It simply must be much more dangerous than the flu for the simple reason that comparatively few people have it and it is overwhelming hospitals.

i wonder if it's really true that "few people" have it. We do observe a huge amount of people going to the hospital in italy, but that could very well be because we reached the peak of the epidemie, where almost all of the population have been infected and remain asymptomatic. in this hypothesis, you implement a lockdown and see number drops over two week but what you're really seing is a population reaching herd immunity.

When we are getting close to the peak of the epidemic, the growth will no longer be exponential, but close to linear. A logistic curve looks exponential for about the first 3rd, close to linear for the second third, and then approaching the asymptote with a negative exponential growth for the last third.

Increasing social distancing will naturally lower the exponential growth rate, but unless the growth becomes linear, there is no reason to believe that herd immunity is anywhere close.

In the asian countries, we more or less know that they have been able to contain the virus, instead of waiting for herd immunity. That meanst that they must have found most of the infected cases (or at least quaranteened them). This gives a good indication of how many deaths are caused by a given number of infected. The numbers are not different enough from Italy to think that more than (at most) a couple of hundred thousand in Italy have reached immunity. If so, it means that we have about 2 more orders of magnitude of growth before herd immunity can be reached. That would be hundreds of thousands of dead, if not millions, just in Italy.

That would be for a homogeneous spread inside a country.

However if the epidemic spreads from region to region, you could reach a local herd immunity in the first regions infected way before seing the country-wide curve becoming linear.

But thanks, trying to find a counter-argument to your post made me realize a single-curve model probably doesn't make a lot of sense to try and understand accurately the spread of an epidemic.

There aren’t large numbers of random people testing positive for it where they’ve done wide spread testing. Korea in particular probably has close to an accurate count, and their mortality rate is still close to 1%

Because differences in testing methodology are a constant factor, which doesn't mean shit to an exponential function.

What do you think of the practice of only testing people who need hospital-level care and people who had contact with confirmed covid-19 patients (and people who traveled to at risk regions)?

I would think that this can only give you the infected from a few clusters and you can be surprised by an unknown cluster of patients suddenly needing intensive care.

Edit: And due to the long incubation period, that cluster can be quite large.

Testing the entire population is not viable. If you don't have many in-country cases, you can delay the epidemics arrival by controlling just the people that come from elsewhere.

Of course, after the disease starts propagating internally, testing becomes irrelevant, and it's time to reduce its spreading speed.