As of April 29th 2020, there were over 3.13 million confirmed cases of coronavirus (COVID-19) globally, with the United States, Spain, and Italy having the highest numbers. This number is most likely an undercount of all cases because the way a country tests for coronavirus can have a huge impact on the number of cases they detect. We explore how different testing approaches can dramatically limit countries’ abilities to flatten the curve of infections.
Statistics on the number of cases are difficult to compare across countries. Countries have vastly different approaches to testing their populations. For example, Iceland has the highest ratio of number of tests to population and has routinely tested persons with no symptoms or no evident contacts with infected persons. Nearly 44 tests per 1000 Icelanders were administered in the first 30 days of their coronavirus outbreak. In contrast, the US only tests some of the people with symptoms, which has resulted in a mere 0.007 per 1000 Americans tested in the first 30 days since the US’s first case.
Countries that fail to test a large or representative portion of their population might face severe adverse effects.  As the virus circulated undetected for weeks, when the number of acute cases and deaths started to increase rapidly in March 2020 it quickly overwhelmed the capacity of local hospitals and funeral homes.The case of Bergamo in Italy is a case in point.
Decisions on testing are essential to flatten the curve of infections. Comparing the town of Vo’, Italy and the Diamond Princess cruise ship provides a natural experiment to understand the consequences of different testing decisions. Both places had a population of between 3,000 and 4,000 people and implemented similar isolation policies during the epidemic. However, the town of Vo’ tested almost all of its residents following the first detected case while the Diamond Princess ship initially did what most countries do: test mainly suspicious cases. As a result, during the first round of testing, 87 people tested positive in Vo’ (an infection rate of 2.6%), whereas only 10 passengers tested positive on the Diamond Princess (an infection rate of 0.3%).
By the final round of tests (two weeks since the first test), Vo’ had 6 people infected (0.2% of the population) and 2 deaths. In contrast, the Diamond Princess had 712 passengers infected (19.2% of the population) and 7 deaths. By testing only suspicious cases during the very first round of tests, the Diamond Princess had missed many infected people who then contributed to spread the virus in the subsequent period. The main difference between the two cases is the number of tests during the first screening, not the policy applied to the infected group. In fact, the Diamond Princess was relatively quick in quarantining all passengers whereas the town of Vo’ initially quarantined only those found to be infected. In other words, the town of Vo’ also reduced the cost of quarantining people.
There is a strong relationship between the number of tests performed and the number of cases identified across countries (see diagram below). There is also evidence that the statistics emerging on Susceptible, Infected, Recovered, and Dead people (SIRD) from tests are not representative of the underlying populations (http://lapogianni.blogspot.com/). This justifies the repeated calls by policy makers for “testing, testing, testing”. It also emphasizes the importance of conducting mass testing immediately after the first cases using samples that are representative of the population.