Infection fatality rate of Covid-19 inferred from seroprevalence data [pdf]

Re:Infection fatality rate of Covid-19 inferred from seropre... by nradov on Saturday October 17, 2020 @08:36PM (#24814706)

As ghoulish as it may seem, I suspect another reason behind the declining fatality rate is that the earlier waves already killed some of the most vulnerable people.

[ Parent | on HN ]

Re:Infection fatality rate of Covid-19 inferred from seropre... by disgruntledphd2 on Sunday October 18, 2020 @08:56AM (#24817520)

Nah, it's a statistical artefact caused by testing policies.

In March, you needed multiple symptoms to get a test, which meant that conditional on having a test, your illness was much more serious.

Because this isn't an issue right now, it looks like the mortality rates have dropped when it's more likely that we are capturing a larger proportion of mild cases.

[ Parent | on HN ]

Re:Infection fatality rate of Covid-19 inferred from seropre... by beagle3 on Sunday October 18, 2020 @07:38PM (#24822033)

Obviously depends on where you look for this kind of reasoning, but I don’t think it is compatible with data in general.

Sweden’s death graph totally looks like “inventory of likely-to-die people exhausted”, regardless of testing (they have not changed recommendations or actions).

Israel had no first wave in April (it had a blip, which turns out was essentially limited to ultra orthodox religious which are about 15% of the population).

In September, the official 2nd wave but really 1st wave struck the entire population - and testing capacity was already high (about 0.7% of the population tested daily). And the stats looked way too similar to other countries’ first wave.

The US is a weakly connected network of hundreds of different repositories, which makes it really hard to observe similar processes - they are not visible on aggregate.

[ Parent | on HN ]