Hacker News Reader

DP excludes the extreme elderly and sick. Remember, people in nursing care don’t go on cruse ships. Further, it’s expected to see additional deaths. So, such simplistic estimates are completely worthless from their data.

What it does do is provide estimates of the rate of infection vs obvious and extreme symptoms. China’s CFR is over 4% and rising, but it’s known that they missed a huge portion of the total case load.

PS: Further to estimate for the US population you need further adjustments. China has a younger population than the US with only 10.8% of the population being 65 and older, but they also have more air pollution etc.

Obesity plays a big role in Covid-19 deaths. 33% of Americans are obese, and 65% are overweight. There's a possibility that more obese people take cruises--it fits their lifestyles (unlimited food availability, and they don't have to move much). So the DP data may be tempered by more obesity even though it has less extreme elderly.

Also many of the crew got infected. So the amount of viral exposure may play a big part. People who had to come in contact with dozens of infected individuals every day may be the most vulnerable. This doesn't bode well for retail grocery workers.

Of the 712 DP infections, 145 occurred in crew average age 36 and 567 occurred in passengers average age 69. Far from every passenger was obese and few crew where obese. So, I doubt obesity was that far from US norms.

> Many obese people take cruises--it fits their lifestyles (unlimited food availability, and they don't have to move much).

Do you have actual data on that or is this just your prejudice speaking?

Anecdata: I've taken three cruises over the last twenty years, and the obesity rate on those was an order of magnitude higher than what I see in my everyday life as a coastal yuppie -- the only other place I've seen that many people who are that fat was at Disney in Orlando. But America as a whole has a way higher obesity rate than what I see normally, so cruises might just be representative of the country in general.

There have been some reports of the obese prefering cruise vacations because they can't fit on airplanes:

http://www.nbcnews.com/id/6661381/ns/health-fitness/t/obese-...

and

https://www.express.co.uk/travel/cruise/1263369/coronavirus-...

and

https://www.bariquins.com/newsletter/2016/august/03/cruising...

Perhaps you should examine your own "prejudice" before pointing fingers at objective, informed people stating facts.

Also, if the theory about BCG vaccination having a beneficial effect turns out to be true, USA never used BCG whereas China and India did (and UK between 1953 and 2008).

I would just like to point out the nursing home facts below:

The length of stay data were striking:

the median length of stay in a nursing home before death was 5 months

the average length of stay was longer at 14 months due to a small number of study participants who had very long lengths of stay

65% died within 1 year of nursing home admission

53% died within 6 months of nursing home admission

Source: https://www.geripal.org/2010/08/length-of-stay-in-nursing-ho...

> DP excludes the extreme elderly and sick. Remember, people in nursing care don’t go on cruse ships. Further, it’s expected to see additional deaths. So, such simplistic estimates are completely worthless from their data.

My estimate above assumes that DP's demographics & health conditions represent >=30% of a country's population and it is a lower bound so additional deaths don't affect it.

Your paragraph quoted here does not negate the logic behind the estimate since "the extreme elderly and sick" and those in nursing care can easily be included in 70% of the country's population.

Calling something a minimum does not make it useful. I can find a minimum bound by taking the total number of deaths divided by total global population. But, while true that’s a worthless bound.

Also, if you’re using passenger demographics then you should use the number of passengers infected excluding the 145 infected crew average age 36.

The goal of establishing a lower bound is to inform decision-making. A lower bound of 0 is trivially true and doesn't help decision-making. A lower bound of 0.5% rules out some policy options, though -- in America, it means over a million people would have to die to achieve herd immunity, which is probably unacceptable to political leaders.

The higher the lower bound, the more policy options can be ruled out, and hence the more useful the bound.

I agree, however if you have two accurate lower bounds then it’s only the higher one that’s useful. Aka X > 5 is not useful if you know X > 10.

If you know of a lower bound that's higher than mine, based on as/more reliable data, you should mention/link to it. People are doubting the Lancet paper based on data source (I don't have a strong opinion on that), so this bound based on a different source of evidence should provide additional information for policymaking.

Your lower bound based on the population of earth is clearly not it.

We can argue about specific examples, but I think you can more easily accept errors in your estimate so:

First, there are 12 not 11 known deaths. https://en.wikipedia.org/wiki/Diamond_Princess_(ship)

Second your using both passengers and crew numbers to hit 712, but 145 of those where crew average age 36 and 567 occurred in passengers average age 69. Thus the demographic estimate of 30% also low.

Finally, for accuracy you need to adjust for the healthier than average population receiving better than average care. But, that’s more debatable.

I use data from worldometers.

https://www.worldometers.info/coronavirus/

You can make adjustments and get better estimates with more work. You are welcomed to and let us know the results.

My point is to invalidate some wildly optimistic estimates like 0.1-0.2% fatality rate which is still getting spread around. With some adjustments on my model, the point would still stand.

“All models are wrong, but some are useful.” —- George E P Box

It also lets you judge whether other models are useful. If you see a model claiming an IFR the lower bound of 0.5%, it's wrong.

In case of China it is helpful to see the breakdown of in-Hubei (epic-center where lockdown enforced after widespread infection) / outside-Hubei. In Hubei CFR is around 4.73%, outside Hubei ~0.86%. It roughly shows how many cases they missed.

China is also lying about their numbers. They quarantined late, had a potluck of 40,000 people in Wuhan, let Wuhan people travel all over, then had hundreds of millions of people travelling over Lunar New Year.

Instead of the expected rise in cell phone customers, there was a 21 million customer drop. You'd expect stagnation or even a rise as most other forms of communication were shut downn.

https://www.bloomberg.com/news/articles/2020-03-23/china-s-m...

The companies got censured by the Chinese gov. Next thing you know, they're spouting this stuff.

https://apnews.com/afs:Content:8717250566

But even if you account for all the edge cases they cite to an extreme degree, you still have millions of mystery cancellations.

Testing may be inaccurate to up to 40% of cases. Not only because the tests are faulty (less likely), but because the virus is not present at the collection spot all the time (more likely) especially if the virus has entered but not yet multiplied enough. It is highly likely based on how this spreads that every single passenger has it right now.

Also, these ships may have a much higher death rate due simply because viral load may be much higher (another factor we are still guessing it). If the body gets a strong dose when the infection first happens, the anti-body building process may not be fast enough to fight it off.

If the body gets a very small load, it is more likely to start combating the virus, perhaps at a rate slower than replication rate, but eventually it can kill off even a major infection as long as it entered slowly.

I think they test the passengers there several times over a period of time. Need to check to be sure.

Viral load is a key issue but will be applicable to many countries too. See my answer to a sibling comment by hellofunk on this.

>DP's demographics is representative of > 30% of population of most countries.

However no one under 70 died from what I've read. So we can only get an upper bound on CFR for people under 70.

For a country with ~10% over 70 years old, and assuming infection fatality rate of ~6% among those, then IFR would already be 0.6%. So suppose that what you've read is the current statistics, it does not contradict the calculation above.

People 65 and over comprise >16% of US population, which is relatively young for a developed country. Here's US age structure: https://www.indexmundi.com/united_states/age_structure.html

Not everyone over 70 will be infected.

edit: bad math

IFR = #fatalities/#infected.

You might think of Attack Rate, the % of population who are infected.

You’re right, my mistake. Doesn’t this still assume even distribution across all ages? With current testing (at least in the US) people >=70 are more likely to have other conditions which complicate infection and therefore more likely to be tested.

Everything you've said here is consistent with the parent comment. You've got a lower bound on the infection fatality rate of the whole population, given the usual distribution of ages. But you've only got an upper bound on the fatality rate of a population artificially chosen to have an age distribution that includes nobody over 70. (Actually you do have a lower bound but it's 0%.) That is relevant to anyone who's under 70 and is asking about a population who's only member is themself.

Yes and I'm saying there's no contradiction. :)

My calculation above simply does not answer the question regarding IFR for each age group. It is based on overall statistics from Diamond Princess.

The sample size from DP might be too small to answer the question, esp for younger age groups with lower IFR. We might need to look at data from South Korea and Germany, for example, to estimate those. It would require a more complicated model since testing in any country is not universal.

> For a country with ~10% over 70 years old, and assuming infection fatality rate of ~6% among those, then IFR would already be 0.6%

That's assuming equal or greater rate of infection for people over 70.

That is, given 1000 people, 100 would be over 70. If 10% of them get infected (10 people), but 50% of younger people get infected (500 people), your IFR could be much lower.

We have some data on the age distribution of severe cases, but asymptomatic cases are virtually untested at this point, and are probably (? may be?) going to shift the distribution toward younger people (right now, we see more cases for older people, as well as more severe cases, because severe cases are more likely to be tested).

You're right. My hypothetical above was for a very limited context of answering its parent post only and not meant as an inference regarding the IFR for other age groups.

> Based on Diamond Princess data, where almost everyone was tested, the lower bound of COVID19's IFR (Infection Fatality Rate) should be > 0.5% for many/most countries.

No, you can't compare a country's likely outcome to that of a cruise ship. On ships, the virus propogates very differently and the viral load will likely be much higher in infections due to constant contact while onboard. When viral loads are higher, the severity of the disease and mortality rate are also much higher [0].

[0] https://www.nytimes.com/2020/04/01/opinion/coronavirus-viral...

That's a good point. It depends on how many infections on DP occurred after passenger isolations & mitigation measures, which would also lower viral load. I'm not sure we have that data. They stayed on the cruise for a long time. A Japanese expert went on the ship to examine the practice and he was very concerned of further infections.

Viral load from infection on DP would be quite close to those living in the same apartment/house/nursing home. In a country with multi-generation households, the viral load could be similar, unless measures are well taken. In other countries, the viral load would depend on how well the elderly are shielded from infection.

Yeah, two things worry me about Covid: the fact that the percentage is non-negligible even for young people and that the variance between case severities is high, even to the point where people think they don't have it but do.

Using the demographics of just the passengers is misleading because there's basically 1 staff per 2 passengers. Most of the staff are 20-40s.

The test they gave only tests for live viral load. It's possible there were cases that were asymptomatic, the immune system took care of it, and when tested they showed no virus, even though they had it.

The only real way we can know for sure is giving mass random antibody tests to a population.

> DP's demographics is representative of > 30% of population of many/most countries.

Regarding age maybe but what about health? Could there be a higher representation of people that have a long illness and want to go on a cruise before they can't?

> 1.54%/0.3 = ~0.51%.

Shouldn't this be 11 / (712 / 0.3) which is 1.54 * 0.3 = 0.462

CFR needs to use closed case. Either dead or recovered. Currently 21% world wide.