Estimation of the true infection rate and infection fatality rate of COVID-19 in each country

Coronavirus disease 2019 (COVID-19) pandemic is now a worldwide peril. Recently, a few local surveys revealed surprisingly high infection rates in general population.^1–4 However, the True Infection Rate (TIR) in the whole population of each country is unknown. We devised a simple method to infer TIR, as well as the Infection Fatality Rate (IFR) based on the open data.

We picked up the data at a website 4 times from April 12th to May 24th, 2020.⁵ Countries or regions having more than 1000 cases on April 12th were included. Counties fulfilling this condition have increased in later surveys, although we did not add such countries because we aimed to investigate the trend for the same population. Positive rate of PCR tests (Infection Rate; IR) and the rate of PCR tests among the population (Examination Rate; ER) were calculated. The cardinal assumption is that IR and ER are negatively correlated because PCR examinations will be restricted to cases with strong suspicion while ER is low, and vice versa. Then TIR of a specific region can be estimated using the linear regression. The ratio of the infected persons in the population to the already identified cases was calculated and named as True/Identified Case Ratio (TICR). IFR was estimated using the number of total deaths at the same website.⁵ For each parameter, 95% confidence interval (CI) was calculated. All statistical calculation was done using Microsoft Excel for Macintosh.

Included were 66 countries or regions. ER and IR on logarithmic scales showed a significant negative correlation at all adopted dates. The correlation coefficients were −0.350, −0.308, −0.332, −0.338 on April 12th, April 26th, May 16th, and May 24th, respectively. The slope value was −0.211, −0.203, −0.248, and −0.268, respectively. The estimated TIR, TICR and IFR on May 24th of representative countries are presented, together with the plot of IR vs. ER on May 24th (Figure 1). The data of Tokyo prefecture in Japan are also added to the table. The data for the whole countries or regions investigated are presented as a supplementary table. TIR took on values between 1% to 10% for most countries. TICR was around 7 to 20 in major western countries, and around 100 in some countries, including Japan. IFR also took on wide range of values, generally higher in western countries than in eastern countries.

• Download figure
• Open in new tab

Figure 1. The correlation of IR vs. ER and the estimated parameters

Left: Two-dimensional plot of IR vs. ER on May 24th, 2020. There was a significant negative correlation (r = −0.338, p = 0.006). A dotted line indicates the linear regression line.

Right: Known and estimated parameters of 15 representative countries and Tokyo prefecture in Japan. The parameters for countries are adopted and calculated from the data on May 24th.⁵ Those for Japanese prefectures are acquired from Toyo Keizai Online. Coronavirus Disease (COVID-19) Situation Report in Japan. https://toyokeizai.net/sp/visual/tko/covid19/en.html [Accessed 26 May 2020].

Abbreviations: IR, infection ratio; ER, examination ratio; IFR, infection fatality rate; TIR, true infection ratio; CI confidence interval; TICR, true/identified case ratio.

A number of factors obscure the correlation between IR and ER. These include the variation of actual situations of conducting PCR-tests among different countries. However, these varieties will be averaged over many countries included in this study. The observed correlation is highly significant, indicating that the inverse relation of IR and ER is sufficiently robust. If countries with high TIR actively expand the PCR tests due to a sense of crisis, the negative correlation would reduce and the slope of the regression line would become shallower as the time lapses. However, such a trend was not observed, and therefore we used the newest result of the regression analysis for the estimation.

Comparing the present results with local surveys is interesting. Keio University in Tokyo found 6% IR using PCR tests for in-patients unrelated to COVID-19.¹ Our estimate of TIR in Tokyo is 5.7%. Reported IRs in USA using antibody tests were 1.5% in Santa Clara², 4.6% in Los Angeles county,³ and 13.9% in New York state,⁴ and our estimate of TIR in the whole USA is 5.0%. These results support the validity of our estimate. We of course understand that the present calculation remains a rough estimate. However, the reliability of the antibody test may not be always guaranteed,^2,6 and therefore our methods must have a certain role.

The trajectory analysis⁷ reveals that the pandemic is still expanding in some countries, or temporarily controlled in others, most probably due to social distancing.⁸ Only limited countries, including China, succeeded in almost completely containing the pandemic (present estimation is not applicable to these countries). If several to hundred times of identified cases are already infected, possibly often asymptomatic, reconsideration of the strategy in the management of COVID-19 may be necessary. Finding out all infected persons and completely containing the infection would be unrealistic in such a situation. Other than complete containment, effective treatment/vaccines, and herd immunity, attenuation of the viral virulence is another way to make the pandemic less damaging.⁹ This is achieved in animals by natural selection but in humans the prevention of in-hospital infection, especially from severe cases, would be the key.¹⁰ This might be a choice, especially in countries with already low estimated IFR, mainly Asian countries such as Singapore, India, Indonesia or Japan.