Coronavirus: How dangerous a threat?

[jmdirt]

The case rate in upstate NY (everything outside of NYC) is about 1.1%, so this would suggest about 2/3 of the infections are asymptomatic (actually a little more, because some in the sample would be recently positive for the virus, but not yet have developed Abs). This is very much consistent with many other studies, mostly of the virus rather than of antbodies, that have been discussed here, which suggest 50-75%.

But also note that the 3.6% figure should be an absolute maximum that the false positive rate could be--even if there were no true positives at all, which obviously wouldn't be the case--so unless the NYC tests were done differently in some way, it would seem that false positives would not have a major effect on the estimate of what proportion of people have Abs in the city. And the case rate for the city is not that much higher than upstate, about 1.6%.

A statistician argues that lockdowns don’t work.

https://www.thepublicdiscourse.com/2020/04/62572/

He defines lockdowns very precisely, as including three types of policies: compelling people to stay at home except for essential purposes; bans of gatherings of small numbers of people; and closing businesses indiscriminately. He isn’t opposed to many restrictive policies, including quarantining sick or exposed people; closing schools; travel restrictions; and outlawing large assemblies. He believes all of these practices clearly reduce spread of the virus.

Some of his main evidence that lockdowns don’t work comes from examining their effect on deaths. He begins by citing studies indicating that on average, most people die from C19 at least twenty or more days after infection. He then looks at the pattern of deaths following lockdowns in several major countries, and concludes that in most cases the number of new deaths peaked and began to fall before this twenty day period. IOW, he believes other, less restrictive measures, that were put in place before the lockdown, must have been the cause of the fall in number of deaths. He then goes on to provide evidence that these other measures work, even estimating their effectiveness (using county data from all over the U.S.) with a statistical model that determines how many days the measure has to be in practice to save one death per 100,000 people.

A well-thought out analysis. However, I think the conclusion lockdowns don’t work is overdoing it. He might argue at best that they don’t add any additional benefit, but to say that stay at home orders, for example, don’t reduce the spread of the virus, and therefore deaths, doesn’t make sense. How could they not?

In fact, his examples, though showing that peaks in deaths begin before this twenty day period ends, don't follow them long enough to see whether they continue to drop. Perhaps a further effect, due to the lockdowns, would be evident.

Also, I don’t know how one interprets the U.S.—granted, a very heterogenous situation—where the number of deaths has been more or less constant for the past two weeks. (By the way, what happened to that projection of 60,000 deaths in the U.S., I think by early August? We're at about 50,000 now, and appear likely to reach 60,000 by the end of the month).

There’s also the example of Michigan, where a stay-at-home order began March 24. The number of new deaths seemed to peak around April 16, 23 days after the order, suggesting that the order did indeed work by the author’s criterion. But after three days of lower numbers, a new high was reached on April 21, and the succeeding two days saw death totals around what had occurred during the peak. There are other factors, of course, that can affect the number of deaths and the author does take into account some of them in his county-by-county model.

By the way--this supports a point Cannavaro was making earlier--some people are touting some of these Ab tests as proof that the mortality rate is very low, only around 0.1%, no worse than the flu. Don't look now, but Belgium is already more than halfway there, with deaths per million of population at 0.056%. So unless half the population or so of Belgium is already infected, the mortality rate for that country almost certainly is greater than 0.1%.




.

I'm as much a statistics man as I am an epidemiologist, but its just a numbers game to solve for a and b when you only know the value for the solution isn't it? There is no way to actually determine what WOULD have happened if there were no stay at home orders in place, right? Sure you can create models with assumptions for that (the main) variable, but I just don't see it coming up with any meaningful solution(s). I think that you can arrive at those same assumptions by guessing or throwing darts at numbers especially since each variable in this situation also has variables of its own.

 

[Merckx index]

Just noticed this whilst I take a break working at home during our lockdown. I wonder how this man would explain Australia's case trend? Our lockdown commenced on March 22. Looking at the daily case trend since then I think the evidence is strong that lockdowns absolutely do work. The restrictions in Australia are very similar to what he mentions.

His model is based on daily deaths, which are too small in number and variable to be meaningful in the case of Australia. I should probably add that while I used coronavirus death data for the U.S. and Michigan discussion, he actually uses a probably more accurate measure, the difference between this year and last year in deaths from all causes.

There is no way to actually determine what WOULD have happened if there were no stay at home orders in place, right?

Well, he also considers countries that did not have stay at home orders, like Sweden and the Netherlands, and claims they came out about the same as those that did--or actually better. Of course, models don't prove anything, but I think they can be better than just guessing. The idea is to learn from our errors. If a model is way off in its projections, we should try to find out why. We really don't have any choice, because we have to act on the assumption that some policies are better than others.

Here's an idea I had wondered about: a way of detecting virus in the air in places where it might be particularly concentrated and hazardous:

https://www.fiercebiotech.com/medtech/swiss-researchers-develop-methods-to-sniff-out-coronavirus-air

Not sure it can distinguish infectious particles from just the RNA, but where there's smoke there's fire.

 

[jmdirt]

His model is based on daily deaths, which are too small in number and variable to be meaningful in the case of Australia. I should probably add that while I used coronavirus death data for the U.S. and Michigan discussion, he actually uses a probably more accurate measure, the difference between this year and last year in deaths from all causes.



Well, he also considers countries that did not have stay at home orders, like Sweden and the Netherlands, and claims they came out about the same as those that did--or actually better. Of course, models don't prove anything, but I think they can be better than just guessing. The idea is to learn from our errors. If a model is way off in its projections, we should try to find out why. We really don't have any choice, because we have to act on the assumption that some policies are better than others.

But there is no way to determine if his model is way off of projections because we can't jump in a time machine and go back to try it without stay at home orders. I agree that many models are better than guessing, but in this case there are too many variables left out to be even close to valid (again, I'm not a statistician).

Its much like (unscientific) people saying "see we didn't need to stay at home, the deaths are low". Maybe the deaths are low because we stayed home?

 

[Merckx index]

The NEJM recently published a study of C19 in Iceland, which has one of the lowest mortality rates in the world. The design, which tested subjects for the virus, is a little confusing. The researchers issued an open invitation for subjects—so the latter were self-selected, and thus could represent a biased sample. In fact, almost half reported respiratory symptoms, though questionnaire data suggested that many of these subjects had a milder flu. This group constituted 10,797 subjects.

They also selected a random group, 6782 subjects, and tested a total of 2283. These consisted of those who accepted the invitation. So this could be a source of bias, though again, the group from which they were taken was chosen randomly.

In the first group, 87/10,797 (0.8%) were positive. In the second group, 13/2283 (0.6%) tested positive. So the proportion of positives was basically the same in either group. The total was 100/13,080 (0.8%). A little less than half—43—of the positives reported no symptoms, though the researchers state, “symptoms almost certainly developed later in some of them.”

There is much more in this study I haven’t discussed. Another group of 9199 subjects were chosen for targeted testing, i.e., they were specifically selected because they were thought to be at risk for being positive. Of these, 1221 (13.3%) were positive. These people were further studied to determine where they might have picked up the virus, e.g., from foreign travel, and more than 500 samples were isolated to get an idea of the diversity of haplotypes.

But the main significance of this study, at least to me, is that it provides a timely counterpoint to the antibody tests suggesting a huge number of asymptomatic individuals. Less than half the positives were asymptomatic, and as I noted, the researchers believe this number would go down over time.

The rate of positives, 0.8%, is lower than other studies of this kind have found. However, Iceland's situation is different from most other countries in that a very large majority of cases (85%) are closed. That means that most of the people who tested positive before wouldn't now. So it's difficult to look at infection rates now and compare them to the actual number of cases reported. The authors say that the infection rate of 0.8% remained constant throughout the testing period. If it remains the same, now, we would want to compare it to the number of currently open cases. That's just 270, or about .075% of the population. The positive rate actually found is about ten times higher than this. Taking into account asymptomatics can at best explain a two-fold difference, so there is still a five-fold difference to account for. That could reflect a possible sample bias. As I noted before, in the larger study, people were invited. In the second study, the original group was thought to b random, but while the positive rate was a little lower, it was still considerably higher than indicated by active cases.

https://www.nejm.org/doi/full/10.1056/NEJMoa2006100?query=featured_home

 

[gregrowlerson]

His model is based on daily deaths, which are too small in number and variable to be meaningful in the case of Australia. I should probably add that while I used coronavirus death data for the U.S. and Michigan discussion, he actually uses a probably more accurate measure, the difference between this year and last year in deaths from all causes.



Well, he also considers countries that did not have stay at home orders, like Sweden and the Netherlands, and claims they came out about the same as those that did--or actually better. Of course, models don't prove anything, but I think they can be better than just guessing. The idea is to learn from our errors. If a model is way off in its projections, we should try to find out why. We really don't have any choice, because we have to act on the assumption that some policies are better than others.

Here's an idea I had wondered about: a way of detecting virus in the air in places where it might be particularly concentrated and hazardous:

https://www.fiercebiotech.com/medtech/swiss-researchers-develop-methods-to-sniff-out-coronavirus-air

Not sure it can distinguish infectious particles from just the RNA, but where there's smoke there's fire.

The link that Merckx index linked on the previous page is well worth a look for the total deaths by day and by region comparisons. I think that you can argue both ways about lockdown vs. mere restrictions, but it was enlightening to get a greater insight into the seriousness of the virus. I had wondered if the total death toll (not just from the virus) wasn't much higher than normal, but we can see on these graphs some significant increases (more than double the daily deaths in Spain, triple in Italy). Of course, what will also be interesting, will be to see at years end the total deaths in 2020 vs. 2019....because then it could be determined a little better as to just how many of those covid 19 deaths were deaths that would have occurred very shortly anyway.

 

[Jagartrott]

In Europe, the total number of excess deaths is still lower than the heavy 2018 flu season. In several countries, it is higher, obviously, but still - I sometimes think we are losing some perspective here.

 

[aphronesis]

Oh, look at this

Weiterleitungshinweis

www.google.com

“The World Health Organization has been working with thousands of researchers across the world since January in a bid to fast-track a development of a vaccine, its director general says, as he announces the launch of a new scheme to help achieve the goal.
WHO’s Tedros Adhanom says the organization is uniting with partners to launch the “Access to #COVID19 Tools Accelerator, or the ACT Accelerator”.
The “landmark collaboration” will accelerate the development, production and equitable distribution of vaccines, diagnostics, and therapeutics for Covid-19”

https://twitter.com/x/status/1253670647290445826

View: https://mobile.twitter.com/WHO/status/1253670647290445826

 

[djpbaltimore]

Anybody who is making the argument that Sweden has been a success to argue against lockdown orders is BSing. Sweden has done markedly worse than their neighbors. They are not facing massive hospitalization problems because there is evidence that they are triaging older patients (>70). If NYC turned away the elderly patients from their hospitals, they would also be doing OK in terms of capacity.

Sweden has a population of around 10 million. It has reported more than 16,700 cases of the coronavirus, and more than 2,000 deaths. In contrast, Denmark, home to almost six million, has reported around 8,000 cases and 394 deaths. Norway, home to over five million people, has around 7,400 cases and 194 deaths.

 

[aphronesis]

One of the things lockdowns do is to buy time; to parse the elements (so that the fear/uncertainty factor is not a vast, amorphous unknown); to save lives, to manage optics, etc. Saying they work or don’t work based on discrete outcomes or disciplinary models is a relic of the limited approaches that allowed this situation to proliferate and to not be presented comprehensively (rather than just quantitatively) in the first place.

 

[djpbaltimore]

Agree 100%. There is ample room to make arguments about the effectiveness of various interventions. Some lockdown requirements are clearly not grounded in science. But, most of the arguments I am hearing on this point are just not very good, made by people who are looking to open the country back up ASAP. They usually go hand in hand with claims that the virus is more widespread and less deadly than everyone thinks. Also, calling what we are doing in most states in the USA a 'lockdown' is a stretch in itself. We are a lot closer to what Sweden is actually doing than many realize.

This does not surprise me in the slightest.

https://twitter.com/x/status/1253659684051001344

View: https://twitter.com/GeoffRBennett/status/1253659684051001344

 

[jmdirt]

I dislike that protests get so much coverage when they represent a minuscule fraction of the population, while responsible people willing to sacrifice for the good of the all get no coverage. I mean I get that watching people carrying assault rifles and yelling is more entertaining than watching someone helping their kid with an online assignment, but...

Disobey Idaho had a rally with less than 200 people, so now they are doing things like going to playgrounds and taking down the police tape, getting arrested on purpose, and then marching (heavily armed) on police officers' laws. Ya know, things that good members of the community do.

 

[Koronin]

The lock down orders buy time to try to keep the hospital systems from becoming overwhelmed (as they did in Italy) and to buy time to hope a treatment becomes available. Now there is an antibody treatment which is helping people and there are several trials going on with other treatment drugs and will eventually get those reports.

 

[Koronin]

I dislike that protests get so much coverage when they represent a minuscule fraction of the population, while responsible people willing to sacrifice for the good of the all get no coverage. I mean I get that watching people carrying assault rifles and yelling is more entertaining than watching someone helping their kid with an online assignment, but...

Disobey Idaho had a rally with less than 200 people, so now they are doing things like going to playgrounds and taking down the police tape, getting arrested on purpose, and then marching (heavily armed) on police officers' laws. Ya know, things that good members of the community do.

There were supposed to be protests in Georgia to protest the reopening of the state. I agree the protests should not have gotten as much coverage as they did get.

 

[Oldermanish]

I dislike that protests get so much coverage when they represent a minuscule fraction of the population, while responsible people willing to sacrifice for the good of the all get no coverage. I mean I get that watching people carrying assault rifles and yelling is more entertaining than watching someone helping their kid with an online assignment, but...

Disobey Idaho had a rally with less than 200 people, so now they are doing things like going to playgrounds and taking down the police tape, getting arrested on purpose, and then marching (heavily armed) on police officers' laws. Ya know, things that good members of the community do.

I would bet most of those Idahoans didn't come from resort areas like Sun Valley. Almost everyone I know there got c19. We scratched a trip there in favor of Whistler, which probably was not much better but didn't involve air travel. They shut down a day earlier than we planned so we bought a 12 pack of toilet paper and headed South. The next day the BC Premier was telling Americans to stay home.
Very few protests in Wash state as they just shut down a huge beef processing plant with rampant employee positives on the East side of the Cascade Mountains. That area can have politics much like "Jefferson County" Norcal and they now have a reality much like the urban sides.

 

[Merckx index]

What I liked about the article arguing against lockdowns is not so much the conclusions as the approach. He's very transparent, stating exactly what he means by lockdowns. By focussing on deaths, he avoids the problem of asymptomatic positives. For any given population, there should be a fairly fixed relationship between deaths and positives. You might have problems comparing different countries by this method, but if you're only examining changes over time within a country, this death/cases relationship should be fairly constant. By using excess deaths over the previous year, he avoids the problem of undercounting or overcounting.

Wrt Iceland, because the fraction of closed cases is so high, we can make an unusually accurate estimate of mortality rate (albeit one with a high amount of uncertainty, because the absolute number of deaths is so low). If we use deaths/total cases, which is an underestimate, the mortality rate is 0.56. If we use deaths/closed cases, which is an underestimate, the rate is 0.66. So the true mortality rate should fall within that range.

This assumes no asymptomatic positives. If we use the 43% the study reported (which might be high, since it appears some of these subjects later developed symptoms), the rate falls to 0.40 - .45, which is close to what several other studies have estimated.

However, the 0.8% figure they report for positives would reduce that rate quite a bit. It projects to about 2900 active cases, whereas the actual number--correcting for asymptomatics--is about 475, or six times lower. So this has to be addressed. As i said before, it might be a selection bias, at least for the large group. If it is, it illustrates just how much that bias can affect results. Keep in mind that the two antibody tests in CA both used a similar method of recruiting.

 

[jmdirt]

What I liked about the article arguing against lockdowns is not so much the conclusions as the approach. He's very transparent, stating exactly what he means by lockdowns. By focussing on deaths, he avoids the problem of asymptomatic positives. For any given population, there should be a fairly fixed relationship between deaths and positives. You might have problems comparing different countries by this method, but if you're only examining changes over time within a country, this death/cases relationship should be fairly constant. By using excess deaths over the previous year, he avoids the problem of undercounting or overcounting.

Wrt Iceland, because the fraction of closed cases is so high, we can make an unusually accurate estimate of mortality rate (albeit one with a high amount of uncertainty, because the absolute number of deaths is so low). If we use deaths/total cases, which is an underestimate, the mortality rate is 0.56. If we use deaths/closed cases, which is an underestimate, the rate is 0.66. So the true mortality rate should fall within that range.

This assumes no asymptomatic positives. If we use the 43% the study reported (which might be high, since it appears some of these subjects later developed symptoms), the rate falls to 0.40 - .45, which is close to what several other studies have estimated.

However, the 0.8% figure they report for positives would reduce that rate quite a bit. It projects to about 2900 active cases, whereas the actual number--correcting for asymptomatics--is about 475, or six times lower. So this has to be addressed. As i said before, it might be a selection bias, at least for the large group. If it is, it illustrates just how much that bias can affect results. Keep in mind that the two antibody tests in CA both used a similar method of recruiting.

I wonder how the 2020 daily, monthly, and yearly death totals will be in the USA compared to other years. The USA is 52,000+ for C19 deaths, but all other causes of death may be reduced due to stay at home orders (less: flu, automobile accidents...maybe more self inflicted deaths though) so we might not be able to do a true comparison.

 

[Merckx index]

That study that found a very high % of asymptomatics in women in a maternity ward was published in NEJM:

https://www.nejm.org/doi/full/10.1056/NEJMc2009316

The original report was 29/33 positives had no symptoms, but as I expected, a few developed symptoms later. One original negative also tested positive. The revised tally is 26/34 asymptomatic, or 76.5%. This is in line with estimates from many other tests (and I think it's likely other women eventually developed symptoms). Also of interest, about 16% of the women tested were positive. This is, of course, far higher than the % of officially confirmed cases. But one could argue that pregnant women would visit a clinic frequently, where their contact with health workers would put them at greater risk for becoming infected with the virus.

At this point, about the only study involving virus testing I'm aware of that showed a far higher proportion of asymptomatics--in line with the conclusion of some of the antibody tests--was of a Boston homeless shelter. It was claimed that 136 individuals, about one-third of the shelter's population, tested positive, and that none showed symptoms. But given the possibility of symptoms developing later, I wouldn't take this number as final.

In fact, I managed to find a paper on this, albeit not yet peer-reviewed:

https://www.medrxiv.org/content/10.1101/2020.04.12.20059618v1

The number is now up to 147, of which:

Cough (7.5%), shortness of breath (1.4%), and fever (0.7%) were all uncommon among COVID-positive individuals.

Thus > 90% were asymptomatic. That's still reduced from 100%, though, and i think it's likely that more individuals would develop symptoms over time. I'm still betting on 50-75%, and I think any antibody test that concludes a far higher proportion exists will have to be really rigorous in ruling out factors such as bias and false positives.

That many asymptomatics may actually turn out to be presymptomatic, though, doesn't necessarily change our notions of risk. A study in China of people infected in clusters, such as families, reported that 44% of infections occurred from people in the presymptomatic stage:

https://www.nature.com/articles/s41591-020-0869-5[/quote]

The huge proportion of positives being reported in homeless shelters from cities all over the U.S. is completely consistent with this.

On a personal note, I just learned that a professor I knew well as a graduate student--I took a course from him, and my roommate was one of his graduate students--recently died from C19. He was almost 90, so clearly at risk.

 

[Jagartrott]

The Belgian government yesterday announced their exit strategy. One of the measures is that schools will not open for everybody yet - first (05/18) the most important grades, and even there only with half classes. Another thing is that they will further ramp up testing, up to 25-30.000 test per day, in emergency situations 45.000/day. That's a lot for a country of 11 million. We happen to have a large number of pharmaceutical companies here and university labs, so I'm guessing that makes it easier then.

Something I'm hearing very little about: the lack of huge outbreaks in Africa and India. Huge clusters of people, but the number of reported cases is (relatively) extremely low: lack of testing, climate factor or genetics? In India, a couple hundred thousand people die each year of tuberculosis, which barely raises eyebrows. Now, a hundreds of millions of people with little or no savings are locked down for a disease that is totally dwarfed by other causes of death in India. India also has a young population, so from my perspective, the lockdown is only there to protect the elites, and does more bad than good for most of the rest.

 

[jmdirt]

That study that found a very high % of asymptomatics in women in a maternity ward was published in NEJM:

https://www.nejm.org/doi/full/10.1056/NEJMc2009316

The original report was 29/33 positives had no symptoms, but as I expected, a few developed symptoms later. One original negative also tested positive. The revised tally is 26/34 asymptomatic, or 76.5%. This is in line with estimates from many other tests (and I think it's likely other women eventually developed symptoms). Also of interest, about 16% of the women tested were positive. This is, of course, far higher than the % of officially confirmed cases. But one could argue that pregnant women would visit a clinic frequently, where their contact with health workers would put them at greater risk for becoming infected with the virus.

At this point, about the only study involving virus testing I'm aware of that showed a far higher proportion of asymptomatics--in line with the conclusion of some of the antibody tests--was of a Boston homeless shelter. It was claimed that 136 individuals, about one-third of the shelter's population, tested positive, and that none showed symptoms. But given the possibility of symptoms developing later, I wouldn't take this number as final.

In fact, I managed to find a paper on this, albeit not yet peer-reviewed:

https://www.medrxiv.org/content/10.1101/2020.04.12.20059618v1

The number is now up to 147, of which:



Thus > 90% were asymptomatic. That's still reduced from 100%, though, and i think it's likely that more individuals would develop symptoms over time. I'm still betting on 50-75%, and I think any antibody test that concludes a far higher proportion exists will have to be really rigorous in ruling out factors such as bias and false positives.

That many asymptomatics may actually turn out to be presymptomatic, though, doesn't necessarily change our notions of risk. A study in China of people infected in clusters, such as families, reported that 44% of infections occurred from people in the presymptomatic stage:

https://www.nature.com/articles/s41591-020-0869-5

The huge proportion of positives being reported in homeless shelters from cities all over the U.S. is completely consistent with this.

On a personal note, I just learned that a professor I knew well as a graduate student--I took a course from him, and my roommate was one of his graduate students--recently died from C19. He was almost 90, so clearly at risk.
[/QUOTE]
Are you still working out in the pool? I understand that if you drink pool chemicals it will kill C19...and you. I hope that your recovery is still going well!

 

[movingtarget]

China and Singapore have had new outbreaks, the latter among migrant workers. Infections also increasing in Russia. Ecuador also seems to be having more problems than their government is letting be known. I would say the same probably goes for Brazil with some of the information leaking out and bypassing the lunatic ravings of their so called leader............

 

[BlueRoads]

Something I'm hearing very little about: the lack of huge outbreaks in Africa and India. Huge clusters of people, but the number of reported cases is (relatively) extremely low: lack of testing, climate factor or genetics? In India, a couple hundred thousand people die each year of tuberculosis, which barely raises eyebrows. Now, a hundreds of millions of people with little or no savings are locked down for a disease that is totally dwarfed by other causes of death in India. India also has a young population, so from my perspective, the lockdown is only there to protect the elites, and does more bad than good for most of the rest.

My thinking... but this is probably a political topic...

 

[Koronin]

I wonder how the 2020 daily, monthly, and yearly death totals will be in the USA compared to other years. The USA is 52,000+ for C19 deaths, but all other causes of death may be reduced due to stay at home orders (less: flu, automobile accidents...maybe more self inflicted deaths though) so we might not be able to do a true comparison.

Remember a few states never put any stay at home orders in place. Also some states only had those orders for about 2 weeks if that before reopening things.

NY did see a huge drop in non covid related hospitalizations.

Due to California having not near the expected number due to going into stay at home orders early they have reopened surgeries as long as they are for medical needs (no plastic surgery or cosmetic surgeries).

 

[aphronesis]

"Immunity passports" in the context of COVID-19

Scientific Brief

www.who.int

 

[jmdirt]

"Immunity passports" in the context of COVID-19

Scientific Brief

www.who.int

My wife and I were discussing this about a week ago. For international travel, I can see your passport now having C19 status. If you don't have *'proof' of immunity, you probably won't be getting in to many/most countries.

*vaccine, or medically accepted antibody level from previous infection?

 

[aphronesis]

Well presuming you were allowed to get on a plane (probably with up to the minute proof of testing) the receiving country can always test and quarantine you. WHO could set guidelines that many countries will choose to agree upon. Or it will fall out like travel during the Cold War. As with most visas there will probably be governmental and pay to play business modifications.