Hacker News Reader

Comment by svara | Hacker News Reader

I think it would be more helpful to cite actual epidemiology research and public health professionals rather than people who seem to use people's fears to garner attention.

Edit: In order to make this a little less dismissive, the actual argument in that Twitter post is that due to lag between first symptoms appearing and death, you can't divide fatalities by total cases to get the mortality rate, and should instead use total cases N days ago. Makes sense superficially, but

* cases that get reported are biased towards people who get more sick

* the 20-56% mortality rate claimed based on guessing N=6 is inconsistent with the mortality rates outside of China, where no one has died so far.

But I'm no epidemiologist, so don't believe me either.

View on HN

The rates for Chinese provinces except for Hubei seem inconsistent with 20-56% mortality too. The ten provinces outside Hubei that have the most confirmed infections are:

* Zhejiang: 428 confirmed; 0 deaths; 4 recovered

* Guangdong: 354 confirmed; 0 deaths; 10 recovered

* Henan: 278 confirmed; 2 deaths; 2 recovered

* Hunan: 277 confirmed; 0 deaths; 2 recovered

* Anhui: 200 confirmed; 0 deaths; 2 recovered

* Chongqing: 182 confirmed; 0 deaths; 1 recovered

* Jiangxi: 162 confirmed; 0 deaths; 5 recovered

* Shandong; 158 confirmed; 0 deaths; 1 recovered

* Sichuan: 142 confirmed; 1 deaths; 1 recovered

* Jiangsu; 129 confirmed; 0 deaths; 1 recovered

Although there's a clearly a lag between confirmed cases and death, there's likely to be less lag in outcomes between patients who recover and patients who die.

On this admittedly small sample, we get 29 recovered and 3 deaths, or about a 10% death rate. That's still pretty high, but it's nowhere near the 20-56% death rate.

If people go from being infected to dying more quickly than they go from being infected to being declared to have recovered, then that would lower the rate. The recovered rate seems to have been increasing faster than the death rate in the last couple of days, which might indicate this is the case. In addition, these cases obviously only include patients who present with symptoms. If some people are asymptomatic, that would also lower the death rate.

That's a very good observation on these 10 provinces. But I would apply caution when interpreting a cherry-picked data subset.

Nitpick: I don't claim "20-56%" death rate, but "9-56%", thus your observation of a 10% CFR is in line with my claim.

One other observation, for what it's worth. According to the John Hopkins site, between 29th Jan and 30th Jan updates (that's the data I happen to have access to), we have worldwide:

* 29th: deaths 133, recovered 126

* 30th: deaths 171, recovered 143

But for Hubei, we get:

* 29th: deaths 125, recovered 88

* 30th: deaths 162, recovered 90

Which looks like the death rate is going up in Hubei.

Subtracting Hubei though, for the rest of the world (but mostly China) we get:

* 29th: deaths 8, recovered 38

* 30th: deaths 9, recovered 53

* difference: deaths 1, recovered 15

Of course these numbers are too small to have real significance, but they're worth watching over the next day or two. It looks like outcomes are very different outside of Hubei from in Hubei.

It's consistent with natural selection. Cases inside Hubei likely have fewer viral generations (= less mutation) than cases outside. Additionally, every case outside has a population bottleneck in that it was transmitted through a host that was well enough to travel. The most virulent cases would be expected to sicken their hosts and discourage travel or extensive social contact.

Surely all current cases have experienced roughly the same number of viral generations? A death outside Hubei today will be caused by a virus with roughly the same number of generations as a death inside Hubei. Yet a smaller fraction seems to be dying outside Hubei.

Other possible explanations are that the healthcare system in Hubei is overwhelmed, so more patients die, or alternatively they're just too busy to accurately report recoveries.

Hubei is estimated as being max at the 5th generation, and genomes so far have indicated less than 0.02% genetic variance. I don't see evolution playing a succincts) significant role in this timescale.

Not really surprising considering medical resources in Hubei are heavily strained, not all severe cases can be diagnosed timely and given enough attention afterwards, which is not true elsewhere.

Edit: Oh you mentioned this yourself in another comment.

Just to update with the latest statistics, since 29th Jan, up to 30th Jan 9:30pm EST, for rest of world outside Hubei, there have now been 1 death and 33 recovered. So the trend seems to be that the death rate (calculated as deaths/recovered) outside Hubei is falling. Still early days and small numbers though.

«cases that get reported are biased towards people who get more sick»

True, the data is biased.

«the 20-56% mortality rate claimed based on guessing N=6 is inconsistent with the mortality rates outside of China, where no one has died so far»

It's not statistically inconsistent. When China had 100 cases confirmed, they had reported only 2 deaths. Now there are 100 cases outside China, so statistically we should expect 2. 0 actual vs 2 expected is statistically insignificant. Also the rest of the world is so much more prepared after the wake-up alarm call from China that out of the 100 confirmed outside China, they are probably receiving superior healthcare than overwhelmed Chinese hospitals. I would expect the mortality rate outside China to be slightly inferior.

It also means though, that the first death outside China should happen soon. I predict in the next week.

> It's not statistically inconsistent. When China had 100 cases confirmed, they had reported only 2 deaths. Now there are 100 cases outside China, so statistically we should expect 2. 0 actual vs 2 expected is statistically insignificant.

My point here was that the China data is probably biased because the medical system there is overtaxed, but outside of China it isn't. Therefore, estimating mortality based on non-China data seems safer, and no one outside of China has died so far.

If I take your low end mortality estimate, 20%, and apply that to the number of non-China cases 6 days ago (12), you get only a 6% probability of all of these people surviving. Note that this already contains a lot of assumptions (going back 6 days, taking your lowest mortality estimate) that work in favor of your alarmism.

When the WHO declares something a "global health emergency" we have valid reasons to be alarmist :)

I think that "going back 6 days" is insufficient, so your conclusion of a 6% probability is unsupported. Given the extreme state of paranoia around the world (eg. authorities testing people coming out of airplanes) it is likely that cases outside China are detected very early after the onset of symptoms. As such, death, when it takes place, will probably take more than 6 days after a confirmed case is detected.

In fact, the studies in The Lancet document cases of patients that take much more than 6 days to die.

We can speculate all night, it remains speculation. All I've done is use the numbers you suggested to point out inconsistencies. Yet you keep finding reasons for why we are almost literally all going to die (with mortality rates of up to around 50%), when you have no more insight into this than anyone else in this thread.

Everybody is taking this seriously, but playing prophet of doom is really shitty and has all sorts of potential negative consequences in the real world.

I suppose I didn't communicate clearly. I never said we are all going to die. For one, we are probably going to find a vaccine or treatment. And the eventual observed CFR may end up being in the lower end of what I quote (9%) instead of the upper end (56%).

You tried to point out 1 inconsistency, and I replied back saying it wasn't an inconsistency ("0 actual vs 2 expected is statistically insignificant.") It's too early to tell if the CFR outside China will differ from the CFR in China.

Yes, and the medical professionals dealing with this are getting statistics about diagnosis and recovery rates and times by analyzing case histories, not applying estimation formulas to totals.

Data gets aggregated and used in mathematical models that can predict outcomes in various hypothetical scenarios, but given that this is a real-world situation, the professionals are going to be looking at the actual data first to find out what’s happening. They can directly answer questions such as, of the patients diagnosed 1 week ago, what’s their status today?

Anyone looking at only total numbers of cases per day simply doesn’t have the data required to compute the things they’re trying to compute with much accuracy. It’s possible to produce estimates with confidence intervals based on sensitivity analysis, but nobody doing that with any care using aggregate data is going to come out with a number as high as 20% for a lower bound. Given that N=6 is a guess, the range reported should be bounded by plugging in plausible extreme values for your guess of N, not assuming your guess of N=6 is exactly correct.

>Cases that get reported are biased towards people who get more sick

True, however I fail to see how this point is relevant when comparing statistics with past diseases - as the same could be said of our historical data we have for SARS or anything else. In any case, we only have reported confirmed cases. Doesn't seem to make sense to me to factor in a unknowable variable here but not in other cases...

Well keep in mind that China, at least Hubei is overwhelmed. But if anywhere else on earth it ramps up to 100k sick quickly will be overwhelmed as well.

It remains to be seen if the rest of the world can keep the infection rate low enough to not repeat the Hubei progression.

Reply on news.ycombinator.com