Which countries are doing the most effective job of testing for COVID-19? And does all this focus on testing really matter, anyway? There seems to be considerable debate about both of those questions. Using a novel look at the data, we’ll see that while testing is ramping up worldwide, the U.S. has a long, long way to go. And we’ll see that testing really does matter.
Let’s start with a comparison of the countries, using the testing, cases, and deaths numbers from the Worldometers database through Thursday, May 14, for a group of over 50 countries. Collectively, these countries account for over 90% of the cases reported worldwide and almost 95% of the deaths. To adjust for widely differing populations, we’ll use per-capita numbers – i.e., per 1MM population – rather than the raw numbers. The first graph below presents Total Cases along the horizontal axis and Total Tests performed along the vertical axis:
Not surprisingly, the results are widely scattered. Testing practices differ dramatically from country to country. In fact, four countries – Iceland, the UAE, Luxembourg, and tiny San Marino – aren’t shown because each one performed more than 90,000 tests per MM. At the other end of the spectrum, some countries simply lack the resources and the necessary healthcare infrastructure to do more testing, or have made different testing policy choices.
Moreover, the cases vs. tests causality runs both ways. Clearly, more testing results in more cases being reported, as asymptomatic or mild cases get added to the COVID-19 Total Cases statistics. But at the same time, more cases drive the need to do more testing. A single case will provoke multiple tests: first, the test to confirm the diagnosis, then more tests to confirm that the patient has recovered – two consecutive negatives are frequently required before quarantine can be lifted – and possibly even additional tests of the patient’s housemates and other obvious candidates for testing. To model this response, we’ve added an upward-sloping red broken line to the above graph, which presumes that every reported case creates a demand for five tests. Countries whose Total Tests put them at or below that red line are doing the minimum amount of testing necessary, given their caseloads. The farther above the red line a country’s datapoint sits, the more “strategic” testing they’re doing – that could include more frequent testing of high-risk people like healthcare and food service workers, and more extensive contact tracing when new cases are discovered. We can apportion a country’s total testing between “necessary” and “strategic” testing, and quantify each.
Now let’s return to the U.S.’s performance on testing. The U.S. had done 10.6 million tests. That’s by far the most in the world, which isn’t surprising given our population of over 330 million. But on a per-capita basis, our 32,200 tests per MM rank us only 28th of the 54 countries in this analysis. And on the metric of “Strategic” Tests, calculated as Total Tests minus 5 times the number of Total Cases – i.e., Total Tests minus “Necessary” Tests – the U.S. fares even worse: we are in 36th place.
Perhaps Total Cases isn’t the right metric to consider in the first place. As with the common flu, anyone with more than mild coronavirus symptoms will consume healthcare resources and be a less productive worker. But with a disease that’s as arguably lethal as COVID-19 – which has so far killed over 300,000 souls – reducing deaths must be the primary focus.
So let’s reconsider our analysis. This time, we’ll graph Total Deaths, rather than Cases, per capita, and consider only “Strategic”, rather than Total, Tests. Here’s the resulting graph, with Total Deaths along the vertical axis presented on a logarithmic rather than a linear scale:
Again, there is quite a scatter here, but it helps to separate the countries into three distinct clusters: (1) the affluent west, which includes most of Europe plus the U.S., Canada, Australia, and New Zealand; (2) a group of other generally affluent countries; and (3) countries in the developing world.
Cluster #1. The affluent west seems to show a clear correlation between increased “strategic” testing and fewer deaths. That correlation becomes even more apparent when we consider that Italy, Spain, and Belgium were by a couple of weeks the first western countries to be truly clobbered by the pandemic, and in that sense their death rates are outliers; the rest of us have had the opportunity to learn from them. (Also note that the graph scale for Total Deaths is logarithmic, so the higher up the scale datapoints sit, the more a visual distinction between them will be understated.)
Cluster #2. This group consists primarily of relatively affluent countries in the Far East and the Middle East, and are somewhat special situations. Several of these countries have used testing intensively, but in generally in specifically targeted situations. A good example of this is South Korea, which has kept fatalities low without needing to test large segments of the population.
Cluster #3. These are mostly countries in the developing world that have not done much testing – in fact, in some the calculation of “Strategic” Tests results in negative numbers. But these countries are also earlier in the pandemic’s progression than the affluent west is. Where there seems to be some modest correlation between testing and lower death rates, these countries will be of increasing concern in the coming months.
This analysis is admittedly limited by the underlying quality of the available data, by the fact that different countries – and different states within the U.S. – have different counting rules, and by the large number of factors we haven’t even considered here. Even so, the data support two principal conclusions: (1) the U.S. needs to accelerate its testing, and (2) testing matters.