Infection Fatality Rate

[Pen]

This YouTube from John Campbell today looks at IFR which since it relates to a number of other discussions I think is worth its own mention not buried inside a long thread:

https://youtu.be/ofGMSH5tjYk

https://youtu.be/ofGMSH5tjYk

[Not_a_Number]

Ugh. This is like 61 studies. I can't discuss 61 studies at once, because these estimates are really dependent on how they sampled and lots of other things. And I'm suspicious of the author of this thing. 

I'm trying to find some good recent studies out of reputable places now. 

[Not_a_Number]

Like, take this study: 

https://www.medrxiv.org/content/10.1101/2020.04.26.20079822v2.article-info

This is ridiculously flawed. It's not a random sample (they are sampling from people who visited outpatient clinics!) We have no idea about the actual specificity of the antibody test they used. Their estimates of prevalence in the actual population is 100s of times greater than by PCR test. 

I mean... MAYBE. But there are many red flags here. 

[Not_a_Number]

And here's their other Japan study. Also from clinics! That is about the opposite of random. 

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7370565/

I really want a paper by someone who is not trying to make the IFR seem smaller. The author of this paper has been working on that project for quite a while. 

I'm not actually willing to look through every single paper on this list, but I would love to have a compilation of results from tests where there was reasonably random sampling, the test sensitivity was known (and preferably yielded a result of above 10%, or the false positives can swamp the actual ones!), and was measured in a country where I trust the death numbers. 

That would really cut down the list, let me tell you. 

Edited October 21, 2020 by Not_a_Number

[Pen]

Obviously there’s a lot we don’t know and great difficulty getting figures.  For Japan there is also the question of the “Neanderthal gene” where if that’s real it would be part of world with iirc low susceptibility to severe cases due to genetics. 

 

18 hours ago, Not_a_Number said:

And here's their other Japan study. Also from clinics! That is about the opposite of random. 

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7370565/

I really want a paper by someone who is not trying to make the IFR seem smaller. The author of this paper has been working on that project for quite a while. 

I'm not actually willing to look through every single paper on this list, but I would love to have a compilation of results from tests where there was reasonably random sampling, the test sensitivity was known (and preferably yielded a result of above 10%, or the false positives can swamp the actual ones!), and was measured in a country where I trust the death numbers. 

That would really cut down the list, let me tell you. 

 

18 hours ago, Not_a_Number said:

Like, take this study: 

https://www.medrxiv.org/content/10.1101/2020.04.26.20079822v2.article-info

This is ridiculously flawed. It's not a random sample (they are sampling from people who visited outpatient clinics!) We have no idea about the actual specificity of the antibody test they used. Their estimates of prevalence in the actual population is 100s of times greater than by PCR test. 

I mean... MAYBE. But there are many red flags here. 

 

 

 

 

Iceland ?   At least was a large sample compared to population. 

 

 

[Not_a_Number]

Just now, Pen said:

Iceland ?   At least was a large sample compared to population. 

 

It literally doesn't matter what percent of a population a sample is. It needs to be random, and it needs to be big enough in absolute size. (Apparently, this is a super common statistics error! I've made it before.) 

[Pen]

https://www.thelancet.com/journals/laninf/article/PIIS1473-3099(20)30769-6/fulltext

 

NYC? 

[Not_a_Number]

1 minute ago, Pen said:

https://www.thelancet.com/journals/laninf/article/PIIS1473-3099(20)30769-6/fulltext

 

NYC? 

I liked the design of NYC's antibody test sampling, and the results had a sufficient percent of the population infected that I think of that as one of the best results 🙂 . But as I'd be quite curious if the IFR is changing, I'd love to know which studies are designed well that are more current. 

[Pen]

5 minutes ago, Not_a_Number said:

It literally doesn't matter what percent of a population a sample is. It needs to be random, and it needs to be big enough in absolute size. (Apparently, this is a super common statistics error! I've made it before.) 

 

How would you do it truly randomly?

Force one out of every ___ people to submit to testing based on a computerized random generator ?

 

The Iceland study had over 30,000 people it looks like. 

 

[Not_a_Number]

31 minutes ago, Pen said:

Force one out of every ___ people to submit to testing based on a computerized random generator ?

NY did grocery stores. It's better than outpatient clinics, that's for sure. 

[Ditto]

29 minutes ago, Not_a_Number said:

NY did grocery stores. It's better than outpatient clinics, that's for sure. 

Woah, I would never go to the grocery store again!

[TCB]

Not sure if I’m understanding the issue correctly but I know they do very regular random sampling in the UK ( or at least England) to try and get an idea of the true infection rate.

[Not_a_Number]

3 minutes ago, TCB said:

Not sure if I’m understanding the issue correctly but I know they do very regular random sampling in the UK ( or at least England) to try and get an idea of the true infection rate.

Yes, that's not a bad idea. Any idea if they have IFR estimates? 

[Not_a_Number]

16 minutes ago, Ditto said:

Woah, I would never go to the grocery store again!

You mean in case you might have to give a sample? I think it was voluntary 😉 

[Not_a_Number]

2 minutes ago, happysmileylady said:

And specifically because we don't have the mitigations in place for this, that we have for the flu, is it really worth thinking about a difference between .1%, .2% and .6%  IFR....when we are really talking about things that have such unrelated levels of understanding and well studied treatment?

Well, the difference is the number of people who will die. So, 3 times deadlier, 3 times as many people will die. 

And we've never tried to figure out the number of people who get the flu as carefully as we've worked for COVID. So the IFR might be much lower. 

[TCB]

3 minutes ago, Not_a_Number said:

Yes, that's not a bad idea. Any idea if they have IFR estimates? 

I don’t know I’m afraid. I haven’t kept an eye out for that recently.

[Not_a_Number]

1 minute ago, happysmileylady said:

I am talking about in terms of the statistical impact once we factor in the mitigation factors that are or are not in place.

What is "statistical impact"? 

[Pen]

57 minutes ago, Not_a_Number said:

NY did grocery stores. It's better than outpatient clinics, that's for sure. 

 

But besides missing people who might be home sick and thus skewing toward people well enough to go to store, wasn’t it then voluntary at grocery store to be tested or not? 

 

[Not_a_Number]

Just now, Pen said:

But besides missing people who might be home sick and thus skewing toward people well enough to go to store, wasn’t it then voluntary at grocery store to be tested or not? 

Yes. It's not perfect. But it's better than outpatient clinics... and better than lots of things that would skew towards people who self-select heavily by volunteering or skew towards people who are out and about more than usual. (Most people go to grocery stores.) 

It's not perfect. But sampling never is. It's an art... 

[Not_a_Number]

1 minute ago, happysmileylady said:

So "statistical impact"....are the decisions being made based on how deadly this thing is.....or are they being made based on how fast our technology and data processing skills can manage.  

I think everyone expects it to get less deadly eventually. Especially after a vaccine... the question is what's going on now. 

[Not_a_Number]

Just now, happysmileylady said:

Right...what is going on now.  

Is what is going on now, any different than what has gone on for other illnesses that are completely brand now that we don't know anything about.

Probably not. Some of them killed a lot of people. Some were the plague and killed a lot more people than this. 

[Not_a_Number]

Just now, happysmileylady said:

So, if that's all the case....is it really so out of the box that there are so many people that think we shouldn't really treat this as so different?

As another new disease? Nope. We should treat it just like another new disease that kills and disables lots of people. We just don't have so many of those. 

[TCB]

3 minutes ago, happysmileylady said:

Right...what is going on now.  

Is what is going on now, any different than what has gone on for other illnesses that are completely brand now that we don't know anything about.

 

A dr I work with was talking to a patients family and said the last one anything like this was 1918.

[Pen]

23 minutes ago, happysmileylady said:

Yes, I realize that we are are looking at hard numbers, that's what we are talking about.  *sigh*

 

I am talking about in terms of the statistical impact once we factor in the mitigation factors that are or are not in place.

 

The thing that has most concerned me about those graphs that @Not_a_Number has posted that suggest that ifr has not budged is it would mean that no “mitigation” has done any good.

 I do not actually believe that.

I think there is something going on with the graphs that might be akin to when someone starts with telling someone to Choose a number and is told to double it, add something, divide etc and then from where they get the person can tell them their original number, or something - because it all algebraically cancels out. 

However  if IFR really has not changed at all then Masks, Vitamin D, better protecting people in Nursing Homes, treatment protocols...   all of that has not helped at all (because it is not “all or nothing” for most mitigations and thus there should be, if some strategies are helping, a reduced mortality rate as mitigations either reduce viral load or increase innate immunity, etc, or both. 

From people I know in medicine I am being told that it does seem like situation is much improved now.

And most graphs I see indicate that cases are skyrocketing, but deaths are not, which is good.

but the possibility that Ifr has not budged a whit does concern me

Edited October 22, 2020 by Pen

[Not_a_Number]

1 minute ago, Pen said:

but the possibility that Ifr has not budged a whit does concern me

Did you see the Arizona IFR calculation I linked before? 

[Pen]

1 minute ago, Not_a_Number said:

Did you see the Arizona IFR calculation I linked before? 

 

No - where?

[Not_a_Number]

1 minute ago, happysmileylady said:

The best mitigations we have against the flu, (or HIV) are medicinal.  They are medicines and vaccines.  And at less than 12m months old on this, it's not surprising that we aren't there yet with this thing.  

That's right. We want to do the mechanical stuff until the medicinal stuff catches up. 

[Pen]

1 minute ago, happysmileylady said:

The vast majority of the mitigations you mention are "mechanical"  Masks, social distancing in nursing home, etc....they are mecanical.

The best mitigations we have against the flu, (or HIV) are medicinal.  They are medicines and vaccines.  And at less than 12m months old on this, it's not surprising that we aren't there yet with this thing.  

 

The best mitigations against most seasonal flu may actually be excellent innate immunity.  (Which May partly depend upon nutritional status.)   The sort of host robustness that allows some people to never be sick a day in their lives despite being in War and other conditions where lots of other people are sick.  

[Not_a_Number]

1 minute ago, happysmileylady said:

So is comparing IFR of a disease we have medicinal mitigation for, to a disease where we have essentially no great medicinal mitigation for, really all that effective, from a public health standpoint.  Which brings it right back around to my original post......is the difference in IFR really all that important when the differences in IFR are mostly like comparing apples to oranges.  

I would think the point of public health is to kill and disable the smallest possible number of people. (And no, I don't just mean with the disease. I think the economic issues are pertinent.) 

[Not_a_Number]

7 minutes ago, Pen said:

No - where?

Here it is: 

https://www.statnews.com/2020/08/24/infection-fatality-rate-shows-covid-19-isnt-getting-less-deadly/

 

[Not_a_Number]

1 minute ago, happysmileylady said:

I recently saw a blurb (so no I don't have a link,) about how there are 300k excess deaths in the US.  Approx 200k of those are due to covid but the other 100k are due to something else.  

Probably most of the rest are also due to COVID. It's just that we don't count all of them, because it's too hard. In a pandemic or other natural disaster, you assume most deaths come from the obvious source. 

If you don't believe it, take a look at excess deaths in equally locked down states with fewer COVID deaths. Here's a document for Seattle, for example, which has been plenty locked down but hasn't had much COVID: 

https://www.kingcounty.gov/depts/health/covid-19/data/~/media/depts/health/communicable-diseases/documents/C19/deaths-associated-with-covid-19.ashx

[Pen]

1 minute ago, happysmileylady said:

So is comparing IFR of a disease we have medicinal mitigation for, to a disease where we have essentially no great medicinal mitigation for, really all that effective, from a public health standpoint.  Which brings it right back around to my original post......is the difference in IFR really all that important when the differences in IFR are mostly like comparing apples to oranges.  

 

Do you mean comparing to flu? 

I don’t think that’s especially helpful for all sorts of reasons, but I do think it *natural* for people to want to compare things of greater familiarity with something new. 

 

I think if we knew that the IFR for our area and age group (or people in our households) or people using a particular strategy or group of strategies were 20% vs 10% vs 1 % vs  0.1%  vs 0.01%  — it could make a difference. 

For example, I have yet to see any indication of concerning IFR for people with vitamin D levels around between 50 and 100 ng/mL, even if in an above age 50 group where Ifr otherwise seems to be higher due to age  (not so sure about any other comorbidity risk factor) — but it is hard to get solid info.  And there are probably so many complex and confounding factors that studies attempting to test that sort of thing don’t help a lot right now .  

 

 

[Not_a_Number]

Just now, happysmileylady said:

You really think so?  You don't think they are due to things like opioid addiction due to support groups being shut down, domestic violence due to access being limited, cancer deaths due to "elective treatments" being put on hold, or other accidental things due to people being straight up terrified to go to the ER?  The truth is that across 330million, 100k isn't all that much and I don't find it at all hard to believe that wonce all those factors are added up, it could easily add up to 100k excess indirect deaths.  

No. Because equally locked down places that didn't have many COVID deaths didn't have the same rates of excess deaths. That's why I linked you that document. Washington has had pretty serious economic impacts and it's not a very open state, so I assume all of those issued occurred there too. 

Someone linked Ohio's opioid deaths this year, and they were very similar to last year. So, there hasn't been any kind of epidemic -- maybe a bit of an increase, nothing spectacular. 

You know, this isn't actually something that can't be tested. You should think about how the data would look like if you were correct, and then you can test the hypothesis. 

[Not_a_Number]

6 minutes ago, happysmileylady said:

And that's really one of the biggest issues isn't it?   Policies created and decisions made based on what can't be tested.   

No, I said it CAN be tested. If you think the excess deaths are coming form non-COVID causes, you should be able to find some evidence. I wouldn't have been surprised if that had been true, but as a matter of fact, there's no evidence for that and there's plenty of evidence that it's just COVID. 

[Pen]

1 hour ago, Not_a_Number said:

Yes. It's not perfect. But it's better than outpatient clinics... and better than lots of things that would skew towards people who self-select heavily by volunteering or skew towards people who are out and about more than usual. (Most people go to grocery stores.) 

It's not perfect. But sampling never is. It's an art... 

 

In NYC that may have skewed away from people who could afford to have grocery delivery of some form.  It may also have skewed away from certain age groups (children whose parents do the grocery shopping?) or other demographics.  Are grocery shoppers in NYC nowadays equally distributed male and female? 

[Not_a_Number]

1 minute ago, Pen said:

In NYC that may have skewed away from people who could afford to have grocery delivery of some form.  It may also have skewed away from certain age groups (children whose parents do the grocery shopping?) or other demographics.  Are grocery shoppers in NYC nowadays equally distributed male and female? 

Sure, it could have skewed away in lots of ways, but it's very hard to be perfectly random. Grocery stores are better than lots of things, because many people do go to the grocery. Most people we know do, for example. 

Since no one is throwing darts at phone books then going door to door (it'd be way too hard), this is close to the best you can do. It's not perfect, of course. 

[Pen]

1 minute ago, Not_a_Number said:

Sure, it could have skewed away in lots of ways, but it's very hard to be perfectly random. Grocery stores are better than lots of things, because many people do go to the grocery. Most people we know do, for example. 

Since no one is throwing darts at phone books then going door to door (it'd be way too hard), this is close to the best you can do. It's not perfect, of course. 

 

They actually did try two versions of random sampling in Oregon, one was done door to door  by OSU students,  and one with contacts by mail from OHSU, iirc.   I don’t know what results were. The state may be too low population and have had too few cases at the time to have gotten meaningful results. 

[Not_a_Number]

Just now, Pen said:

They actually did try two versions of random sampling in Oregon, one was done door to door  by OSU students,  and one with contacts by mail from OHSU, iirc.   I don’t know what results were. The state may be too low population and have had too few cases at the time to have gotten meaningful results. 

Oh, neat. If you find what they got, I'll be curious. 

Anyway, anything but outpatient clinics, lol! That's the worst sampling I've ever heard of. Well, I guess the really heavily self-selected samples are similarly bad, but yeesh. 

[Pen]

1 minute ago, Not_a_Number said:

Oh, neat. If you find what they got, I'll be curious. 

Anyway, anything but outpatient clinics, lol! That's the worst sampling I've ever heard of. Well, I guess the really heavily self-selected samples are similarly bad, but yeesh. 

 

1 minute ago, Not_a_Number said:

Oh, neat. If you find what they got, I'll be curious. 

Anyway, anything but outpatient clinics, lol! That's the worst sampling I've ever heard of. Well, I guess the really heavily self-selected samples are similarly bad, but yeesh. 

 

This link may help you find more — Corvallis where OSU is is not the only area being sampled. I heard at least also Bend and Hermiston and maybe other places too. 

I heard some of the summer sampling plans had to be skipped due to wildfire and air quality problems. But Apparently they have gone back to doing it now. 

They have also been sampling waste water

https://theworldlink.com/news/local_free/osu-covid-19-sampling-suggests-three-people-in-1-000-have-virus/article_bb9c1b12-0830-11eb-bbd2-ffa50d426e65.html

 

[Pen]

Ah. OHSU one was stopped:

https://news.ohsu.edu/2020/08/27/key-to-oregon-study-transition-6812195

 

[Pen]

https://trace.oregonstate.edu/

@Not_a_Number 

[Ditto]

22 hours ago, Not_a_Number said:

You mean in case you might have to give a sample? I think it was voluntary 😉 

Yep!  I thought it was mandatory though.  

[Not_a_Number]

Just now, Ditto said:

Yep!  I thought it was mandatory though.  

I doubt they had enough energy to sample every single person in a grocery store!! 😄 

[Ditto]

Just now, Not_a_Number said:

I doubt they had enough energy to sample every single person in a grocery store!! 😄 

Very true.  I just had visions of me cluelessly walking along thinking about my food list and then getting stopped at the entrance by a Nurse Ratchet (sp??) type nurse with a cotton swab the length of my arm in her hand demanding to swab me before I could enter.  And I have no patience nor tolerance for that bull.

[Not_a_Number]

Just now, Ditto said:

Very true.  I just had visions of me cluelessly walking along thinking about my food list and then getting stopped at the entrance by a Nurse Ratchet (sp??) type nurse with a cotton swab the length of my arm in her hand demanding to swab me before I could enter.  And I have no patience nor tolerance for that bull.

I believe there was just a station inside a grocery store! 😄 And I think people actually tried to find out where they were, and went to find them, because this was before antibody tests were prevalent... 

[Pam in CT]

re definitions, and whether or not "mitigation strategies" are affecting IFR:

23 hours ago, Pen said:

The thing that has most concerned me about those graphs that @Not_a_Number has posted that suggest that ifr has not budged is it would mean that no “mitigation” has done any good...

However  if IFR really has not changed at all then Masks, Vitamin D, better protecting people in Nursing Homes, treatment protocols...   all of that has not helped at all (because it is not “all or nothing” for most mitigations and thus there should be, if some strategies are helping, a reduced mortality rate as mitigations either reduce viral load or increase innate immunity, etc, or both....

IFR measures the proportion of deaths among infected individuals.  Mitigation strategies aimed at lowering transmission -- masking, distancing, doing stuff outdoors, individuals avoiding and/or policies prohibiting large gatherings -- all of those strategies aim to reduce the number of infected individuals. But that won't affect IFR.  Even if those strategies, or a near-complete lockdown, reduced what is today 73K new cases down to just 730 infected individuals -- which would, obviously, be a *massive improvement* for affected individuals, their families, hospitals, downstream health conseqquences, downstream bills... it wouldn't necessarily affect the IFR. Just the denominator on which IFR were calculated.

 

23 hours ago, happysmileylady said:

The vast majority of the mitigations you mention are "mechanical"  Masks, social distancing in nursing home, etc....they are mechanical.

The best mitigations we have against the flu, (or HIV) are medicinal.  They are medicines and vaccines.  And at less than 12m months old on this, it's not surprising that we aren't there yet with this thing.  

As a language thing, I'd call what you're calling "medicinal" strategies either "treatments" or "vaccines."

Vaccines are also aimed at reducing the number of affected individuals, so -- like what you're calling "mechanical" strategies like masking and distancing -- would actually not reduce the IFR (just the denominator).

Effective treatments *would* affect IFR, by reducing the number of people who do become infected from dying.

[Frances]

23 hours ago, happysmileylady said:

You really think so?  You don't think they are due to things like opioid addiction due to support groups being shut down, domestic violence due to access being limited, cancer deaths due to "elective treatments" being put on hold, or other accidental things due to people being straight up terrified to go to the ER?  The truth is that across 330million, 100k isn't all that much and I don't find it at all hard to believe that wonce all those factors are added up, it could easily add up to 100k excess indirect deaths.  

There’s also research going on looking at the possible reasons for reduction in some common causes of death due to things related to the shutdown such as less air pollution and driving. I recently read an article about a study that used data from a large number of hospitals across the country. 

Have elective cancer treatments been put on hold long enough anywhere in the US to actually lead to deaths already?