key: cord-0942006-86ovj7xg authors: Kurita, Junko; Sugawara, Tamie; Ohkusa, Yasushi title: Forecast of the COVID-19 outbreak, collapse of medical facilities, and lockdown effects in Tokyo, Japan date: 2020-04-06 journal: nan DOI: 10.1101/2020.04.02.20051490 sha: d156565e44bc5d6868c889b8199f52bacfc70514 doc_id: 942006 cord_uid: 86ovj7xg Background: The number of patients of COVID-19 in Tokyo has been increasing gradually through the end of March, 2020. Object: Support for policymaking requires forecasting of the entire course and outcome of the outbreak including the date of collapse of medical facilities if a lockdown is not initiated. Moreover, the effects of a lockdown must be considered when choosing to initiate one. Method: Data of Tokyo patients with symptoms during January 14 − March 28, 2020 were used to formulate a susceptible−infected−recovered (SIR) model using three age classes and to estimate the basic reproduction number (R0). Based on the estimated R0, We inferred outbreak outcomes including the date of collapse of medical facilities if a lockdown were not enacted. Then we estimate the lockdown effects. Results: Results suggest R0 as 2.86, with a 95% confidence interval of [2.73, 2.97]. Collapse of medical facilities can be expected to occur on April 26 if no lockdown occurs. The total number of deaths can be expected to be half a million people. If a lockdown were enacted from April 6, and if more than 60% of trips outside the home were restricted voluntarily, then a collapse of medical facilities could be avoided. Discussion and Conclusion: The estimated R0 was similar to that found from other studies conducted in China and Japan. Results demonstrate that a lockdown with reasonable cooperation of residents can avoid a collapse of medical facilities and save 0.25 million mortality cases. The initial case of COVID-19 in Japan was that of a patient who showed symptoms when returning from Wuhan, China on January 3, 2020. As of March 28, 2020, 1150 cases had been announced as infected in the community of Japan, excluding asymptomatic cases, those for which the onset date or age was not reported, those infected abroad, and those infected on a large cruise ship: the Diamond Princess [1] . In metropolitan Tokyo, which has about 13 million residents, 234 symptomatic cases were identified as of March 28, 2020. The entire course of the outbreak must be predicted to evaluate the necessary medical resources for policymaking. Moreover, one must evaluate, as a worst case scenario, the collapse of medical facilities which can occur when medical needs far exceed the capacity of medical resources. Under such circumstances, the case-fatality ratio (CFR) rises considerably. Especially, the capacity of intensive-care-unit (ICU) facilities is usually not so large. They are expected to be allocated quickly to patients. Therefore, escalation of case mortality represents an important concern. To forecast these phenomena, we construct a simple susceptible-infected-recovered (SIR) model for Tokyo incorporating the necessary medical resources. Then we predict whether a collapse of medical facilities occurs, in addition to the magnitude of mortality if a collapse were to occur. We applied a simple SIR model [2, 3, 4] with three age classes: children 19 years old or younger, adults 20-59 years old, and elderly people 60 years old or older. We assumed some protection of children [5] : 40% of children were protected t infection [4] . The incubation period was assumed to be equal for all people of the three age classes and following the empirical distribution inferred for the outbreak in Japan. All rights reserved. No reuse allowed without permission. author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.04.02.20051490 doi: medRxiv preprint Experiences of Japanese people living in Wuhan until the outbreak provide information related to mild cases because complete laboratory surveillance was conducted for them. During January 29 -February 17, 2020, 829 Japanese people returned to Japan from Wuhan. Each had undergone a test to detect COVID-19; of them, 14 were found to be positive for COVID-19 [6] . Of those 14, 10 Japanese people had exhibited mild symptoms; the other 4 showed no symptom. Moreover, two Japanese residents of Wuhan exhibited severe symptoms: one was confirmed as having contracted COVID-19. The other died, although no fatal case was confirmed as COVID-19 by testing. In addition, two Japanese residents of Wuhan with mild symptoms were refused re-entry to Japan even though they had not been confirmed as infected. If one assumes that the Japanese fatal case in Wuhan and that the two rejected re-entrants were infected with COVID-19, then 2 severe cases, 12 mild cases, and 4 asymptomatic cases were found to exist among these Japanese residents of Wuhan. We therefore apply these proportions of asymptomatic cases to symptomatic cases in the simulation. Assuming that the degrees of infectivity among the severe patients and mild patients are equal also for asymptomatic cases, half of the symptomatic cases can be assumed. This assumption about relative infectiousness among asymptomatic cases compared with symptomatic cases was used also in simulation studies for influenza [7] [8] [9] [10] [11] . We sought to ascertain R 0 to fit the number of patients during 14 January -28 March and to minimize the sum of squared residuals among the reported numbers and the fitted values. Its 95% confidence interval (CI) was calculated using 10,000 iterations of bootstrapping for the empirical distribution of epidemic curves. Contact patterns among children, adults, and elderly people were estimated in an earlier study [12] . We identified the following contact patterns: the share of children contacting with other children accounted for 15/19 of all contacts, contact with adults accounted for 3/19, and contact with elderly people accounted for 1/19; the share of adults contacting children accounted for 3/9 of all All rights reserved. No reuse allowed without permission. author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.04.02.20051490 doi: medRxiv preprint contacts; those contacting with other adults were 5/9, and those contacting with elderly people were 1/9; elderly people contacting with children were 1/7, with adults were 2/7, and with other elderly people were 4/7. We assumed that contact frequencies in the same age class decreased in the same proportion in all age classes if the Tokyo Metropolitan Government (TMG) were to declare a lockdown in Tokyo. However, contact frequencies among other age classes were assumed not to be changed by a lockdown because most of the contact with other age classes can be presumed to occur at home. Contacts at home will probably be unaffected by a lockdown. Experience in Japan has revealed the pneumonia incidence in elderly COVID-19 patients as 30.6%. That among adults is 22.2%. We used these ratios to assume ratios of severe cases to symptomatic cases. Among children, no pneumonia case has been reported. Only pneumonia cases were received hospital treatment. The length of hospitalization was assumed as 30 days. Of those severe cases, 30% were assumed to require the use of an intensive care unit (ICU) and respirator for 20 days. The case fatality rate among ICU patients was assumed to be 50% from experience in Tokyo up through March 25, 2020. However, if patients requiring care at an ICU or respirator cannot receive it, then we assumed that the CFR among them was 100%. We define the collapse of medical facilities as circumstances under which the necessary ICU bed number becomes greater than 70% of all existing ICU beds. In Tokyo, there are 1000 ICU beds. Therefore, if the necessary number of ICU beds becomes greater than 700, medical service at the facilities will collapse. The CFR among patients who need care at an ICU but do not receive it is 100%. We used data of the COVID-19 community outbreak of patients in Japan who showed any symptom during January 14 -March 28, 2020 in Tokyo. We excluded some patients who had been infected abroad and who returned from abroad and those who were presumed to be infected persons All rights reserved. No reuse allowed without permission. author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.04.02.20051490 doi: medRxiv preprint from the Diamond Princess. They were presumed not to be community-acquired in Japan. Published information about COVID-19 patients with symptoms from the Ministry of Labour, Health and Welfare (MLHW) Japan or TMG was usually affected adversely by some delay because of uncertainty during onset to visiting a doctor or in the timing of a physician's suspicion of COVID-19. Therefore, published data of patients must be adjusted at least a few days. To adjust the data, we applied the following regression analysis. We set Xt-k|t as the number of patients for whom the onset date was t-k published on day t. The dependent variables are the degree of reporting delay, where k>m for several m and k. Here, m denotes the difference of the publishing dates between the two published. Date t represents the publishing date of the latest publishing. The explanatory variables were 1/k, 1/m, and 1/km. The degree of reporting delay was estimated as [estimated coefficient of constant term] + [estimated coefficient of 1/k]/k, when m was sufficiently large and time had passed. Therefore, this estimated degree of reporting delay multiplied by the latest published data is expected to be a prediction of the number of patients for whom the onset date was t-k. We used this adjusted number of patients in the latest few days, including those after VEC was adopted. We used published data of 2, 5, 6 and 9-17 March, 2020 provided by MLHW [1] . First, we estimated R 0 . Then we predicted the peak date, total number of symptomatic cases and mortality cases, maximum number of newly infected symptomatic cases per day, beds, and ICU beds. We also predicted the date of collapse of medical facilities. Moreover, we predicted the effects of the lockdown from April 6 and measured whether the medical system would collapse. Then for cases of collapse, we inferred the date of collapse. We calculated the 95% CI through 10,000 bootstrapped distributions. All rights reserved. No reuse allowed without permission. author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.04.02.20051490 doi: medRxiv preprint All information used for this study has been collected under the Law of Infection Control, Japan. There is therefore no ethical issue related to this study. During January 14 -March 28 in Tokyo, 4 cases among children, 145 cases among adults, and 85 cases among elderly people were identified as community-acquired COVID-19 for whom the onset date was published. Figure 1 depicts the empirical distribution of incubation period among 62 cases for which the exposed date and onset date were published by MHLW. Its mode and median were six days; the average was 6.74 days. Table 3 according All rights reserved. No reuse allowed without permission. author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.04.02.20051490 doi: medRxiv preprint to the proportion of voluntary restriction of people leaving their home. A collapse of medical facilities could be prevented if more than 60% of trips outside of the home were restricted. The collapse of medical facilities could e postponed by three months if more than 50% but fewer than 60% of trips outside of the home going were restricted. We applied a simple SIR model with three age classes including asymptomatic cases and assuming some proportion of children as protected. An earlier study [13] [14] [15] estimated R 0 for COVID-19 as 2.24-3.58 in Wuhan. Our obtained R 0 of 2.86 was very similar. However, one study revealed that R 0 in Japan up through 26 February was just 0.6 [16] . Their evidence given was the number of secondarily infected people. Of 110 primary cases, they found that 83 cases did not infect anyone. Nevertheless, their findings must be considered carefully. They specifically examined only those patients who had been infected by the known confirmed cases. However, half of the patients had no link to any known confirmed case. They were probably counted as primary cases but not as secondary cases. They were infected by someone in the community. In fact, until 26 February, there were 302 infected persons in the community. Of those, 161 cases were unlinked. Therefore, those unlinked 161 cases were actually secondary cases that had not appeared as secondary cases in the figures of an earlier study [14] . All rights reserved. No reuse allowed without permission. author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.04.02.20051490 doi: medRxiv preprint One might assume that the primary cases which proceeded to infect the unlinked secondary cases were the same as the 110 primary cases in an earlier study. If so, then R 0 should be 2.1 (=0.6+161/110). Alternatively, one can infer that the primary cases of the unlinked patients were not identified by a public health center (PHC) and that their distribution of the secondary cases was the same as the figure in an early study. That would mean that 28 primary cases infected 66 secondary cases. Therefore, the average number of secondary cases conditional on those which infected more than one person was 2.44 (=66/27). Presumably, the 161 unlinked cases were infected by 66 hypothetical primary cases. Moreover, these hypothetical 66 cases represented infection by 27 other cases, and so on. Therefore, 112 cases infected 273 cases which were not identified by a PHC. In total, the average number of secondary cases is expected to be 1.5 (=(66+273)/(110+112)). Therefore, the true number of R 0 is expected to be in the range of 1.5-2.1. These numbers are slightly smaller than other estimates, but far greater than one. Moreover, on April 26, medical services can be expected to collapse. Eventually, the cases ending in mortality will be approximately 0.5 million in Tokyo if no lockout is conducted. The number of deaths will be about five times greater than the average number of deaths during an equivalent period. Results also demonstrate that if a lockdown is initiated on April 6 and if more than 60% of trips outside the home were restricted voluntarily, then collapse of medical facilities might be avoided. It is noteworthy that no law exists to enforce curfews in Japan. Therefore, a lockout would ask, not force, residents to avoid leaving their home voluntarily. Consequently, cooperation with a lockout must achieve voluntary cooperation to a great degree. Evidence related to compliance is All rights reserved. No reuse allowed without permission. author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.04.02.20051490 doi: medRxiv preprint scarce because no lockdown has been conducted in Japan to date. An exceptional study asked people about restricting movement outside the home if the government asked them to do it [17] . Results showed that 93.3% would comply voluntarily with such a government request. Collapse of medical facilities from COVID-19 might be avoided in Japan, even with no enforcement of a lockdown, because strong compliance can be expected. Our unpublished research suggests that school closure since March 2 decreased contact frequencies among children by 40%. Voluntary event cancellation since February 27 decreased it among adults by 50%. Although a collapse of medical facilities has been postponed, it appears to be unavoidable if we apply these numbers to a lockdown effect. The SIR model is too simple to incorporate households, firms or schools. It is a completely mixed model. It therefore ignores some difference inside and outside of those groups. It can adjust contact patterns to mimic some policies including lockdowns or school closures as in the present study. A model highlighting differences inside and outside of those groups is an individual-based model (IBM), which mimics movements and contacts of individuals. It can therefore evaluate behavioral changes of individuals directly [6] [7] [8] [9] 17] . Therefore, we must use IBM for evaluation of a lockdown instead of a SIR model. No IBM exists for COVID-19 but an IBM exists for pandemic influenza. Especially, the most precise IBM, RIBM, has been developed in Japan using actual data of transportation [17, 18] . It indicated that lockdown with 60% voluntary restriction to going out can reduce prevalence by 40 percentage points for pandemic flu [18] . Therefore, it shared the same result that 60% voluntary restriction out of trips outside the home can avoid collapse of medical facilities with the present study for COVID-19. All rights reserved. No reuse allowed without permission. author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.04.02.20051490 doi: medRxiv preprint We predicted outcomes of COVID-19 and lockdown effects in Tokyo. We estimated a collapse of medical facilities in late April and about 0.5 million cases with mortality in Tokyo if a lockdown were not applied. We estimated the effects of lockdown enacted from April 6 and reasonable compliance with voluntary restrictions on trips outside the home. Such a lockdown might avoid a collapse of medical facilities. However, it is noteworthy that such a lockdown might continue until a vaccine for COVID-19 can be developed. Costs of a lockdown would be huge if it were to last for more than a half of year. Its cost-effectiveness is expected to represent a concern in this case. This study represents the authors' opinion. It does not reflect any stance of our affiliation. author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.04.02.20051490 doi: medRxiv preprint author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.04.02.20051490 doi: medRxiv preprint Note: "Collapse of medical facilities" is defined as necessary ICU beds greater than 70% of capacity. All rights reserved. No reuse allowed without permission. author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.04.02.20051490 doi: medRxiv preprint Notes: Bars represent the number of patients by incubation period among 62 cases whose exposure date and onset date were published by the Ministry of Labour, Health and Welfare, Japan. All rights reserved. No reuse allowed without permission. author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.04.02.20051490 doi: medRxiv preprint Japan Ministry of Health, Labour and Welfare. Press Releases of Domestic Situation Preliminary evaluation for as voluntary event cancellation as counter measure to COVID-19 outbreak in Japan as of Insignificant effect of counter measure for coronavirus infectious disease -19 in Japan Japan Ministry of Health, Labour and Welfare Strategies for containing an emerging influenza pandemic in Southeast Asia Containing Pandemic Influenza at the Source Mitigation strategies for pandemic influenza in the United States Strategies for mitigating an influenza pandemic Simulation Model of Pandemic influenza in the Whole of Japan Social contacts, vaccination decisions and influenza in Japan From 2019 to 2020: A Data-Driven Analysis in the Early Phase of the Outbreak The reproductive number of COVID-19 is higher compared to SARS coronavirus All rights reserved. No reuse allowed without permission author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint (which was not peer-reviewed) is the Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) and Coronavirus disease-2019 (COVID-19): The Epidemic and the Challenges Closed environments facilitate secondary transmission of coronavirus disease 2019 (COVID-19) Survey of pandemic behavior: to stay at home or not pandemic Simulation: application of the mathematical model for infectious disease. Gijutsuhyouronsha All rights reserved. No reuse allowed without permission author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint (which was not peer-reviewed) is the We acknowledge the great efforts of all staff at public health centers, medical institutions, and other facilities who are fighting the spread and destruction associated with COVID-19.