Age-specific mortality and immunity patterns of SARS-CoV-2 | Hacker News Reader

Age-specific mortality and immunity patterns of SARS-CoV-2

nature.com

69 pointsdivbzero58 comments

Every single paper on IFR estimation seems to mostly ignore that poorer countries without infrastructure aren't capable of and are not capturing a large fraction of the covid-19 caused deaths.

They all show a clear trend of lower IFR in places like Kenya, Peru, etc and significantly higher IFR in places like Switzerland, England, France, Germany, etc. It's not hard to figure out which is closer to reality.

The only odd-man out is China, and that's obviously because China kicked out all journalists and started faking numbers back in March and have continued to report falsehoods since.

I’m not sure that they ignore anything. Most poorer countries have a younger population, and we know that mortality goes up exponentially with age. This is why the IFR is 1/10,000 for someone in their 30s and 1/10 for someone in their 80s

Yes that's a factor. On average poorer countries also tend to have lower rates of the co-morbid conditions which increase COVID-19 risk such as obesity, diabetes, hypertension, and vitamin D deficiency.

IFR = "Infection Fatality Ratio" (for those who were unfamiliar with the acronym like me).

I could guess form the text but was about to look it up so thanks for disambiguating the term here. Wish people would add in parens the meaning of the acronym they use at the first occurence in the text then refer to the acronym only.

> Wish people would add in parens the meaning of the acronym they use at the first occurence

I find that style utterly patronizing. For uncommon abbreviations it would be better to introduce them in parenthesis after their meaning is spelled out on first use. The reader gains by looking-up common abbreviations.

nature publishes research papers. Those tend to address primarily fellow researchers familiar with the terms.

HN is a generalist forum. People who use specialized acronyms in threads here inconvenience the generalist HN user base. Just as we would expect researchers writing for specialist journals to write for their specialist audience, we should expect HN posters to write for the generalist HN audience.

It might not entirely be incorrect either. Very few deaths are officially reported but the deaths which weren't going to be reported, were of the same subset of people who never tested either, despite symptoms.

So it would be safe to assume while both the infected and fatality numbers are way off, they are at least representative. Hence, I believe the IFR isn't incorrect.

https://www.businessinsider.com/coronavirus-china-back-to-no...

I thought it was because of the aggressive response in the beginning that China is back to normal?

You are correct. OP is spreading conspiracy theories.

People are having a hard time swallowing it. There is still a lot of Western hubris in the world, and I say that as a westerner.

> They all show a clear trend of lower IFR in places like Kenya, Peru, etc and significantly higher IFR in places like Switzerland, England, France, Germany, etc. It's not hard to figure out which is closer to reality.

Of course it's hard to figure out! The life expectancy in Kenya is 65 and 83 in Switzerland. It could be that there aren't many octogenarians in Kenya to kill.

fatality numbers in general is captured quite well by many countries.

The infection numbers are admittedly unreliable.

This means that we still get a very valuable signal of deaths per million population and deaths per million in hospitals (weaker for the 2nd one).

Testing coverage has the biggest variance accross countries.

Given how quickly the virus spreads, given that China stopped all of its lockdowns in May, and given the nature of exponential growth, if China were lying about its numbers, by this point, they would have tens or hundreds of millions of cases.

If that were the case, your Chinese coworker with family back home would have known about their entire extended family getting sick.

... Yet, their extended family is not sick, and life in China goes on as normal.

What is the most plausible explanation for this?

PS. China's self-reported IFR was ~6%. If they are making up all the numbers, do you think they might have fabricated a lower number?

You’re downvoted for the obvious truth, which is that China’s numbers are almost certainly real.

People who still don’t get it can’t accept that China has beat us here. If you’ve been there in the last 5 years, I think it is entirely unsurprising that they did.

For Peru at least we have some data that shows their excess mortality is far greater than their official numbers:

https://www.economist.com/graphic-detail/2020/07/15/tracking...

Data as of Apr 1-Oct 31

Covid 19 Deaths: 34,446 Excess Deaths: 82,076

Back of the envelope they are estimating a little under 2 undiagnosed cases for every confirmed case. Putting our “actual” total around 42 million as of today.

I have no reason to doubt this particular study but I’ve seen other studies estimating up to 10x actual for every confirmed. It seems there is no great way to know this without some consistent, nationwide, antibody testing effort. Which hasn’t happened and by now may be too late to ever happen.

The fraction of infections which are diagnosed varies quite a bit by circumstance, e.g. available test capacity. In general I get the impression that 10x was typical early on and 3x may be more typical now.

https://www.acpjournals.org/doi/10.7326/M20-3012

Metanalysis mostly based on data from the first spike.

Looks like about 50% of cases are asymptomatic in the general population, with a higher proportion of asymptomatic cases in younger populations?

Remember that not everyone with symptoms is going to bother getting tested. In fact, everyone I know who has had possible COVID symptoms in the past few months hasn't bothered to get tested.

On some level I don't really know why you would get tested, except if you're curious. I'm a person in my 30s. If I got covid symptoms, I would warn everyone I've had close or transitive close contact with -- that is, our nanny, my son's tutor, and my son's classroom. But I don't know if I would bother to get tested. How does knowing you have it help? I guess if I was sure I had it it would be nice to get confirmation so that I could have a strong belief I'm immune going forward. But other than that I'm not sure how it helps.

It helps because some of those people you are warning have jobs with actual consequences if infection, or live with people that do. Like your sons teachers.

You go get tested so you can tell them “false alarm it’s all good”.

no you can't, and that is part of the problem. even if you test negative, the test isn't 100% and you still have to ask as if you have it

Right that's my understanding. I guess this is a good reason to consider getting tested though.

It can help if you get results quickly enough. And if you trust the results. If I came into close contact with someone who tested positive I wouldn't trust a negative test. If I had no symptoms I'd quarantine for 10-14 days. For people with tertiary exposure, through me, I have no idea how they should handle it.

I was severely ill with upper-respiratory symptoms about a month ago. Knowing if I had covid was critical information. If I had covid, going to a hospital or clinic for help was safer (for me) I can't catch covid twice. If I didn't have covid, I was already sick and getting covid on top of that would be a serious danger to my life.

I got close to needing to go to the hospital but decided not to. Turns out that was a good idea, I didn't have covid (PRC test).

Even antibody tests can underestimate the number of infections as the levels can drop below the detection threshold in some patients after several months. This effect will become more pronounced the longer the pandemic continues. In order to get really accurate data on infections we have to repeatedly test the same subjects, like in the UK's REACT study.

https://www.imperial.ac.uk/medicine/research-and-impact/grou...

Or alternatively test for T cell response. But that's a more complex and expensive assay.

It feels like the more aggressive the estimate is for the upper bound the less accurate it sounds, whether that's true or not. Somehow it just feels more believable that it's 2x instead of 10x without considering evidence either way. I think epidemiology is a tough field because you never actually have any verifiable ground truth.

The paper doesn't say so, but it seems there may be some process by which a the scenario where there are a large number of simultaneously infected individuals living in close proximity tends to produce worse outcomes in all those individuals.

I also came away with this impression. It seems fairly easy to explain, though. There are probably some bottlenecks in the treatment process. E. g. getting everyone to the hospital in time might be challenging if there is a large outbreak in a nursing home, even if there are enough beds and nurses at that hospital.

There’s pretty decent evidence that viral load is a big determinant of infection severity.

This guy I work with, his mother had it quite badly. He had to care for her in a temporary hospital for almost 3 weeks while she recovered. This was in a pretty poor area in India.

He said he saw about 25 people die during his time there.

This story either negates or supports your case, but either way , I’d imagine they were exposed to a pretty severe viral load in that time. I didn’t ask about PPE, I doubt he has much.

He is fine now AFAIK never really got sick, maybe be was already immune ?

There are many reports of households where the first person is infected outside the household, but only in passing, gets a low initial exposure, and develops a mild case. But then the subsequent infected people in the household develop much more severe cases because of higher initial exposure in the household.

I am sick and tired of everyone saying African countries are resisting better on paper because they can't test.

During the Ebola, AIDS, Yellow Fever crises/épidémies we did not have the means either... But we could count the bodies.

Eg: Gabon is a small country with a pop just under 2M. The capital is 60%+ of that. There are 2 morgues there. All are seeing a bump in deaths, but no unprecedented activity.

Something is really happening. There must be a reason why sub-Saharan countries are relatively ok, but everyone is only saying: it's because of poor data.

(rant) HN feels like the_donald sometimes. Incredible that such a high number of interesting & intelligent people can stop thinking altogether. (/rant)

I feel like the answer is obvious. The countries are younger and healthier.

Look at the age distrabution of Gabon and compare it to the US or Italy. We know that covid is most deadly to the elderly, with 80+ being 100X more likely to die than someone in their 20s. The entire country of Gabon has <4,000 people over 80!

https://www.populationpyramid.net/gabon/2019/ https://www.populationpyramid.net/united-states-of-america/2... https://www.populationpyramid.net/italy/2019/

If you lived in an African country and you couldn't handle this flu you would be dead already.

Vulnerable people live in rich countries.

Given the article say 5% have had it what can we infer about total deaths in the do nothing/herd immunity strategy?

If you can wait for two months then just visit https://www.worldometers.info/coronavirus/country/us/ and you will find the results.

They are estimating infection rates by combining mortality stats specifically for the younger part of the population and seroprevalence studies.

So if you accept the known biases in the seroprevalence studies, and make the (strong, IMO) assumption that an individual’s risk of severe disease is not correlated to behavior that reduced their odds of infection before September (the cutoff in this study), and account for the age distribution in the as yet uninfected population, maybe you could get reasonable ish bounds.

I’m skeptical because a change in behavior probably changes relative risk of infection, and I think people with known risk factors are probably more careful on average.

Anyways, the first thing to do would be to check such an estimate against data for October and November

I see what you mean here, it makes sense that those at risk didn’t go to illegal raves or BLM protests. Therefore the infection rate could be higher than 5%. My guesstimate from the small sample of people I knew is 10-15% of people lost their smell in late February to early April and some had horrendous symptoms and a few had those symptoms last for months! I wouldn’t say many people I know have been affected by the second wave so who knows.

From that stat alone, not very much.

- Your best estimate on a do nothing strategy would be that total deaths would be 20x what it currently is,

(I'd assume everyone catches covid before herd immunity becomes relevant: here, aiming for herd immunity is a different strategy than 'do nothing')

- plus likely a whole bunch more due to overwhelmed health services, likely a large proportion of people needing intensive care would also have died,

- plus likely a whole bunch more people would have had a high viral load initial exposure and more likely to have consequential symptoms,

- plus likely a whole bunch more, reflecting a likely relatively safe age and risk distribution of people who have caught covid to date.

> I'd assume everyone catches covid before herd immunity becomes relevant...

That's just not how it works. As the number of immune people (whether through survival, natural immunity or vaccine) in a population increases, the chances of anybody who is contagious spreading the disease begins to decrease. Herd immunity is going to be reached before everybody catches it, no matter what.

Yes I read 60-70% is enough for herd immunity.

It takes weeks before people stop being contagious. That's plenty of time to pass it around to everybody in a "do nothing" scenario.

(Of course, that is in a naive model with homogeneous population. If we've learned anything from this pandemic is that the population is heterogeneous enough for breaking those.)

That is not correct. The vast majority of patients stop being contagious within 10 days of infection.

https://www.evms.edu/media/evms_public/departments/internal_...

Patients may continue have positive PCR test results for weeks, but that doesn't mean they are contagious.

>The vast majority of patients stop being contagious within 10 days of infection.

Nonetheless, if growth starts out exponential then a majority of people would have been recently enough infected.

You need to play around with a simple herd immunity model. You'll see what I'm talking about.

This makes a fractional difference at most. And, I disagree on the following grounds -

Let's say we assume the point of herd immunity is when 60% are infected (which I also doubt: models were not taking into account superspreading events when this was established as an estimate)

Then we can agree that if ~60% of the population is seropositive and no longer infectious, everyone (everyone else) is now safe. Cool.

However, if there is uncontrolled infection, then we reach the 60% mark of herd immunity, you bust through that ceiling because most people are still infectious and continuing to infect others.

Herd immunity will be reached at a percent of the population being infected. What percent that is depends on a lot of factors, including the rate of infection (R0), the length of contagion, etc. Nevertheless, herd immunity will be reached at less than 100% infected with Covid-19. I'm not saying that it will be as low as 60%. I highly doubt that. I'd say closer to 80-85%, as a guess based on my simplistic math. I'm no epidemiologist. Then again, neither are you, I'm guessing. :)

Months ago there was a fascinating link on HN that had models you could play with to adjust all the variables in a simple model. You should find it and play around. I'll try to find a link when I can come back to this later.

OK, I get you. In my model of no changed behaviour, people who are coincidentally isolated during the final weeks of virulent spread might ultimately avoid infection. That's still going to be 99%, 98% infected? 95% seems doubtful to me under that model - but perhaps you are right.

100% with a rounding error.

Your model is absurd. In what world would nobody change their behavior?

> * Herd immunity is going to be reached before everybody catches it, no matter what*

That’s only true for a very slow-moving epidemic. For a fast mover, like this one, there is always -always- an overshoot of the steady state number needed for “herd immunity” (which, btw, is a term created to describe vaccinated populations, not wild infections).

No. It's true for every epidemic except extremely fast-moving and long-lasting ones. Covid is not that.

Public health and epidemiology professionals say otherwise.

https://www.nytimes.com/2020/05/01/opinion/sunday/coronaviru...

That article literally shows less than 100% of people getting Covid.

As a parent of a 7-month-old, it is very frustrating to see the age range 0-4 years old. Everything I read says children under 1 year is at higher risk, but then all the statistics group that higher-risk age group with the lower risk age group 2-4. How am I supposed to understand the risk to my child when I know the stats are including lower-risk populations? How much higher is the risk for 0-1? If anyone has any insights into the under 1 year age group, I would really appreciate it.

My understanding is that a vaccine for children will not be available for some time. I will have to make some decisions about daycare this spring and want to make the most informed decision possible.

Numbers for age 0-1:

https://www.heritage.org/data-visualizations/public-health/c...

Newborns and infants are higher-risk for everything, simply because their immune systems are very immature. There are also minimum ages for all the common childhood vaccines. Not particularly comforting, but something every parent deals with. I'd avoid daycare under age 2 if you can.

Reply on news.ycombinator.com