The highly publicized errors in the recording of the results of COVID-19 testing are, in my opinion, only common glitches that can be corrected as one goes along. The critical issue regarding testing is not how massive they need to be, but how they can be used for guidance on policies that affect the entire population, and not only those that need medical treatment or isolation/quarantine.
Random testing for COVID-19 in Austria. Just today (5/15/20), I was pointed to: https://www/sora.at/fileadmin/downloads/projekte/Austria_Spread_of_SARS-CoV-2_Study_Report.pdf, which says that it is “the first study in continental Europe based upon nationwide PCR testing in a representative national sample.” This SORA study, dated 4/30/20, uses precisely what I have advocated as the most scientific way to estimate the full (i.e. including the hidden) COVID-19 infection rate, and thereby analyze the full impact of the pandemic on the population (“Random testing versus clinical testing,” 5/2/20).
Austria now has a national total of 16,058 cases and 626 deaths from COVID-19, which means 1,783 cases and 70 deaths per million Austrians; the Philippines has a national total of 11,876 cases and 790 deaths, or 108 cases and 7 deaths per million Filipinos (worldometers.info, 5/15/20). The Austrian case rate is 17 times, and its death rate is 10 times, that of the Philippines.
The SORA study is based on a statistically representative national sample of only 1,544 respondents, PCR-tested by the Austrian Red Cross on April 1-6, and then interviewed about their contacts, mobility, etc. by telephone on April 6-10. A scientific survey is actually very systematic: stratified into large cities, medium towns, and small towns; chosen at random from telephone directories, with only a few per sample point; prior permission obtained by telephone; households visited personally or else asked to come to a testing center; the entire family was tested, including children if permitted.
The SORA study’s national sample infection rate was 0.33 percent (3,300 cases per million), implying a national estimate of 28,500 non-hospital or “dark” cases. Allowing for statistical error, SORA puts the range of “dark” infection at 0.12-0.77 percent, and the number of non-hospital cases at 10,200-67,400. Along with the telephone survey data, SORA analyzes the effect of the Austrian lockdown on the true infection.
Random testing is on the way in Indonesia. Meanwhile, the Indonesian government is about to start its own random testing, using samples of 1,000 per province (https://www.thejakartapost.com/news/2020/05/13/covid-19-government-to-conduct-pcr-pool-tests-in-8-provinces.html).
Indonesia has a present national total of 16,006 cases and 1,043 deaths; this means 59 cases and 4 deaths per million Indonesians, or much lower than the rates among Filipinos. It will need 8,000 tests to do an Austrian-type study seeking data for eight specific provinces, each of which is the size of a small country.
The Philippines will need only 4,000 tests for a comparable study, applicable to the nation as a whole and to the National Capital Region, the rest of Luzon, Visayas, and Mindanao separately. I think this is quite affordable.
For every million persons of its population, Indonesia has done 635 tests, the Philippines has done 1,729 tests, and Austria has done 38,262 tests. Yet all those tests produce data only pertinent to the specific persons tested, and not to the rest of the people of the country. Whatever the size of the country, the accuracy of estimation of the true infection depends on the absolute number of tests—assuming they are applied to a statistically random sample—and not on the proportion of the tests to the population.
Contact mahar.mangahas@sws.org.ph.