Excess Mortality probably due to COVID-19 in Tokyo, Japan in August, 2020

1. Introduction

To date, excess mortality has mainly been used to assess the social effects of influenza activity [1–6]. However, since the emergence of COVID-19, excess mortality attributable to COVID-19 has been attracting attention [7] as a measure of the total effects of the disease because it can reflect cases which have not been identified as polymerase chain reaction (PCR) positive. Especially in Japan, PCR tests administered per capita have been few. Therefore, concern has arisen about the possibility that some deaths caused by COVID-19 have not been recognized heretofore. Moreover, excess mortality related to COVID-19 might be expected to contribute to evaluation of vaccine effects. For these evaluations, the estimated excess mortality without the effects of a vaccine should be regarded as a baseline. Nevertheless, no such a trial has been undertaken to date. This study might be the first trial to measure that figure in Japan.

As of the end of July, 2020, the COVID-19 outbreak showed its highest peak on April 3. In all, about 17 thousand patients and about one thousand deaths from the outbreak have been reported in Japan. Although Japan has about one third of the population of the U.S., these figures are vastly different in scale from those of the U.S., which has reported 1.93 million cases of morbidity and 110 thousand cases of mortality (deaths) [8]. In light of the much lower number of patients in Japan, some criticism has arisen that low PCR testing rates might have led to the lower number of documented patients [9]. In this sense, one might regard the number of deaths as reflecting the actual situation in Japan, but with no testing-related bias.

Concerning deaths, the case-fatality rate (CFR) is about 5%. In fact, the CFRs in both countries are not much different. The lower PCR testing in Japan might be related to some problems. Therefore, we specifically examined excess mortality attributable to COVID-19 in Japan, irrespective of the cause of death.

In Japan, excess mortality was estimated using the National Institute of Infectious Diseases (NIID) model [10], which has been the official procedure for more than ten years. It was applied to two data sources: the national monthly deaths of all causes and the respective weekly pneumonia and influenza deaths in the 21 largest cities and their total. The latter is published regularly in Japanese during the influenza season as https://www.niid.go.jp/niid/ja/flu-m/2112-idsc/jinsoku/131-flu-jinsoku.html. Unfortunately, that publication ceased in March 2020 because it is intended for influenza. The first peak in Japan was April 3[11]: excess mortality cannot be detected until March. Instead, we applied NIID model to the all causes of death in April to October, 2020 in Japan.

2. Method

Excess mortality is defined as the difference between the actual number of deaths and an epidemiological threshold. The epidemiological threshold is defined as the upper bound of the 95% confidence interval (CI) of the baseline. The baseline is defined as the number of deaths that are likely to have occurred if an influenza outbreak had not occurred. Therefore, if the actual deaths are fewer than the epidemiological threshold, then excess mortality is not inferred.

The data used for this study were monthly deaths of all causes from 1987 through October 2020 [12]. NIID model, the Stochastic Frontier Estimation [13–19], is presented as where D_t represents all causes of death in month/year t, T_t denotes the linear time trend, and M_it is the dummy variable for a month, which is one if t is the i-th month and otherwise zero. Moreover, ν_t and ω_t are stochastic variables as ν_t ∼N(0, μ²) and ω_t∼N(0,ξ²); they are mutually independent. Although ν_t represents stochastic disturbances, ω_t denotes non-negative deaths attributable to influenza. These disturbance terms in this model are parameterized by two parameters: ξ/μ and (μ²+ξ²)^0.5. If the null hypothesis ξ/μ=0 is not rejected, then the Stochastic Frontier Estimation model is inappropriate.

Study areas were the whole of Japan and its capital, Tokyo. Study period foe estimation was from 1987 to October 2020 for the whole of Japan and up through July for Tokyo. We adopted 5% as significant level.

3. Results

Table 1 summarized the estimation results in the whole of Japan and Table 2 for Tokyo. Figure 1 presents observed deaths, the estimated baseline, and its threshold in Japan. Figure 2 specifically depicts the last year in Japan. We found 12 and 104 excess mortality in August and October, 2020

• Download figure
• Open in new tab

Figure 1:

Observations of the estimated baseline and threshold since 1987 until October 2020 in Japan

Note: The blue line represents observations. The orange line represents the estimated baseline. The gray line shows its threshold.

• Download figure
• Open in new tab

Figure 2:

Observation of the estimated baseline and threshold since October 2019 in Japan

Note: The blue line represents observations. The orange line represents the estimated baseline. The gray line shows its threshold.

Figure 3 and 4 showed the estimated result in Tokyo. We found 595 excess mortality was detected in August, which was 3.1% of baseline.

• Download figure
• Open in new tab

Figure 3:

Observations of the estimated baseline and threshold since 1987 until August 2020 in Tokyo

Note: The blue line represents observations. The orange line represents the estimated baseline. The gray line shows its threshold.

• Download figure
• Open in new tab

Figure 4:

Observation of the estimated baseline and threshold since May 2019 until August 2020 in Tokyo

Note: The blue line represents observations. The orange line represents the estimated baseline. The gray line shows its threshold.

4. Discussion

This study applied the NIID model to all causes of death to detect excess mortality attributable to COVID-19. We found 12 and 104 excess mortality in August and October, 2020.

Moreover, we found 595 excess mortality in August in Tokyo which correspondeces to approximately 3.1% of the baseline. It may not be a few excess mortality. It might be increasing after September, especially in winter, even though excess mortality in the whole of Japan were very few at lease until October.

On the other hand, about 2000 mortality cases caused by COVID-19 confirmed by PCR testing, were reported in MHLW officially as of December throughout Japan [12]. Therefore, even if COVID-19 actually caused external mortality, neither the NIID model nor another statistical model might would detect significant effects attributable to COVID-19. Of course, some deaths from COVID-19 might not have been tested and might have been excluded from these official numbers of mortality cases. However, total deaths might include deaths without diagnosis as COVID-19, but actually those associated with COVID-19. Estimation results suggest that such unrecognized deaths associated with COVID-19 are not significant, even if they exist.

Some researcher in Japan emphasized substantial excess mortality until May around six thousands due to COVID-19 [20] using Farrington algorithm [21] and EuroMOMO [22]. However, in nationwide, we found only 13 excess mortality in August In fact, they also showed that the observation were less than baseline in nationwide until May. However, they summed up the number in the prefectures where the excess mortalities were found and emphasize six thousands excess mortality in nationwide. Therefore, their too much excess mortality in Japan probably mislead risk participation for COVID-19 among general population.

In particular, in Tokyo, though they found 18 excess mortality in May, we did not found it. It probably suggested that their adopted procedure have upper biased for excess mortality comparison with NIID model, which was suggested logically [23].

Conversely, because we found that the number of total death were lower than baseline in nationwide until July and very few excess mortality in August, we might emphasize negative excess mortality which means precaution among people including wearing masks, washing hand by alcohol, and keep social distance reduce the infection risk of not only for COVID-19 as well as other infection diseases. Therefore, total number of death had decreased offsetting completely the effect of COVID-19.

Using pneumonia death data instead of total death data might be better to evaluate excess mortality caused by COVID-19. However, application rule of the International Classification of Diseases was revised on January 2017, after which pneumonia deaths decreased by approximately 25%. April 2020 was the fourth April since that of 2017. Therefore, April 2020 data must be compared to data of only three earlier years. Therefore, it might be inappropriate to measure excess mortality attributable to COVID-19 using pneumonia deaths only.

5. Conclusion

We found substantial excess mortality since the outbreak of COVID-19 had emerged in Tokyo in August. It might be increasing after September, especially in winter. It should be important to continue to monitor excess mortality of COVID-19 carefully hereafter.

The present study is based on the authors’ opinions: it does not reflect any stance or policy of their professionally affiliated bodies.

7. Conflict of interest

The authors have no conflict of interest to declare.

8. Ethical considerations

All information used for this study was published on the web site of MHLW [12]. Therefore, no ethical issue is presented.