Superspreading

Social classes

In the previous section, we've considered the possibility that not all people are identical - while some are highly mobile, others may not be. To account for this, the software allows to define multiple social classes which are randomly distributed across the grid. Each class can have its own mobility (and of course probability to move). Thus, we can simulate a population in which most people more or less stay at home, but some do not.

Superspreading events

To see this effect in the simulation, let's introduce a class society. Create the following config file:

simulation
num_timesteps 1000
snapshot_interval 200
filename_base spreader

disease recovery_time 7
p_transmission 0.23

population grid_size 1000
num_classes 3

class mobility 3
p_mobility 1.0
fraction 0.8

class mobility 5
p_mobility 1.0
fraction 0.2

class mobility 100
p_mobility 1.0
fraction 0.0

end

There's now three classes defined - initially 80% of the people are rather restricted in their interaction and only 20% (with a mobility of 5) have a larger radius. In the previous section we saw that with a mobility of 3 for everyone, the disease hardly propagates at all. As you can verify when you run the scenario, adding 20% of people with a mobility of 5 changes the picture - now the infection spreads much faster.

Now, let's introduce a very small fraction of highly mobile (and hence efficient) spreaders. Change the fraction of the second class to 0.19 and of the third class to 0.01 and re-run. The effect is rather pronounced and the infection spreads much faster (though not exponential - still we can observe daily decreasing growth percentages). Make the fraction even smaller - let's have 0.199 for class two and 0.001 for class three. Still there is an effect (all scenarios are summarized here):

Comparison of scenarios with different mobility classes.

Why is that the case? The reason is that we've seen that low mobility equals a geometric constraint - infections have to remain inside the expanding flame front and hence early on large parts of the grid are simply unaffected:

Spatial distribution after 200 days of the two-class scenario

The picture changes when even a small percentage of highly mobile people is introduced - these can 'jump ahead' of the front and seed new outbreaks of the infection elsewhere.

Spatial distribution after 200 days with 0.1% high-mobility people.

Once more people are highly mobile, the seeding is no longer contained in individual spots but generates a fuzzy front that can be filled by locally propagating infections fairly efficiently.

Spatial distribution after 200 days with 1% high-mobility people.

Consequences

The above illustrates quite well the importance of restricting situations in which people can mix with a potentially huge social group - such as sports events or concerts. Even if only a small fraction of the population participates in such superspreading events, this can significantly accelerate the spread by seeding new infection clusters.

Continue with Prior immunity.


Back to main index     Back to science     Back to numerical epidemic

Created by Thorsten Renk 2017-2020 - see the disclaimer, privacy statement and contact information.