One of the most important things that countries need to be doing to help understand and stop the spread of COVID-19 is testing. However, countries differ in their response to corona. Some countries opt for herd immunity, some implement weaker forms of lock-down than others. Yet others, such as Taiwan, are actively using extensive testing and tracing of the infected people. When comparing countries this method seems to perform very well as Taiwan only has a few a few hundred cases, while it is in very close proximity to China and there were cases of infected very early on. We see much less extensive testing in European countries, this may also be due to the limited availability of tests. Questions that come up when thinking about testing are. When is testing most relevant? Is it better to do distributed tests or is it better to do single large scale testing at a specific point in time. What would then be the best time to do the large scale testing? One could imagine that if you start too quickly without many people infected, this will give you hardly any new information. However do it too late and it is already too late to flatten the curve.
In this scenario, we explore the effect is of testing people on the spread of the disease. Based on the current implementation of the model we want to evaluate four aspects: 1) testing family of sick persons, 2) testing colleagues of sick persons, 3) random tests, and 4) a single full large scale test.
For this scenario, we assume that agents are not directly aware they are infected at the moment of contagion. During the incubation period infected agents are able to infect others but they are not aware of this. After a couple of ticks symptoms appear and agents become aware of their sickness. This results in a difference between the variables Infected and Epistemic Infected. When aware of infection, the simulation assumes that the agents will stay home until they believe to be immune, in order to minimize further spread of the disease. Sometimes when an agent just became immune he can still think he is infected for a couple of days. After not having symptoms for a while it will believe it is Immune and becomes Epistemic Immune.
The simulation focus on the variables epistemic infection error, this is the amount of people that are infected but do not yet know it, and epistemic immunity error is the amount of people that are immune and do not yet know that. The current simulation does not yet take into account that people who receive intensive care have a lower chance of dying. Therefore one should only look at the infection and immune curve and in practice a flatter curve will actually mean less people dying. However this is not yet reflected in the upcoming graphs.
Simulation settings / Model configurations
We test the spreading of the coronavirus, using the standard NetLogo model under the following conditions:
- No testing at all
- Testing at home when someone in houseohold is sick
- Testing at home and workplace when someone is sick
- Randomized tests of population: ratio-population-randomly-tested-daily: 0.005, 0.02, 0.08, 1 per simulation step (there are 4 steps per day)
- Full population test at a certain amount of ticks: 15 ticks (right after 2 weeks)
Overall comparison of testing strategies:
Effect of the different testing strategies on number of cases:
Discussion & Conclusion
The results show in some cases a positive effect for flattening the curve when testing for the virus. Some methods seem more effective than others. For example testing 100% of the population flattens the curve the most in these single runs. However this is also the most resource intensive and not feasible in real-life situations. Testing family members of a sick person or testing colleagues does not seem to have a large effect on reducing the spread.