IEEE Computational Intelligence Magazine - November 2020 - 31

103

Estimated Rt

5
102

4
3

101

2
0

Feb-20
Feb-25
Mar-01
Mar-06
Mar-11
Mar-16
Mar-21
Mar-26
Mar-31
Apr-05
Apr-10
Apr-15
Apr-20
Apr-25
Apr-30
May-05
May-10
May-15
May-20
May-25
May-30

Reconstructed Daily Infection

104

3
103
Lockdown

Lockdown +
Centralized Quarantine

102

Feb-29

Feb-24

Feb-19

Feb-14

Feb-09

Feb-04

Jan-30

Jan-25

100

Jan-20

0
Jan-15

101

Jan-10

Date
FIGURE 8 Estimated evolution of transmission dynamics in Wuhan.
The black line represents the reconstructed daily infection number and
the green bar is the posterior mean of estimated R t. Two major events
(city lockdown measure from 23rd Jan and centralized quarantine
from 2nd Feb) are annotated with red arrows.

104

7
6

Wuhan, China

100

Date
FIGURE 7 Estimated evolution of transmission dynamics in UK. The
black line represents the reconstructed daily infection number and the
green bar is the posterior mean of estimated R t.

Relative % Reduction in Dt
Suppression Effect

United Kingdom

Reconstructed Daily Infection

sharp decrease from 23rd Jan. When the centralized quarantine
policy has been enforced from the beginning of February, the
R t values then largely remain below zero (the spike around
14th Feb is due to misreporting).
Figure 9 compares the reductions in < R t , p t , D t > for
different response levels between European Countries and
Wuhan. From the analysis of Wuhan data, the strong impact of
lockdown is clearly demonstrated with the immediate relative
reduction of R t by 58%. We also observed that the combination of lockdown, centralized quarantine and immediate
admission of confirmed patients starting from Feb 2nd in
Wuhan was associated with a more substantial relative reduction of R t with strong suppression and mitigation effects.

Estimated Rt

results of 14 European countries. It shows different reduction
rates of for different response levels. The relative
reduction of compared to the minimal response
(Level 0 where R t is set to R 0 ) was estimated for each response
level. With soft response (Level 1), the corresponding intervention measures (e.g. school closure, quarantine of international
arrivals from high-risk regions) are correlated with a relative
reduction of R t by 35%, showing both strong suppression effect
(D t shortening 22%) and mitigation effect ( p t reduction 29%).
With strong response (Level 2), the relative reduction of R t
increases to 60% with a strong mitigation effect ( p t reduction
56%). But the suppression effect (D t shortening 26%) is similar
to that of Level 1, indicating marginal incremental suppression
effect. This observation shows a consistency with the aim of
representative intervention measures on this level (e.g. canceling public events, restrictions on gathering and internal movements) to reduce the contact rates among the population.
The emergent response (Level 3) shows substantial relative
reduction of reproductive number (R t reduction 71%) with
suppression (D t shortening 37%) and mitigation ( p t reduction
67%) effects, correlated to the intensive measures (e.g. workplace closure and stay-at-home requirements). A similar degree
of reductions is found for Level 4 (R t reduction 74%; D t shortening 40%; p t reduction 70%) while the stringency of intervention measures is higher. We find that our estimated evolving
patterns of p t and D t correspond well to the serial strategies
taken by some European countries, such as the 'contain-delaylockdown' route taken in the UK.
In addition to the results of 14 European Countries,
Figure 8 shows the results of applying our method to the
Wuhan data, where the greens bars indicate the posterior mean
of R t during the outbreak of COVID-19. We can see that at
the early stage of the pandemic, the R t levels are above 1. After
the lockdown intervention has taken effect, R t experienced a

Soft
Response
∆R = 35%

20
30
40

Wuhan
Lockdown
∆R = 58%

50
60

Europe

Wuhan

Strong
Response
∆R = 60%
Emergent
Responses
∆R = 71%
∆R = 74%

∆R = 79%
Lockdown
+ Centralized Quarantine

20
30
40
50
60
Relative % Reduction in pt
Mitigation Effect

FIGURE 9 The relative reduction of mitigation factor p t and suppression factor D t under different response levels compared to minimal
response level.

NOVEMBER 2020 | IEEE COMPUTATIONAL INTELLIGENCE MAGAZINE

IEEE Computational Intelligence Magazine - November 2020

Table of Contents for the Digital Edition of IEEE Computational Intelligence Magazine - November 2020