Estimation of true number of COVID-19 infected people in Japan using LINE questionnaire

Shin-ichiro Tanaka; Shinya Oku

COVID-19 has been evolving worldwide in the first half of 2020 and also in Japan, the confirmed number of infected people was reported as 5,438 on the website of Ministry of Health 1 , Labour and Welfare (MHLW) as of 11 a. m. of April 12, 2020. This number, confirmed by Polymerase Chain Reaction (PCR) tests, only shows the subpopulation of the infected and its true number is presumably more than this. While the estimation of true number is very important in order to prevent from socalled collapsed healthcare system or overwhelming hospitals as well as in order to acquire herd immunity in the future, it is challenging to estimate the true number with the number of tested positive. Its difficulty comes from 1) the extremely limited number of PCR tests with various reasons and 2) as PCR test is performed when the subject is suspicious of having COVID-19 infection with certain medical situation such as high fever or continuous general fatigue, etc., sampling of PCR test is not randomized. That being understood, if the PCR test subjects are randomized and the number of PCR tests is sufficient, by calculating the product of national population and (confirmed number of infected)/(number of PCR tests), a satisfactorily reliable true number of infected can be obtained. However, since only the people suspicious of being infected undergo the tests actually the positive rate is much higher than that of the randomized samples. Using the current (confirmed number of infected)/(number of PCR tests) leads to estimate the true number quite excessively. For example, as of April 9, 2020, MHLW website 2 states that 4,667 out of 54,284 was tested positive. If this rate is multiplied by 126 million of entire Japan population, approximately 10 million is the assumed true number of the infected, which would not reflect the actual situation. Under the status quo as such, a large-scale randomized questionnaire (1st) by MHLW/LINE corporation was conducted from March 31 to April 1, and the result was made open 3 on April 10. The authors analyzed these data statistically and estimated the true number of COVID-19 infected patients.