A simple model of covid-19 contagion

In the first set of simulations, eight thousand nodes were connected at random in a single network. Density was fixed to give a degree of about 10.  Each time period, each node "interacted" with all its direct contacts. The incubation period was 14 days after which symptoms should have manifest, leading quarantine and preventing further infection by that node. After another 7 days, the node was flagged as recovered (no long counted as a case) and deemed to be immune.

In the figure below the proportion of sick individuals was plotted for 364 time periods for six different transmission probabilities ranging from 1% to 2% . Halving the transmission probability reduced the peak number of concurrently infected people from over 25% at day 75 to about 7.5% at day 147, a 70% reduction.

The second set of simulations held the transmission rate constant 1.4% and tested different network density from 0.1% ( degree ~ 10 ), to 0.2% ( degree ~ 20 ).

What appears rather striking in comparing these two figures is how much more effective reducing transmission probability is that reducing network degree.

It may be that the social distancing strategy, reducing contact with network ties and in effect making the network significantly more sparse, is less effective than finding ways to reduce the probability of transmission at each interaction.