A photo shows a mascot character of confectionery store Fujiya Co.,Ltd., wearing a face mask at a branch shop in Matsubara City, Osaka. The Japanese government a week ago expanded a national emergency declaration's target area to the entire country. The period will be until May 6th. Japan's Prime Minister Shinzo Abe asked people to refrain from going out to the extent possible. Photo: AFP / The Yomiuri Shimbun

What is the real number of people infected with Covid-19 in Japan? Or of any country, for that matter?

“Testing, testing, testing!” most people say. “Without more testing, we don’t believe the numbers.”

As of Apr 21 Japan had conducted 124,550 tests or 1 test per 1,000 people.  How does that compare with peer countries? 

Comparing testing

Iceland leads the world with 448 tests per 1,000 people followed by Luxembourg with 56 per 1,000.  Statistically, however, they may be anomalies with relatively low populations and rather pronounced Covid-19 prevalence. 

Among other high testers are Italy, Norway, Switzerland and Germany. They have all performed more than 20 tests per thousand citizens.

Among countries facing the Pacific Ocean, New Zealand, Australia, Canada, the United States and South Korea have all done more than 10 tests per 1,000.

Among the low testers in Europe are Spain, the United Kingdom, Sweden and France, all with fewer than eight tests per 1,000.

The US now claims over four million tests done, led by the state of Rhode Island with 35 tests per 1,000 state residents, New York (32), Louisiana (31) and Massachusetts (25). The lowest in the US are Texas (6.6) and Kansas (6.4).

Tests, positives and positivity rates by country

Various sources dated April 16-21, 2020

Two Outliers

And then there are Taiwan and Japan. 

The vice president of Taiwan, who is also an epidemiologist by profession, says Covid-19 prevalence is still less than one per 100,000, argues that tests will only increase false negatives and believes resources can be better spend screening inbounds and tracing those who had been infected. 

This was also the strategy for Japan, until something happened. The positivity rate began to rise.

The positivity rate is the number of positive test results divided by the total number of tests conducted. There is much controversy among epidemiologists about the data’s utility, but here is how it goes.

Positivity Rate

Imagine you have red marbles and green marbles in a darkened jar.  You grab a fistful and there are two reds and eight greens. You grab a bigger fistful and get four reds and seventeen greens. If the ratio remains stable over a number of tries, it may be fair to conclude that around 20% of the marbles are red.

What would it mean if over time that percentage began to rise?

In France, the positivity rate was 69% while the testing was a still a low 3.5 tests per 1,000 people. Spain’s positivity rate was 56% on 7.6 tests per 1,000 people. These high results probably reflect the focusing of tests on known cluster zones.

In the US state of New Jersey the positivity rate was 50% on 20 tests per 1,000. The state of New York had 39% positivity on 33 tests per 1,000.

All these results suggest that the actual Covid-19 prevalences are probably higher than the confirmed numbers of cases reported.

Positivity rate tends to be volatile during the early phases of a pandemic because of occasional spikes while the cases remain scarce. The data become more readable from around the third month, which in this case was March 2020.

Throughout most of March, Japan’s positivity rate remained low and stable between 5% to 6%.

That compares with Norway’s 5.6% on 24 tests per 1,000 and Canada’s 6.5% on 15 tests per 1,000. Neither of these countries is known as a hot spot.

Source: Ministry of Health, Labor and Welfare

But what does it mean when the positivity rate begins to rise as the number of tests rises? That is exactly what happened to Japan in April.

It probably meant this: Within the three-pronged strategy of testing, tracing and isolating, Japan probably needed to do more testing and do a better job of isolation.

Months behind Germany

In response, Prime Minister Shinzo Abe declared a partial shutdown on April 7 and called for more testing. Prefectural governments began to hire entire hotels for isolating asymptomatic and moderate cases. They initiated drive-thru testing.

Abe has pledged testing capacity of 20,000 a day – or four times the recent run rate of 5,000. Is that enough?

Assuming the 20,000 a day is achieved by the middle of May, Japan’s number of tests per 1,000 people will reach 10.5 by July.

That’s where South Korea is today and its new Korean case count is now fewer than 10 a day. 

By September Japan’s tests per 1,000 should reach 21 per 1,000, which is where Germany is today – and that country is getting ready to partially open up its economy.

Wrong Question

South Korea and Germany are both considered to be in pretty good shape – with quite different test counts per thousand. (Never mind Iceland, with its 448 per thousand.)

What the numbers suggest is that the test count per 1,000 necessary for taking control of Covid-19 may vary rather widely.

The numbers also suggest that asking for the “actual number of Covid-19 cases” may be the wrong question. It is a fast moving target. The figure recently doubled every nine days in Japan. The recent “speed-of-doubling” in the US was five days. At the slow end is Taiwan with 17 days.

How can we account for those differences? We don’t really know.

And we need to think about that, because the key question is when this speed-of-doubling will begin to decelerate to the point that doubling no longer takes place. And why.

For that we need to know more about the virus – and our own behavior.