No significant effect of lockdowns on Covid-19 spread.

[PeterNSteinmetz]

A peer-reviewed study showing no significant effect of more restrictive policies like lockdowns on the spread of Covid-19 in a trans-national comparison. "After subtracting the epidemic and lrNPI effects, we find no clear, significant beneficial effect of mrNPIs on case growth in any country."

Cautions in interpretation would be that this is one study which must be weighed with others and is an observational study. Some of the strengths are that it attempted to account for normal growth of the pandemic and less restrictive measures like recommending social distancing.

https://doi.org/10.1111/eci.13484.

I will leave it to the moderators to decide whether this post is still permitted under the Covid-19 discussion policy, or if due to being a trans-national comparison, is too violative of the no politics rule.

 

[PeterNSteinmetz]

@PeterNSteinmetz Care to comment on the overall data comparing the rates of death per capita between Sweden, Denmark, and Norway as David M has shown? Raw death data does not have any implicit bias built in as studies often have. Unless you can claim that the death data is truly an over-counting of COVID-related deaths (even though many sources show that the death data is actually an under count of COVID deaths).

That is interesting to look at the graphs. What is interesting here is that Sweden was one of the less severe interventions which was studied by Benavid at al. linked at the start. They controlled for time course of development of the pandemic using a temporal development model. I do not know what the results of applying such a model to the current more recent data would be or including Denmark and Norway in the comparison.

I just checked and I don't see any peer-reviewed articles or preprints of this question on the servers. That itself makes me suspicious of this graphical argument. If it were that obvious, I should think it would already be published or at least in preprint form. I see that this is a popular argument and graphical comparison in the lay literature with a lot of people making arguments on both sides of the issue.

OTOH, the following article says that there was little change in all-cause mortality in Norway and Sweden during the initial parts of the pandemic, which strongly suggests your hypothesis that some of this has to do with how things are counted could account for at least a substantial portion of this graphical effect (https://www.medrxiv.org/content/10.1101/2020.11.11.20229708v1). This article looked through July, which includes a large portion of the initial difference which is shown in the graphical comparison presented above.

In general I don't think one can draw strong causal conclusions from trans-national comparisons generally without controlling for a lot of other variables and temporal development.

I do agree that death rates are subject to less bias than case counts, which are correlated with testing levels. Though they are still subject to some reporting bias as death cause assignment is not a cut and dry sort of determination.

What is rather clear in terms of death rates is that in the US the introduction of coercive lockdown measures was not significantly correlated with a decrease in deaths attributed to Covid-19 when comparing states. So that is definitely at variance with what the graphical comparison of death rates in these 4 countries would suggest.

 

[PeterNSteinmetz]

This study is comparing countries with mandatory lockdowns versus voluntary ones. It is reasonable to expect that some people in non-mandatory countries voluntarily self-isolate and at the same time people in mandatory countries ignore lockdown rules. You can't therefore use this study to draw much of a conclusion of how social distancing effects viral spread. It is more a study on the relationship of how laws alter human behavior—and the answer might be not a whole lot. The same thing is seen when measuring the effect of speed limits on the actual speed people drive.

This is an important point to realize, though I would argue that if one is interested in the effects of such lockdown policies, this study has a fairly good utility. Considered by itself, it argues that such policies don't slow the spread of Covid-19. That could be due to the fact that people don't really follow them.

The Google mobility data actually would argue that this may well be the case. People started voluntarily reducing their travel by about 40-50% 2-3 weeks before any of the coercive lockdowns were put in place. In most states, average mobility after the lockdown orders actually began to increase and was back near baseline about 4 weeks later. One could make a Devil's argument that the lockdown orders caused those increases, but I suspect it was just an effect of people reaching their limit with that type of thing.

 

[PeterNSteinmetz]

Interesting, so that would give little utility to any study that had variables for which there is no realistic way to control . . .

I think it would be fair to say that all scientific studies have their limitations. I would not agree that the study has "little" utility in terms of telling us whether coercive lockdown measures slow the spread of Covid-19. I would argue that this study, considered individually, argues fairly strongly that such measures do not slow the spread. Even though it did not find a significant effect of such interventions, one can consider what the data imply as an estimate of the effect on the rate of spread. And that analysis says the lockdown orders actually INCREASED the rate of spread by 7-13%, but there is rather large chance that increase was simply due to random chance.

I have said it before in these discussions that when considering these results it is important to recognize that studies in the biomedical sciences are very rarely as certain as those in the hard sciences like physics or something like engineering testing. The data will always be softer given the inherent limitations of the field. The presence of an uncontrolled variable does not necessarily imply a study is somehow worthless or fatally flawed. Rather, it must be judged relative to the totality of the evidence available in all studies. These studies can generally be grouped into 3 categories in terms of decreasing strength of the evidence: randomized controlled studies, observational studies, and in-vitro or laboratory studies.

There are several observational studies on the effects of lockdown policies on Covid-19 cases or deaths. This study adds to the body of evidence that such policies have no or a negative effect in terms of health outcomes.

 

[PeterNSteinmetz]

Notice that the flu hasn't been truly discussed in 2020, or so far into 2021. The flu didn't just disappear, they are lumping flu numbers in, to raise the number of cases they call covid. It has been blown way out of proportion!

I suspect this true from what data I have seen, though have not investigated deeply. I suppose it is possible that basically COVID-19 has taken over as the flu strain for this season.

 

[PeterNSteinmetz]

Peter, if you don't trust any of the doctors that are the source of the data on flu and COVID rates and numbers, then you also can't trust any of the studies that are created by the doctors either. So there is no point at looking at anything.

I agree and as I say have not investigated this very deeply. @asicer’s post suggests that the seasonal flu may not have decreased, but a comparison to the numbers over the last several years would be informative on this point.

Here is a link to the cases through 2019. https://www.statista.com/statistics/861113/estimated-number-of-flu-cases-us/

 

[PeterNSteinmetz]

But you agreed with @YKA

Please read what I wrote. I said I suspect it is true and have not looked deeply.

I am looking for a direct comparison of flu cases from 2020 to years before presently.

 

[PeterNSteinmetz]

View attachment 93441

So that suggests that the 2019-20 season was perhaps a bit worse in terms of % of visits for influenza than 2017-2108. Though not shockingly so. Not a ton lower or higher.

But would that estimate include the cases for Covid-19? The source of this graph would be useful in that regard.

Another interesting item is that the start of the 2010-2021 season appears to starting on the lower end.

 

[PeterNSteinmetz]

Yes. Exactly, and expected with the measures in place to prevent respiratory virus spread.

Israel had an absolutely terrible flu season last year and was super worried about a repeat this year along with COVID, but they have also had remarkably low influenza infections so far.

That interpretation seems at variance with the other graph and with the CDC data estimating flu cases here https://www.cdc.gov/flu/about/burden/past-seasons.html.

Normally they are estimating 35 million cases per year.

If you look carefully at the @asicer graph it seems like the axes may be mislabeled. It would be good to have a link for the source.

 

[PeterNSteinmetz]

I would guess that some small amount of covid cases in Jan-March 2020 were misreported as Flu as there were no available tests in the US at that time. But the number is probably only a few percent otherwise across the country. In NYC, it could have been significant as COVID was already raging there by April.

Perhaps the source of the graph so we might understand their methodology? The critical question is whether a case of Covid-19 would be counted as an ILI by this reporting network.

If so and that is the intention of their definition, I think the graph argues for exactly @YKA’s point. There has been no appreciable change in the rate of ILIs when when counts seasonal influenza and Covid 19 together as ILIs.

 

[PeterNSteinmetz]

Yes, he was, but I felt he got a non response

I agree. I think the idea there were 40 cases seems incredible. And that there is something wrong with the scale of those graphs. Need the source to understand it.

 

[PeterNSteinmetz]

Both charts are from this link from the CDC.
https://www.cdc.gov/flu/weekly/index.htm

My chart does come with a waiver that it could be picking up COVID, but the one from @ascier is for lab confirmed tests of influenza.
There have been 468,064 lab tests for influenza with only 1,159 positive (0.2%) this year so far.

Thanks. So doesn’t that suggest that the total number of cases of ILIs in 2019-2020 is about the same as 2017-2018? That the majority of these are Covid-19, which has essentially displaced the normal seasonal flu.

And isn’t that sort of what @YKA was suggesting was the case?

I find this plausible in terms of how viruses evolve and compete with each other. A more “successful” virus like Covid-19 infects all the susceptible hosts that the seasonal flu used to.

ETA: I note the CDC says the graph of the seasonal comparisons explicitly states it will capture SARS-Cov-2, the pathogen in Covid-19.

 

[PeterNSteinmetz]

Influenza is way down because it needs human-to-human contact to spread, and there's been a lot less of that.

The causes of the decrease of type A and B are likely a bit difficult to determine exactly.

But do you think there is data on less human - human contact? The mobility data, which is only a loosely correlated measure, were down for about 4-6 weeks back in February - March 2020, but I think have rebounded.

I have not looked into this deeply either, but would be curious about measures of actual close contact other than the mobility data.

 

[PeterNSteinmetz]

It's extremely unlikely that most doctors and nurses believe this, or that they've managed to bribe the labs and state public health authorities to collude with them.

I agree that some sort of underhanded bribes or collusion are highly unlikely.

OTOH, consider this scenario. Given that the total ILIs may be about the same with COVID as past bad flu seasons, it is possible that previously people who came in and passed and this was partly contributed to by a type A or B flu would simple have been reported as dying of the primary other cause like heart disease, diabetes etc. And the hospital would have received no special reimbursement for reporting the flu.

In the present day, person comes in with same situation except Covid rather than type A or B flu. Now there is a significant financial incentive to report the Covid.

People respond very strongly to such financial incentives and so whereas a case might have previously not had the flu mentioned or it was a questionable cause of death, now there is a strong driving force to identify and list the Covid. This could easily bias numbers reported in a significant way. And hospitals have entire departments dedicated to optimizing revenue by listing the right causes of illness and death.

 

[PeterNSteinmetz]

Apparently part of the stimulus package “It is true, however, that the government will pay more to hospitals for COVID-19 cases in two senses: By paying an additional 20% on top of traditional Medicare rates for COVID-19 patients during the public health emergency, and by reimbursing hospitals for treating the uninsured patients with the disease (at that enhanced Medicare rate).

Both of those provisions stem from the Coronavirus Aid, Relief, and Economic Security Act, or CARES Act.”

https://www.factcheck.org/2020/04/hospital-payments-and-the-covid-19-death-count/

As I noted above, hospitals have entire departments devoted to finding the proper codes for billing which will optimize revenue.

 

[PeterNSteinmetz]

I am a critical care doc. We screen by PCR for multiple viruses, the panels vary by hospital, most have limited supplies. I have yet to have an influenza case this season. That is weird, but we are testing for it. Our panels test for influenza, parainfluenza, multiple corona viruses including Covid19, metapneumovirus, RSV etc. We are still mainly just seeing COVID. I think it is as likely that covid is undercounted as over-counted, and excess death estimates seem to suggest more people are dying from something. I haven’t seen any signs of a conspiracy, and nobody has asked me to overcount COVID cases. If anything, we have been asked to be judicious with testing, as until recently we have been short on testing supplies. It is not a perfect science. The tests aren’t perfect, patients are complicated, but I think most medical people are trying to be accurate.

Would you happen to have any good sources regarding the excess deaths.

I have also not looked into this deeply but remember the JHU professor having given a lecture stating that in fact there did not appear to be an increase in total mortality. It was subsequently removed from the website.

 

[PeterNSteinmetz]

Just a quick look shows they have been tracking influenza during the pandemic.
See here: https://www.cdc.gov/flu/weekly/pastreports.htm

See link above- CDC has influenza numbers. COVID-19 != influenza different viruses altogether.

Of course. But if you look at the data from the CDC involving surveillance from outpatient visits, it does strongly suggest that the total ILIs (which includes COVID and other flus) in 2019-2020 season are similar to a bad prior flu season like 2017-2018.

 

[PeterNSteinmetz]

JOOC, how effective are cloth masks compared to surgical mask in this regard? If surgical masks block, say, 95% of the virus, what percentage does a cloth mask block?

There is a recent study on this which suggested that the cloth masks are considerably less effective than surgical masks.

OTOH, we had the whole other thread on the DANMASK study which showed that when you recommend people wear surgical masks and they report they wore them, that such intervention does NOT significantly reduce the likelihood of them being infected by COVID-19. Perhaps a small effect, 15%, but that was very likely due to chance.

Putting these two together would strongly suggest that cloth masks would also fail to have a significant effect at reducing infection for the wearer.

 

[PeterNSteinmetz]

It is still available, without retraction watermark, on The Wayback Machine.

There certainly is a strong contrast here between the links from the CDC above and the Briand seminar reported at this link.

The CDC estimates appear to depend on some established though complex model fitting.

To really make sense of this I think would require obtaining the datasets and analyzing it independently. May have to try that.

 

[PeterNSteinmetz]

And what’s your point? They still managed to avoid the worst of the virus. The paper ignored those countries completely in their study.

Perhaps read the study to see why they chose those countries?