COVID is characterized by clusters, rather than sustained transmission like e.g. flu. This determines variability in numbers (high) and actions we should take (early and decisive).

A short thread on the article of @zeynep about overdispersion 1/6

( example transmission)

2/6 Overdispersion (the k parameter @zeynep talked about) indicates there's more spread in the numbers than expected. This means that, compared to flu:

- more people don't transmit the disease at all
- more people are superspreaders, infecting a lot of people.

3/6 as a consequence, outcomes differ greatly between regions pure by chance: some get lucky, and some are flooded with cases _under the same circumstances_.

But with only few cases, far more regions can escape an exploding COVID epidemic compared to flu. Good news.

4/6 When you let cases rise, it becomes increasingly difficult to control the epidemic. The chance of having no new cases after 10 steps (40-50 days) decreases dramatically with rising # infections. So the longer you wait, the more effort it takes to gain control again.

5/6 As @zeynep wrote, R can be suppressed by:
- early, rigorous contact tracing to identify clusters
- distance, masks, sanitation, ventilation, ...

Or, if you don't want that:
- long and painful lockdowns when hospitals get flooded with patients https://www.theatlantic.com/health/archive/2020/09/k-overlooked-variable-driving-pandemic/616548/

6/6 Japan, Taiwan, New Zealand, South Korea, ... show us it's possible to have a relative normal life and healthy economy if you take action in time. "In time" being the crucial factor here (looking at you, EU and US...)

Code + extra explanation:

https://jofam.github.io/covidBE_analysis/Overdispersion.html