This is a component of the ad hoc covid19 data project connected to the FUFF platform (fuff.org)

The example is made with the simulation model that is linked from there.

example result only for simplified 4 cluster simulation

the risk with protecting risk groups

I chose the simplified model of 4 clusters to demonstrate that (-depending on the chosen parameters!-) the intention of protecting a risk group cluster against clusters with a considerable virus activity can go wrong badly.

And this, even mathematically. However, -disclaimer-, I want to encourage you to read on the pitfalls of applying mathematical models in the ressources under the link at the top of this page.

example case:

an example of 4 clusters

#1 has a higher activitiy rate and/or is more frequently exposed through jobs

#2 is not in physical distancing but working out of home

#3 is very careful and able to distance themselves, constrained to necessary contacts.

#4 is a protected risk group. But several of its members are locally or socially interconnected and are dependable of help through hub structured contacts with high exposure.

Especially those who are gathered together in nursery homes, hospitals etc.. So the inner spreading probability of the cluster will be higher than the one of group #3, if the virus penetrates over the protection walls somehow.

The result with the selected parameters (see the screenshot):

In average -in this model- you will have significant spillover effects, so that the cluster that you intended to protect will instead come out with a huge number of infections nevertheless.

Provided that cluster was to be protected for a reason, this effect is devastating.

Even cluster #3, who would normally have no problems with a prob of 0.8 get a reasonable amount of infections through the leakage from cluster #1 (and #2)

As long as the spread probability per person within the risk cluster is not a lot below 1, it is very difficult to protect a risk group.

(Why a lot below? See the dangers of averages example. Because it is only an average of subclusters/indivdiual persons with different probabilities.)

This may be difficult if not impossible for nurse homes, hospitals etc.

The reason is you have to consider the caring persons as a (potentially superspreading) subcluster of the protected risk cluster (and not as a separate cluster overspilling).

The overspill then is either the infection of a caring person from outside or through a person from inside that has been infected by a visitor or another accident.

Even if you are careful, mathematically it is a matter of time and the frequency of occurences is dependent on the situation in the other clusters.

Consider that risk groups that are not 'seperated' into hospitals and nurse homes, are to a good part socialised in an analogue way.

In reality you have numerous of those cluster situations, all with different sets of parameters (and more complex).

Such you will have various results, what implies the enhanced probability of at least some of those risk group protection scenarious will fail.

I deliberately chose that simplified model so the point of overspilling into a high potential cluster is easier to understand.

For further understanding I recommend to experiment with the 32 cluster variant, where you can define more clusters of different context.

experiment yourself…

http://fuff.org/data/cr0.html