key: cord-0828504-i034d3cc authors: Kurita, J.; Sugawara, T.; Sugishita, Y.; Ohkusa, Y. title: Negative Excess Mortality in Pneumonia Death caused by COVID-19 in Japan date: 2021-01-22 journal: nan DOI: 10.1101/2021.01.22.21250283 sha: f78b181165118095158be39d1d41b6bbb3ff9fe3 doc_id: 828504 cord_uid: i034d3cc Background: Since the emergence of COVID-19, cases of excess mortality from all causes have been very few in Japan. Object: To evaluate COVID-19 effects precisely, we specifically examine deaths caused by pneumonia and examine excess mortality attributable to pneumonia in Japan. Method: We applied the NIID model to pneumonia deaths from 2005 up through August, 2020 for the whole of Japan. Introduction of routine pneumococcal vaccination for elderly people and revision in ICD10 were incorporated into the estimation model. Results: No excess mortality was found for 2020. However, negative excess mortality was observed as 178 in May, 314 in June, and 75 in July. Discussion and Conclusion: Significantly negative excess mortality might reflect precautions taken by people including wearing masks, washing hands with alcohol, and maintaining social distance. They reduced the infection risk not only of for COVID-19 but also of other infectious diseases causing pneumonia. Since the emergence of COVID-19, excess mortality from all causes has been low in Japan [1] . Actually, data obtained throughout Japan show 12 and 104 cases of excess mortality in August and October, 2020. These very few cases of excess mortality in Japan might be attributable to the lower number of traffic accidents from voluntary restrictions against going out. Therefore, we specifically examine deaths caused by pneumonia for precise evaluation of COVID-19 effects on Japan. The present study examines excess mortality attributable to pneumonia in Japan in 2020. Regarding pneumonia deaths, two points can be ignored among all causes of death. Therefore, those points are incorporated into the model for excess mortality related to pneumonia deaths. The procedure of estimation was almost identical to that used for an earlier study [1], except for some points. Excess mortality was defined as the difference between the actual number of deaths and an epidemiological threshold if the actual number of deaths was larger than an epidemiological threshold. The epidemiological threshold is defined All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. where D t represents pneumonia deaths in month/year t, T t denotes the linear time trend, V t is a dummy variable for routine pneumococcal vaccination, C t is a dummy variable for revision in ICD10, and M it is a dummy variable for the month. That is, V t is one after All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. October, 2014 and otherwise zero. C t is one after 2018 and otherwise zero. M it 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, μ 2 ) and ω t ~N(0,ξ 2 ); 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 (μ 2 +ξ 2 ) 0.5 . If the null hypothesis ξ /μ=0 is not rejected, then the Stochastic Frontier Estimation model is inappropriate. The study area was the entire nation of Japan. The study period for estimation was 2005 through August 2020. We adopted 5% as the level at which significance was inferred for results. Pneumonia is defined as J12-J18 in ICD10. Because the estimated coefficients of the cross term of routine vaccination initiation and time trend, the ICD10 revision dummy, and the cross term of ICD10 revision and time trend were not found to be significant, we dropped these terms from the estimation equation (1). Table 1 presents estimation results without these variables. The estimated coefficients of ICD10 revision and its cross term with the time trend were not significant. Moreover, although the estimated coefficients of routine pneumococcal vaccination for elderly were significant, its cross term with a time trend was not significant. Figure 1 presents observed deaths, the estimated baseline, and its threshold based All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted January 22, 2021. ; https://doi.org/10.1101/2021.01.22.21250283 doi: medRxiv preprint on the model including insignificant terms. Figure 2 specifically depicts the last 12 months in Japan. We found no excess mortality in 2020. In fact, we observed negative excess mortality as 178 in May, 314 in June, and 75 in July. This study applied the NIID model to pneumonia deaths to detect excess mortality attributable to COVID-19. No excess mortality was found in 2020, but 1074 negative excess mortality cases were identified in the last three months. Its volume was about 3.1% of the baseline in the corresponding period. Throughout Japan, about 4000 cases of mortality caused by COVID-19, as confirmed by PCR testing, were reported to the MHLW officially as of December [12] . Therefore, even if COVID-19 actually caused external mortality, neither the NIID model nor another statistical model would detect significant effects attributable to COVID-19. Of course, it is possible that some deaths caused by COVID-19 had been classified as pneumonia deaths. Even so, COVID-19 is associated with significantly reduced pneumonia mortality in May, June, and July, in Japan. The result was not attributable to the wide spread of pneumococcal vaccination or revision in ICD10 because some excess mortality was found in 2019. Results suggest that precautions among people, including wearing masks, washing hands with alcohol, and maintaining social distance reduced the infection risk All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Therefore, those excess mortality findings for Japan have misjudged the risk participation for COVID-19 among the general population. The effect of ICD10 revision was not significant at all. It is probably attributable to overlapping of its period with routine pneumococcal vaccination. In other words, its effect was not so great if one incorporates vaccination effects into the estimation model. Actually, before 2015, the increasing pace was decreasing gradually and almost flat in All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The present study has some limitations. First, our results reflected data through August, 2020 when the outbreak of COVID-19 was not severe. Subsequently, and especially in winter, a severer COVID-19 outbreak emerged. Negative excess mortality from pneumonia deaths might disappear. Continuous monitoring is expected to be necessary. Second, by rules governing the determination of the cause of death, some cases of death attributable to COVID-19 were not classified as pneumonia or as COVID-19 infection. The rule for cause of death requires classification by the most fundamental cause. For example, if cancer patients were infected by COVID-19, but then showed pneumonia and died, their proximate cause of death was probably cancer, and not COVID-19 or pneumonia. Therefore, some fraction of negative excess mortality by pneumonia might be classified as caused by other than pneumonia. However, even for periods when deaths of all causes in some months were fewer than the baseline, but not significantly in 2020 [1], the total deaths during the entire COVID-19 outbreak period were fewer than the usual level. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Thirdly, similarly to the point raised earlier, if a healthy person was presumed to be infected with COVID-19, then showed pneumonia, and finally died, then the cause of death should be classified as COVID-19, or as U07.1, for an identified virus or as U07.12 for an unidentified virus under ICD10. Conversely, for a case in which a healthy person tested positive for COVID-19, showed pneumonia, and died, but who was not presumed to be infected by COVID-19, the cause of death should be classified as pneumonia. Therefore, negative excess mortality related to pneumonia does not necessarily represent negative excess mortality of COVID-19 itself. We found no excess mortality in 2020, but 1074 cases of negative excess mortality during the last three months. Continued careful monitoring of excess mortality of COVID-19 is expected to be important. The present study is based on the authors' opinions: it does not reflect any stance or policy of their professionally affiliated bodies. We acknowledge Dr. Nobuhiko Okabe, Kawasaki City Institute for Public Health, Dr.Kiyosu Taniguchi, National Hospital Organization Mie National Hospital, and Dr.Nahoko Shindo, WHO for their helpful support. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The authors have no conflict of interest to declare. All information used for this study was published on the web site of MHLW [12] . Therefore, no ethical issue is presented. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. μ 2 represents the variance of the disturbance term. "Vaccination" is a dummy variable for routine pneumococcal vaccination: one after October 2014 and zero otherwise. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted January 22, 2021. ; https://doi.org/10.1101/2021.01.22.21250283 doi: medRxiv preprint 1 Figure 1 : Observations of the estimated baseline and threshold for pneumonia deaths from 2005 through August 2020 in Japan (persons) Year Note: The blue line represents observations. The orange line represents the estimated baseline. The gray line shows its threshold. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted January 22, 2021. ; https://doi.org/10.1101/2021.01.22.21250283 doi: medRxiv preprint Cost inefficiency in Washington Hospitals: A stochastic frontier approach using panel data Frontier Estimation: How useful a tool for health economics Managed care and technical efficiency Alternative methods to examine hospital efficiency: Data envelopment analysis and stochastic frontier analysis Cost efficiency of US hospitals: A stochastic frontier approach Figure 2: Observations of the estimated baseline and threshold for pneumonia deaths since Month/Year Note: The blue line represents observations. The orange line represents the estimated baseline. The gray line shows its threshold No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted