Genetic_Programming_Theory_and_Practice_XIII

256 S. Stijven et al.

a

1.2 1.4 1.6 1.8 2.0 2.2 2.4

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

R 0

Attack Rate

0 10002000300040005000

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Infected people seeded

Attack Rate

b

1.2 1.4 1.6 1.8 2.0 2.2 2.4

0

10

20

30

40

50

60

70

1.2 1.4 1.6 1.8 2.0 2.2 2.4

0

10

20

30

40

50

60

70

R 0

Day Epidemic Peak

0 10002000300040005000

0

10

20

30

40

50

60

70

Infected people seeded

Day Epidemic Peak

Fig. 10Response prediction plots of the high-quality surrogate models obtained with SR.
[Adapted from Willem et al. ( 2014 )] Response prediction plot for the cumulative clinical attack
rate (a) and the day on which the epidemic reaches its peak (b) when seeding occurs only once
(black) or on a daily basis (gray). Predictions for R 0 assume a fixed number of infected seeds,
indicated by the dot in the panel on the right, and vice versa

or very late peaks (Fig. 10 b). There is no consensus in the literature on pandemic
influenza models about how and to which extent infectious individuals should be
seeded. Some studies (Ferguson et al. 2005 ; Chao et al. 2010 ; Halloran et al. 2008 ;
Andradóttir et al. 2011 ) have been published with static seeding of 1, 10 and 100
individuals while others used dynamic seeding. There seems to be no concern
about the potential impact of these different seeding approaches, as only a shift
of the epidemic curve due to seeding has been reported (Germann et al. 2006 ). We
explored a wide range of seeding values using both static and dynamic approaches,
and observed that the seeding approach has impact on the results. The surrogate
model divergence for small seeding values was very large so these conditions needed
to be sampled more intensively.