest are at, or below, general Western European levels. About half the countries are over the 10% in cases per test, whereas in the core of Europe only the Netherlands was up there. The falls from the first peak are far less steep.

Here, Cyprus is the high outlier in tests per population. The patterns in cumulative cases per test are more complex than further west or north; this area seems to have experienced several moving waves of cases in succession.

In this group of mostly small countries, both numbers of tests and cumulative tests per case to date are comparable with the remainder of Western Europe.

Here is the ordered list of European countries by tests per hundred thousand:

Here is the ordered list of cumulative cases per test in Europe:

The seven countries at the bottom of the first list and the top of the second are those which have not reported cumulative testing data. They include Czechia, Sweden and France. And all the top 14 reporting countries, except the Netherlands, are in Eastern Europe.

There is a considerable degree of inverse correlation between the two graphs. Cyprus, Austria, Slovakia, and Denmark are the top four and bottom four respectively; and Luxembourg is not far behind. Hardly surprising, as the more liberal you can be with test kits, the less chance there will be of each individual test coming out positive. Though this isn’t the whole story; Norway and Finland, at least, have done relatively few tests, yet also have low cumulative cases per test.

But you can see something interesting when you compare these lists with the ordered list of deaths per million for the same countries:

Some of the same names are near the top of both cases per test and deaths per million lists: Bosnia and Herzegovina, North Macedonia, Slovenia. Even stronger is the resemblance of the bottom of the league table in cumulative cases per test to the bottom in deaths per million: Denmark, Cyprus, Iceland, Finland and Norway are all in the bottom ten in both (ignoring the Vatican). And Cyprus, Denmark and Malta are all in the top ten in tests and the bottom ten in deaths per million. In short, good testing for COVID saves lives.

To check this, I feel a scatterplot (or, better, a series of scatterplots) coming on!

Americas

Looking at the spaghetti graphs as a whole, the history of how many tests were carried out when, and when the cases per test peaked and troughed, is of less interest than the total tests per hundred thousand population and the cumulative cases per test over the course of the epidemic. So, I’ll concentrate, from now on, on the ordered lists of countries in a region.

In general, these numbers are way lower than in Europe. Only the top five have reached the mark of one test per head of population.

For cumulative cases per test, I’ll show the top 20 world-wide:

Those numbers are remarkable. More than 50% of tests in Brazil have been positive over the whole course of the epidemic! More than 40% in Peru, and almost 40% in Argentina, too. There must have been very serious shortages of test kits in these countries!

Let’s compare with the top 20 in the world in deaths per million:

Peru, Brazil, Colombia, Argentina, Paraguay and Mexico are all in the world top 20 in deaths per million. Brazil, Peru, Argentina and Mexico are four of the bottom six in the Americas in tests per hundred thousand. They are also the top four world-wide in cumulative cases per test. And Paraguay and Bolivia are seventh and ninth respectively. This suggests that “lack of testing kills.” And that failing to find and isolate the cases that are out there, kills to an even greater extent. A general shortage of test kits in South, and to a lesser extent Central, America may be a significant part of the reason why those countries have performed so badly against the virus.

Middle East and North Africa

The UAE, Bahrain and Israel are comparable with high-testing countries in Western Europe. There is a second group, from Jordan down to Lebanon, which compares with the lower-testing countries there. The rest are lower yet.

Not as bad as South America, but still not good, except for the bottom four.

Sub-Saharan Africa

These are all orders of magnitude lower than European figures; even Gabon, which has not yet quite reached 50% in tests per head of population.

These are higher than the corresponding figures in Europe, but not by nearly as much as the tests are lower. It does look as if it’s harder for the virus to spread in Africa than elsewhere. And Benin is top of the list in tests, and second from bottom in cumulative cases per test.

It’s interesting that Namibia and South Africa, while second and third in the list of tests per hundred thousand, are also high in the list of cumulative cases per test. As these are two of only four countries in mainland Africa to be generally considered functioning democracies (the others being Ghana and Botswana, the latter of which is not reporting testing data), this suggests to me that their data is probably better than other data from Africa. So, I will put these countries on my list for later follow-up.

Rest of Asia

Hong Kong, Singapore, the Maldives and Georgia are all well above the 1 test per head of population level. The countries doing least testing in this region are Laos (which has very low cases per million), Bangladesh, Myanmar and Indonesia.

Those at the top of this list are comparable with the more reliable countries in Africa, and with the Middle East. East and South-east Asian countries seem to have lower cumulative cases per test than the rest.

Australasia and Oceania

Australia and New Zealand have done quite a bit of testing; Fiji much less; PNG hardly any.

Again, a fair degree of inverse correlation with the level of testing.

Top and bottom 20

Here is the list of the top 20 countries in tests per 100,000:

The Singaporeans have done something remarkable. They are in both the top 20 in the world in tests per 100,000, and the bottom 20 in deaths per million. But Singapore is the only country to have achieved anything like this. Moreover, the idea that, in general, a good rate of testing ensures a good performance in terms of deaths per million, does not hold up. A counter-example is provided by the UK: seventh in the world in tests per 100,000, and 20^th in the world in deaths per million.

It isn’t worth showing the lowest testers, as many countries are not reporting testing data at all.

Here are the world top and bottom 20 in cumulative cases per test:

The top 20 list is a catalogue of horrors. But a lot of countries, which have done well in terms of deaths per million, are in that bottom 20.

The follow-up list

These are countries which may be of interest for follow-up because of their testing history. The ones marked ✔are already in the list.

Country	Already In?	Reason
Cyprus	✔	Highest tests per 100,000: Low cumulative cases per test
Austria		High tests per 100,000: Low cumulative cases per test
Slovakia		High tests per 100,000: Low cumulative cases per test
Luxembourg		High tests per 100,000: Low cumulative cases per test
UAE		High tests per 100,000: Low cumulative cases per test
Denmark	✔	High tests per 100,000: Lowest cumulative cases per test
Bahrain	✔	High tests per 100,000 for its region
Israel		High tests per 100,000 for its region
Tunisia		High cumulative cases per test for its region
Brazil	✔	Highest cumulative cases per test
Peru	✔	High cumulative cases per test
Argentina		High cumulative cases per test
Mexico	✔	High cumulative cases per test
Ecuador	✔	High cumulative cases per test
Hong Kong	✔	Lowest cumulative cases per test
New Zealand		Low cumulative cases per test
Australia		Low cumulative cases per test
Bhutan	✔	Low cumulative cases per test
Singapore	✔	Low cumulative cases per test
Taiwan	✔	Low cumulative cases per test
Laos	✔	Low tests per 100,000 for its region: Low cumulative cases per test
Benin		Low cumulative cases per test
Namibia		Probably good testing data from Africa
South Africa		Probably good testing data from Africa
Bangladesh		Low tests per 100,000 for its region: High cumulative tests per case for its region
Myanmar		Low tests per 100,000 for its region: High cumulative tests per case for its region
Indonesia		Low tests per 100,000 for its region: High cumulative tests per case for its region

Scatterplots

Here are the scatterplots I referred to. First, the negative relation between tests per 100,000 and cumulative cases per test in Europe does carry over into the world as a whole:

I didn’t try to show a trend, because there are two substantial sets of outliers. At the right are countries which have done 5 or more tests per head of population; and these have uniformly low cumulative cases per test. However hard you try, you can’t (or shouldn’t be able to) find cases that aren’t there! At the top left, on the other hand, are the countries which have recorded very high cumulative cases per test. These seem to have rather similar levels of tests per head, suggesting that test kits are in short supply in those countries.

Next, cases per million versus tests per 100,000:

Unsurprisingly, this tells us that the more testing you do, the more cases you are likely to find. But it’s almost as if there are two different types of countries on the same plot; the high testers and the low. Here are the plots just for Europe and just for the Americas, which have done relatively much and relatively little testing respectively:

The trend is 15 times higher in the Americas than in Europe, reflecting that the marginal value of each test is greater when there are fewer tests being done than when there are many.

Now for deaths per million against tests per 100,000:

There’s not much trend here at all, and it’s positive! More tests mean more deaths per million? But it’s the low testers in the bottom left corner that drag the line down. And the low testers have a much bigger spread of deaths per million than the high testers.

Again, I’ll show the separate plots for Europe and for the Americas:

Now, that’s a bit counter-intuitive. The Y-axis intercepts are very similar, and the downward “trend” in the Americas is about twice that in Europe. I expected a higher Y-axis intercept in the Americas, and a far bigger downward trend. It seems that Peru, with 6,000 deaths per million, is an outlier rather than a norm. The higher trend in the Americas also suggests that there may be a “law of diminishing returns” from testing, which some European countries hit, but even the USA, the most active tester in the Americas, has not.

Lastly, here are deaths per million against cumulative cases per test:

60 more deaths per million for each percent of rise in cases per test world-wide? That’s more evidence that shortages of test kits kill.

To sum up

The main conclusion is that a high ratio of cumulative cases per test, likely brought on by a shortage of test kits in a country, tends to go with a high rate of deaths per million from COVID. So, shortages of test kits kill. A general shortage of test kits in South, and to a lesser extent Central, America may be a significant part of the reason why those countries have performed so badly against the virus.

Cumulative cases per test, in turn, is roughly inversely correlated with the number of tests per head of population. But the direct effect of test numbers on deaths per million is less obvious.