What would explain kids being so immune? ~1000 confined cases under 19 year old while you have 10 000 50-59 year olds?

According to today's interview with a NY Times reporter on The Daily podcast, the immunity is explained by the close relationship to the common cold. Children tend to get lots of common colds. Around 1/3rd of common colds are a form of coronavirus, because of this, their immune systems are more prepared for this new coronavirus. Over time, this immunity fades, but it takes many years.

https://www.nytimes.com/2020/02/27/podcasts/the-daily/corona...

This is the aspect of the disease that is so interesting to me. With many diseases, like the flu, the most at risk are the elderly and the very young. But barely 1% of infections in China are children younger than 10, and 0 deaths.

Can it be that smoking plays big part in this? As this also explains gender statistics.

An early preprint from the team of China’s leading expert Zhong Nanshan tagged smokers and non-smokers among the 1099 cases they analyzed. 85.4% supposedly never smoked, with 14.5% of those cases severe. So it’s not like non-smokers are very much immune.

https://www.medrxiv.org/content/10.1101/2020.02.06.20020974v...

https://www.medrxiv.org/content/10.1101/2020.02.06.20020974v...

I would say that it really dismisses smoking as a big factor then.

I wouldn't: https://news.ycombinator.com/item?id=22435408

The percentage of severe cases among smokers is definitely higher, although the sample size is small.

Looking at the numbers from Wei-jie Guan et al (https://www.medrxiv.org/content/10.1101/2020.02.06.20020974v...):

nonsmokers = 927 cases, 793 non-severe, 134 severe

exsmokers = 21 cases, 12 non-severe, 9 severe

smokers = 137 cases, 108 non-severe, 29 severe

That gives:

nonsmokers: 14% severe cases of presented

ex-smokers: 43% severe cases of presented

smokers: 21% severe cases of presented

The low number of ex-smokers as well as other factors (did they quit because of advanced smoking-related diseases, etc.?) makes that number hard to interpret.

But smoking carries a normalised average 40% increased risk of severe cases.

    100 * (21-14)/((21 + 14)/2) = 40

Isn't smoking much more common among the Chinese than this sample suggests? I would expect a random sample to include more smokers than not. It may imply that the act of smoking or the associated health states reduce the likelihood of infection, even if they increase the likelihood of a bad infection if it does land.

Or length of time exposed to air pollution, where smoking habit increases the likelihood of mortality?

Seems like the larger correlation in deaths is related to societal status - farmers, laborers, and retirees have much higher death rates.

Farmers — likely crappy rural healthcare.

Retirees — obviously age-skewed.

Quite possibly also: arriving sicker.

Hypothesis: If you're a rural / non-urban farmer, under quarantine and travel restrictions, you don't go to a hospital unless you're already very sick.

One of the real challenges here is in separating data artefacts from ground truth. Testing, surveillance, self-presentation, doctor-lobbying (by the patient or family) can all influence that.

This makes all data already somewhat suspect, before any overt manipulations are considered (and makes those manipulations all the more toxic to a full understanding).

And of course, there are numerous other factors. Farmers may well have poorer overall health (nutrition, behavioural risk factors (smoking, drink), preexisting conditions, latent or opportunistic infection, etc., etc.). All of which I'm very openly speculating on -- these are hypotheses which would have to be demonstrated through clinical assessment or systematic observation.

Note that "systematic" and "complete" aren't the same -- if you are following a defensible sampling method, then 100% coverage isn't necessary. But you'd want to capture samples from other populations.

The overall prevalence of COVID-19 within even Wuhan City and Hubei Province is still low, based on assessed cases (clinical symptoms or the more accurate PCR genetic screen for the SARS-nCoV-2 virus), still under 1%. Which means that population sampling would be unlikely to detect many cases without something on the order of many hundreds, possibly thousands, of samples. You're looking for an accurate central-tendency descriptor of a ~1:100 event.

Knowing what the actual exposure rate has been would be useful. Whether it's sufficiently useful to come up with a comprehensive large-scale sampling methodology is another question. Note that it's not merely the sampling size, but the selection and successful measurement from each of the selected individuals for sampling which is necessary.

(I've some statistics background, and some healthcare background. Enough to know basics of sampling, error, and bias. Epidemiological public health monitoring is beyond my pay grade however.)