 HOME
 Reports
 Simulation of infection spread and control
 Estimated effects of human flow suppression
Estimated effects of human flow suppression
 Date
 2021.08.17
 Researcher
 Setsuya Kurahashi
 Organization
 Graduate School of Business Sciences, University of Tsukuba
/

Graduate School of Business Sciences, University of Tsukuba
Summary
1.The number of people staying in downtown areas at 19:00 has a strong influence on the infection rate
As a result of statistical estimation of the effective reproduction number by the population staying in downtown areas of Tokyo and Osaka by time of day and the population coming in from outside the areas, it was shown that the population staying in downtown areas at 19:00 has a strong relationship with the infection rate.
2.40% reduction in the number of people staying in downtown areas at 19:00 could result in 90% reduction in the number of positive cases\
Using 2021/8/13 as a baseline, it was estimated that 40% reduction in the number of people staying in downtown areas at 19:00 would result in 90% reduction in the number of new positive cases. Similarly, the number of severely ill patients is reduced by more than 90%.
3.It is important to go home early and avoid stopping in downtown areas after work hours as much as possible
In addition to reducing the number of people staying in downtown areas at 15:00, the effect of reducing the number of people staying at 19:00 was significant, suggesting that it is important to finish work on time and go home early. In addition, in the case of Osaka, it was suggested that the number of people staying in downtown areas at 21:00 was at the same level of risk.
4.Effects of diminishing vaccine efficacy are anticipated
If the effect on infection suppression and immunity effect decrease to 64% after 180 days of the second vaccination (Israel Ministry of Health), a third vaccination may be required after the end of the year, as the slight increase in infections after November continue.
https://www.gov.il/en/departments/news/0507202103
https://www.gov.il/en/departments/news/0607202104
Estimation model of the infection change rate by the population flow in Tokyo
➡︎As a result of statistical estimation of the effective reproduction number by the population staying in downtown areas of Tokyo by time of day and the population coming in from outside of Tokyo, it was shown that the population staying in downtown areas at 19:00 has a strong relationship with the infection rate.
Chart: Estimation of the infection change rate by the population flow in Tokyo
Effect of suppressing the number of people staying in Tokyo on the reduction in the number of positive patients
➡︎The number of positive patients was estimated from the SEIR model by decreasing x1 (the number of people staying in downtown areas of Tokyo at 19:00) in the estimation model of infection change rate to obtain Dt.
➡︎40% reduction in the number of people staying in downtown areas of Tokyo at 19:00 could result in 90% reduction in the number of positive cases.
1.Based on the average number of people staying during 8/7~13, continued at 100%
2.Based on the average number of people staying during 8/7~13, reduced the human flow to 80% after 8/16*
3.Based on the average number of people staying during 8/7~13, reduced the human flow to 60% after 8/16*
Estimated from the number of people staying in downtown areas of Tokyo (Shinjuku, Ginza, Shibuya, Ueno, Ikebukuro, Roppongi) at 19:00
However, the number of positive patients is aged 15 and over, and when aged 14 and under are included, it increases by about 1.1 times
* Full effect after 7 days, 7% decrease in the number of people coming in from outside of Tokyo
Red: Number of new positives (15 years old or older)
Green: Number of new positives (1539 years old)
Blue: Number of new positives (4064 years old)
Purple: Number of new positives (65 years old or older)
Solid line: Actuals / Wavy line: Estimates
* Numbers are 7day moving averages for 15 years old or older
Effect of suppressing the number of people staying in Tokyo on the reduction in the number of severely ill patients
➡︎A 40% reduction in the number of people staying in downtown areas of Tokyo at 19:00 could result in a 90% reduction in the number of severely ill patients.
➡︎The number of severely ill patients aged 65 and over is high. It is necessary to consider measures such as further increasing vaccination rates and a third dose of vaccine.
1.Based on the average during 8/713, continued at 100%
2.Based on the average during 8/713, reduced the human flow to 80% after 8/16*
3.Based on the average during 8/713, reduced the human flow to 60% after 8/16*
Estimated from the number of people staying in downtown areas of Tokyo (Shinjuku, Ginza, Shibuya, Ueno, Ikebukuro, Roppongi) at 19:00
* Full effect after 7 days
Red: Number of seriously ill inpatients (15 years old or older)
Green: Number of seriously ill inpatients (1539 years old)
Blue: Number of seriously ill inpatients (4064 years old)
Red: Number of seriously ill inpatients (65 years old or older)
Solid line: Actuals / Wavy line: Estimates
* Numbers are 7day moving averages
Insights about Tokyo
The results of the analysis, which included the number of people staying at 15:00, the effect of reducing the number of people staying at 19:00 was greater than reducing the number of people staying at 15:00, suggesting that it is important to finish work on time and go home early.
Estimation model of the infection change rate by the population flow in Osaka
➡︎In Osaka, same as Tokyo, the number of people staying in downtown areas at 19:00 has a strong influence on the infection rate.
Effect of suppressing the number of people staying in Osaka on the reduction in the number of positive patients
➡︎In Osaka, same as Tokyo, a reduction in the number of people staying in downtown areas at 19:00 could result in a reduction in the number of positive cases.
1.Based on the average during 8/7~13, continued at 100%
2.Based on the average during 8/7~13, reduced the human flow to 80% after 8/16*
3.Based on the average during 8/7~13, reduced the human flow to 60% after 8/16*
Estimated from the number of people staying in downtown areas of Osaka (Kyobashi, Juso, OsakaMinami, Shinsekai, Tennoji, Abeno) at 19:00
However, the number of positive patients is aged 15 and over, and when aged 14 and under are included, it increases by about 1.1 times
* Full effect after 7 days
Red: Number of new positives (15 years old or older)
Green: Number of new positives (1539 years old)
Blue: Number of new positives (4064 years old)
Purple: Number of new positives (65 years old or older)
Solid line: Actuals / Wavy line: Estimates
* Numbers are 7day moving averages for 15 years old or older
Insights about Osaka
The results of the analysis, which included the number of people staying at 21:00, the effect of reducing the number of people staying at 21:00 was as great as that of the number staying at 19:00, so it is also important to take measures for the number of people staying around 21:00.
This is a trend that was not seen in Tokyo, suggesting that measures may need to be taken in the late night zone in Osaka.
Model Settings
1.Infection model by SEIR mathematical model and AI optimization method
The SEIR model, which takes into account population flow and AI technology (evolutionary optimization + quasiNewton method), were used to optimize infection model estimation within and between three age groups (1539 years, 4059 years, and 65 years or older)*1. The positive patient influx from outside Tokyo was estimated from mobile spatial statistics data and LocationMind xPop*2, and incorporated into the model, and the model was trained from the data from March 1 to July 30, 2021. The number of severely ill patients was estimated from the transition of the number of positive patients in each age group by constructing a statistical model from the data from June 21 to August 1, 2021. In addition, assuming that there was a behavior change of the citizens of Tokyo that was equivalent to the cancellation of the 1st state of emergency last year, a simulation was conducted by applying the effective reproduction number since last summer and the data of the floating population in Tokyo. The Delta variant was assumed to have 1.4 times the infectivity (basic regeneration arithmetic) of the estimated Alpha variant.
2.Estimating circuit breaker strength and vaccination effectiveness
For the Alpha variant (remaining conventional variant) and Delta variant, the strength of emergency declaration mitigation was set.
3.Effects of vaccine and behavior change
・The vaccine effect was 57% for the 1st dose, 94% for the 2nd dose for the Alpha variant to prevent the infection, and 0.9 times for the Delta variant. Measured values are used for changes in the number of effective reproductions and the number of population flows from 3/1 to 8/13. After 8/14 uses the most recent 7day moving average infection change rate.
・Vaccination rate setting
After 3/5 0.05% of the population (1st measured number of medical staff)
After 3/27 0.032%, 0.033% (number of 1st and 2nd medical staff measurements)
After 4/12 0.069%, 0.030% (1st and 2nd actual measurements of medical staff) 0.01% (1st actual measurement of elderly people)
After 5/4 0.064%, 0.078% (1st and 2nd actual measurements of medical staff) 0.065%, 0.006% (1st and 2nd actual measurements of elderly people)
After 6/1 0.064%, 0.078% (1st and 2nd expected medical staff) 0.08%, 0.065% (1st and 2nd expected elderly)
After 6/21 k/2%, k/2% (1st and 2nd expected medical staff) k/2%, k/2% (1st and 2nd expected elderly) k = 1.0%
After 8/15 1.3%
・Diminishing vaccine efficacy
The infection suppression effect was assumed to diminish to 64% (Israel Ministry of Health) after 180 days of the second vaccination. Since no data was available on immunity effect, it was assumed to be equivalent.
・https://www.gov.il/en/departments/news/0507202103
・https://www.gov.il/en/departments/news/0607202104
* The number of new positive patients was about 10% lower than the total number of positive patients because the number of new positive patients aged 15 years or older (publication date) was set.
*2 "LocationMind xPop" data is data that NTT DOCOMO collectively and statistically processes from mobile phone location information sent with permission from users of applications* provided by NTT DOCOMO. Location information is GPS data (latitude and longitude information) that is measured at a minimum of every five minutes, and does not contain information that identifies an individual. * DOCOMO Map Navi Service (map application, local guide) and some other applications
Model Details
Agespecific Vaccine Effect SEIR Model
Infection Transition Probability by Age
(Propagate from right to left)
Y is 15 to 39 years old, M is 40 to 64 years old, and E is 65 years old or older.
Infections between the ages of 15 and 39 are 87% from the same age group (15 to 39 years old), 13% from 40 to 64 years old, and 0% from 65 years old or older.
The same applies to infections between the ages of 40 and 64 and those aged 65 or older.
Inverse Simulation Model