About COVID-19 epidemic modeling

How to model the COVID-19 as accurately as possibly?

Different types of prediction models

Exponential fit

Describe pros, cons


Describe pros, cons

Agent-based modeling

Describe pros, cons

The most important thing which is lacking in basically all the models I have seen, is segmentation. The models treat the entire population as a uniform mass, even though both the fatality rates and the number of contacts IRL vary a lot between age groups. For instance, based on Swedish data, the death rate for 70-79 is about 100x compared to 40-. If clothing manufacturers would produce clothes based on similar market studies, they would definitely go bankrupt immediately.
Another important aspect is the time factor. There is no point in modelling these kinds of epidemics without a time limit. In this case it is max 6 months, because when people start losing their jobs and run out of money, nobody will care about COVID-19 anymore. The budgets of states are just a fraction of the size of the businesses, so they can’t help. Elderly people will die of isolation and lack of exercise, mental problems are already skyrocketing, destroying the future of the youth, and the funding of the healthcare system will result in much more victims than COVID.
When you add these essential factors, the conclusions will be the opposite of the ones you end up with without them. That is also the explanation to the Swedish approach, because the likeliness to become infected is more or less a function of the number of contacts in your daily life, and normally it is orders of magnitude higher among the resistent youth than among the vulnerable elder. In a lockdown, on the other hand, the Estonian data shows that the infection rate is actually higher among the old, which is a catastrophy for the total fatality rate.
So please, update your way of modelling asap, and publish the results.

The most important question a simulation should answer is: how can we cross this epidemic in x months with minimal damage.