 HOME
 Reports
 Simulation of infection spread and control
 Projection of the number of new positive persons assuming the prevalence of Omicron variant
Projection of the number of new positive persons assuming the prevalence of Omicron variant
 Date
 2021.12.14
 Researcher
 Akimasa Hirata
 Organization
 Center of Biomedical Physics and Information Technology
/

Center of Biomedical Physics and Information Technology
Prediction system based on AI (deep learning)
Based on deep learning (LSTM model), directly predicted the number of new positive cases/number of seriously ill patients per day (oneweek average value), etc.
The data volume in appearance can be increased by using standardized data instead of that of prefectures. Estimate based on nonlinear regression (the number of parameters is virtually unlimited)
*The values can be calculated within several months to the extent that the accuracy can be secured
*Prepared using the Twitter data provided from NTT DATA by Toyoda Lab, Institute of Industrial Science, the University of Tokyo
E. A. Rashed and A. Hirata, “Infectivity upsurge by COVID19 viral variants in Japan: evidence from a deep learning modeling.” Int. J. Environ. Res. Public Health, 2021.
(Update) Factors of the decrease in infection in the 5th wave
・There is a possibility that the decrease in the number of newly infected people appeared to be drastic due to (1) an end to the temporary increase in the effective reproduction number associated with consecutive holidays in the expansion period and (2) the infectionprevention effect of vaccination.
・Used machine learning to virtually reproduce cases where (1) vaccination did not proceed from July 11 and onward, where (2) behaviors do not change (selfrestraint was not promoted), and where (3) there is no natural infection.
・The vaccination effect, natural infection, and behavioral change reduce the effective reproduction number and contribute to ending the pandemic earlier.
Learning up to April 2022. After that, used actual data as much as possible for reproduction.
*The assumption of the case without behavioral change associated with selfrestraint is that flows of people is equivalent to the level of early July (30% compared with February 2020) and Twitter (drinking party) rose by 30% on average. If flows of people are equivalent to those of October 2020, the approximated effect is about 10%.
**The number of asymptomatic infected people is assumed to be four time larger than the number of positive cases (approximate).
(Update) 6th wave without the appearance of the omicron variant?
*Difference: Update of the prevention effect of vaccination (12/7), If Twitter “drinking party” is used
The main reason for the difference is the change in the prevention effect of vaccination
Medium model (infectionprevention effects of the first and second vaccinations: 65% and 75%, respectively),
Prevention effect of the third vaccination: 95%
The vaccination effect is assumed to keep its peak for 14 days after vaccination and then linearly attenuate
(The prevention effects of vaccination of the second and third vaccinations of half a year later are 35% and 55%, respectively (attenuation of 40%))
↓
High model (infectionprevention effects of the first and second vaccinations: 70% and 95%, respectively),
Prevention effect of the third vaccination: 95%
The vaccination effect is assumed to keep its peak for 14 days after vaccination and then linearly attenuate
(The prevention effect of vaccination attenuated by 25% in half a year (70% for both second and third vaccinations))
Spread of the omicron variant (assumption based on multiple reports)
Parameters which play important roles in analysis:
(1) Infectionprevention effect of vaccination: Limited,
(2) Effective reproduction number (infectability): That of South Africa is limited
(3) Flows of people and behaviors: Unpredictable
As of now, researchers have to set scenarios with multiple assumptions with limited knowledge
Assumption: Replacement with the omicron variant
Scenario: Community spread of the omicron variant was confirmed in midJanuary and it was assumed that infection spread twice as fast as the delta variant (to be replaced by the delta variant in midFebruary). The infectability is assumed to be 1.25 or 1.5 times stronger than that of the delta variant (transmissibility 0.9 to 1.35)*.
*Barnard et al, Modelling the potential consequences of the Omicron SARSCoV2 variant in England, https://cmmid.github.io/topics/covid19/reports/omicron_england/report_11_dec_2021.pdf
Model of the infectionprevention effect of vaccination
Model of the vaccination effect on the delta variant (Japanese version) (from the material of the report on December 9)
Based on the high model (infectionprevention effects of the 1 and 2nd vaccinations: 70% and 95%, respectively) and the report of 10/5, it is assumed that the prevention effect of the 3rd vaccination will be 95% and the prevention effect will decrease by 25% half a year later (decreases linearly up to 70%). The vaccination effect is assumed to keep its peak for 14 days after vaccination and then attenuate (Reference:
https://doi.org/10.1038/d41586021022618）
Model of the vaccination effect on the omicron variant
According to the reference, it is assumed that the prevention effects of the second and third vaccination on the delta variant are 65% and 75%, respectively. It is assumed that the decrease in the vaccination effect will reach its peak 14 days after vaccination as with the delta variant and then attenuate by 25% in half a year (linearly decreases to 40% and 50%). https://khub.net/documents/135939561/430986542/Effectiveness+of+COVID19+vaccines+against+Omicron+variant+of+concern.pdf/f423c9f491cb0274c8c570e8fad50074 ）
Scenario: If the third vaccination is conducted six to eight months after the completion of the second vaccination
Assume the timing at which to start the third vaccination is December 1, and assume the number of vaccinated people on the corresponding date before it will increase immediately after the start. (Start date of the second vaccination: Tokyo from 2021/3/10 and onward.)
Change in the number of new positive cases depending on the infectability of the omicron variant
Calculated assuming that behavior control would start from the end of February, referring to the flows of people of the previous year (black dotted line).
The implementation of countermeasures against infection is assumed. *The standard of flows of people is the median value of each day of the week for the five weeks from January 3 to February 6, 2020
*The timing of the start of spread was set arbitrarily
Change in the number of new positive cases depending on the third vaccination timing
Calculated assuming that behavior control would start from the end of February, referring to the flows of people of the previous year (black dotted line).
The implementation of countermeasures against infection are assumed. *The standard of flows of people is the median value of each day of the week for the five weeks from January 3 to February 6, 2020
*The difference depending on the vaccination interval is small.
Summary
・Supplementary report that the infectionprevention effect of vaccination decreased by 30% to 40% (This has the strongest impact on the trial calculation)
・What is the effective reproduction number of the omicron variant?
That of the delta variant was said to be twice as high at first, but actually it was 1.5 times or lower (estimated by this group; refer to 12/9). There is also a report saying that the transmissibility of the omicron variant is 0.95 to 1.35 stronger. It has to be corrected in the future.
・The trial calculation based on the above assumptions suggests a possibility that the peak may be lower than that in the 5th wave. Because a certain vaccination effect is expected.
・Acceleration of the vaccination rate (%)
It may temporarily push diffusion, though the effect will be limited (population level, not individual level)
・Data is insufficient for evaluating aggravation (in the current approximation, it is about 1/10 but may increase from now on).
Related reports
Report by Akimasa Hirata

Projection of new positives and deaths in the 8th wave ~The 8th wave in metropolitan areas depends on the mutation strain~

Convergence Factors of the 5th to 7th Waves ～In considering future measures～

Prediction of the number of new infections and deaths in the future ~Changes due to vaccination period and yearend and New Year activities~