Hacker News Reader

Comment by cycrutchfield | Hacker News Reader

This is not really valid due to fatality rate being heavily dependent on age, comorbidities, and healthcare quality (which is itself dependent on case load), and the distributions of these differ greatly across countries. But, sure, you could try to normalize these to infer the actual case load.

View on HN

There is a lot of assuming going on so this doesn't actually prove anything but I think it's something worth thinking about.

Is Spain and S. Korea really that much different to have a 4x difference in death rate?

It would be useful if this could be broken down into separate categories - estimated CFR without medical intervention, and likely CFR at various levels of intervention.

The former suggests what your population mortality rate will be if your healthcare system is overwhelmed, and could possibly be derived from Italian data. The latter gives you a resources vs effectiveness sweet spot that maximises health system throughput, so you can save the maximum number of lives over time given the resources you have (or can build/find.)

The age profile is relevant, but worryingly there seem to be a number of reports that it's not just the over-65s who are at risk.

I haven't seen anything more recent about this than the Chinese estimates from a month or so ago.

CFR can't be used to determine population mortality rate. It in no way reflects r0, and only tangentially at best hints at percentage of asymptomatic infections (and that's recent as there was never drive-by testing for any previous epidemic). Those two dimensions add unfathomable complexity to the equation. In many ways it's better to leave those dimensions out to improve the utility of CFR. A doctor doing triage or health department coming up with orders wants a simple, basic number, not a panoply of options that require data they won't have at hand.

We by all means should collect more data, just don't shoehorn it into such a very primitive yet very essential statistic. Give them new names and new, more appropriate semantics. I suspect that after this is over they're going to tighten the criteria around CFR, and in particular exclude by definition asymptomatic cases and possibly non-hospitalized cases. They never had to do that before as "cases" usually implied someone sick enough to be given a diagnosis, which in turn implied someone at least moderately sick--e.g. actual or suspected case of pneumonia.

Right - eg when the ICU is full and refusing new patients, the death rate is going to be higher.

Don't you first of all have to offset the infection rate by the time-to-death? I've been puzzling over all these extremely bad estimates we hear, and the only reason for it that I can come up with is that nobody wants to point at the much darker truth.

One study I've seen pinned the median time-to-death after symptom onset at 18 days. Even in Korea you won't get less than a 2-3% CFR with that estimation.

edit: I don't know anything about statistics by the way, please tell me that I'm wrong.

Yes, absolutely. But that's why we should use case fatality rate and not just divide #deaths/#infected

It seems like the biggest confounder is age. If we stratify across age, we might be forgiven for ignoring differences in healthcare quality. The death rates by age in Korea are given here:

https://www.businessinsider.com/coronavirus-death-rates-by-a...

20-29: 0%

30-39: 0.1%

40-49: 0.1%

50-59: 0.4%

60-69: 1.5%

70-79: 4.3%

80+: 7.2%

honest question: what is the death rate for an average person in their 80s?

ie: if you have cov2 you have a 7% chance of dying from it but dont you already have say a 50% chance of dying in your 80s anyway?

to be strictly logical and not emotional about this moral question: are we unhinging society (effects we cannot begin to calculate) to keep the fatality rate for the population 80+ at 50% instead of 57%?

in terms of costs to the healthcare system, might this equate to a net relief if the peak were accelerated instead of flattened?

Here are the death probabilities for males in their 80s based on the 2016 social security life table:

https://www.ssa.gov/oact/STATS/table4c6.html

> honest question: what is the death rate for an average person in their 80s?

That's a standard actuarial problem, answered with life tables (https://www.cdc.gov/nchs/data/nvsr/nvsr61/nvsr61_03.pdf -- pdf, US 2008). In said life tables, there is a 7% death rate between ages 83 and 84.

To an order-of-magnitude approximation, infection with the novel coronavirus seems to impose roughly one year's worth of mortality risk on the victim.

That’s not an entirely fair assessment given that most people dying from Covid-19 have other serious health issues.

It’d be interesting to see how many healthy 80+ year olds have died.

Mortality rates are freely available: https://www.ssa.gov/oact/STATS/table4c6.html

There's a ~6% chance of dying during the year you turned 80. That chance goes up dramatically every year you get older. By 85 it's nearly 10%.

It would be interesting to calculate how this virus changes those rates. I feel like these kind of analysis are completely lacking in any reporting.

It's not only about the old people. There are enough younger people that need medical care, even if they don't die from it.

And then it's also about the rest of the population that needs medical attention for a different reason.

Imagine the medicial facilities being flooded by COVID-19 patients. There are enough severe cases already in many countries that non essential operations are suspended.

Good luck getting adequate treatment for your broken leg.

so if you have flu like symptoms are you going to the hospital now? the vast majority of cov2 cases result in mild flu like symptoms and only 25% of the severe cases require medical intervention (according to WHO). so shouldnt the advice be to not seek medical attention at all? if you need the medical intervention youre in the 80 plus cohort that is already "flooding medical facilities" (been in one lately?). im just trying to do the long division and work out the logic that leads to the conclusion we should gamble with the livlihoods of huge percentage of the population for 7% mortality reduction in seniors. couldnt the exact same effect be achieved by quarantining seniors?

I mean, this is what the UK is going to try. We will see how successful it is. I am not optimistic.

You have to remember that these fatality rates are with treatment, and old people will experience roughly 2 years of death rates within a few months in that case, leading to massive numbers of hospitalizations. That case load is absolutely not normal regardless of who's driving it. What's your plan exactly -- deny treatment to old people who have COVID-19? Not only will that increase death rates substantially, it's horrible.

https://www.statnews.com/2020/03/16/coronavirus-model-shows-...

And all of that is ignoring the possibility of people suffering nonfatal chronic health problems, which has been reported.

Meanwhile, remdesivir and lopinavir/ritonavir are in trials and once we actually have drugs available the hospitalizations would be shorter and the death rates much lower. At that point quarantine becomes less important.

Reply on news.ycombinator.com