Hacker News Reader

Comment by acangiano | Hacker News Reader

Also, 2% is very dubious. Italy is at 7%. The Philippines at 8%. I mean, sure these figures are out of people who were tested, but 2% is looking awfully optimistic at this point. So the potential for death reduction is even greater than the intentionally unassailable 1% figure you picked.

View on HN

In terms of fatality, the Diamond Princess cruise ship [1] has the helpful property that everyone on board was tested. Nobody recovering/asymptomatic not getting tested, nobody being refused a test.

The wikipedia page lists 700 infections and 7 deaths.

Of course there are still certain biases: on one hand it happened early, when hospitals weren't overloaded; on the other, cruise ships tend to have more older people. And as I don't speak Japanese I haven't traced those numbers to primary sources.

[1] https://en.wikipedia.org/wiki/2020_coronavirus_outbreak_on_c...

WHO Sitreps have the aggregated data.

https://www.who.int/emergencies/diseases/novel-coronavirus-2...

From: Situation report - 55, 15 March 2020

3.5% mortality worldwide (there are totals at the end of country listings)

..and they were trapped on that ship for days. Amesh Adalja (infectious disease MD at John Hopkins) has been in a few interviews where he commented, SARS-Cov2 is similar to other viruses where the initial dosage does impact the severity of the infection.

Note that in both cases there is a selection bias, due poor testing, and having the majority of initial reported cases being severe enough to reach hospitals. In my opinion, South Korea has better data, due to massive testing, and not having a completely overwhelmed health system. The rate there is 0.6%, which is still significant.

0.6%-1% is what we would be looking at when everyone gets the medical care they need. But many people are suggesting that instead of shutting down a large portion of the economy we let this thing rip through our population. And if we do that, a very small percentage of the population will get the healthcare they need. So the rate we should use when considering the "rip the bandaid" off approach should be closer to 6% than 0.6%.

> And if we do that, a very small percentage of the population will get the healthcare they need. So the rate we should use when considering the "rip the bandaid" off approach should be closer to 6% than 0.6%.

And if you take the "rip the bandaid off" approach, you can't restrict your analysis to just CoVID-19 deaths. The strain on the medical system will mean there will be an increase in deaths from other causes that didn't get the treatment they need.

Those tested represent a tiny fraction of those likely infected.

Yes. But South Korea is still at 0.6%, and they've tested everyone and their dog, and provided top care to everyone.

This is incorrect. As of the weekend gone past South Korea had tested 248,000 people out of a population of 50,000,000, with 8,086 +ve cases and 72 deaths.

<300k is not "and they've tested everyone and their dog"

There is a lot of memetic nonsense flying around about this online and this is part of it

Hm, ok. That would make it in the region of 1%. Thought my numbers were out of date since I heard 0.6% somewhere else.

1% of what?

8000-ish cases. 70-ish deaths => ~0.9% fatality rate. Very thorough testing, so 0.9% is likely to be a reasonable estimate of the real fatality rate.

Granted, 250k-ish of 50 million is a small sample, but you've got to assume that they haven't tested randomly but tested people who have been in touch with known cases.

You are almost there.

> but you've got to assume that they haven't tested randomly but tested people who have been in touch with known cases

Yes. Well with symptoms etc. So of the people who got tested they are much more likely to test +ve. I agree. Which means real cases are massively more widespread than those tested.

It is believed here in the UK that this has been fairly rife for a while now, but those who got it are not statistically counted anywhere. The real fatality rate is way less than 0.9

I don't follow your reasoning. South Korea has tested 35 times the number of positive cases. That's a lot. Presumably, it's not a random sample, but people who have been in touch with people who have had obvious symptoms. Maybe some random sampling in outbreak areas, I don't know their methodology.

Certainly there's a number of people that are not picked up throughout that process. This means that the real death rate is not ~70/8000. The real denominator is bigger.

But it wouldn't be on the order of 10 times bigger. In most of the groups of sick people, there's likely to be some that show enough symptoms to prompt a test, with a positive test leading most of that group to be tracked down. The exception would be tiny groups that somehow did not spread the disease further, or untracably spread it to similar tiny groups.

I don't buy that there's a massive amount of such lonely cases out there. Twice as many as detected? Sure. But not a dramatically bigger number than that.

Maybe there's a way to show mathematically that this reasoning is wrong, if so, please feel free.

I'll use London for an example. I personally know many people who believe they have this, and, I believe I and 3 friends picked it up from each other after one visited Milan in Feb. Loads of people have this and are not counted.

> But it wouldn't be on the order of 10 times bigger.

I believe it is of that order.

> In most of the groups of sick people, there's likely to be some that show enough symptoms to prompt a test, with a positive test leading most of that group to be tracked down

Literally not because most people won't be tested unless they are in need of hospitalisation

248k out of 50M is still a very impressive sample size; especially compared to most scientific papers coming out these days.

There is a persistent assertion that "everyone" in SK is being tested. It's not true. There is a pretty high chance that a high % of the infected are almost asymptomatic. Fatigue might be the only thing some get. Others nothing. The problem is people asserting without statistical extrapolation a base fatality rate predicated on a division of [+ve tests/deaths]. It is exactly what WHO have done before in other epidemic like situations and it doesn't make sense.

Previous epidemics have seen these massive similar ratios promoted only to later be revised aggressively. I can offer examples if required

I am curious of those examples. The big outbreaks in the past the WHO has focused on include MERS, Ebola and H1N1. There's a really good Wendover video on the WHO Ebola response and it's pretty impressive.

That said, Ebola got people sick quickly (compared to SARS-cov2) and it started in areas that weren't as globally connected, limiting the spread.

Still, there was a lot of hype of H1N1 that never really panned out. Still this is a totally new virus, from a family of viruses that has historically never been bad (a quarter of colds are caused coronaviruses).

If it's May 1st and there are <50k dead in the US with the patient rates going down, I think people are going to start to seriously question anything the WHO says going into the future. The alternative is that by May we'll be hearing news reports of how 90% of hospital in <insert big city .. London, Brussels, Atlanta> are at capacity with people being shipped to suburbs and individual 3D printing ventilator parts.

Honestly, I will be surprised either way.

Reply on news.ycombinator.com