COVID-19: Masks and ventilation more effective than social distancing

[dmspilot]

Estimating COVID-19 exposure in a classroom setting: A comparison between mathematical and numerical models featured
Physics of Fluids 33, 021904 (2021); https://doi.org/10.1063/5.0040755

"The risk from airborne SARS-CoV-2 exposure does not appear to be strongly correlated with the distance, and many of the peak exposures were observed outside of physical-distancing guidelines. This indicates that mask mandates, well designed HVAC systems, and the combination of exposure time with number of occupants are of increased importance compared to physical distancing."

 

[PeterNSteinmetz]

A pure mathematical modeling study. And the results appear to be at variance with the observational studies in the WHO meta-analysis on distancing.

This is consistent with the prior data on the effects of masking In the general sense that the in vitro studies suggest masking should have an effect (and this is an in silico study). The observational studies give a variety of answers. And the randomized controlled studies, to the extent they bear on this question, fail to show a significant effect of the general public wearing cloth masks on slowing the spread of COVID-19 and some evidence of possible significant harm.

It really is too bad we don’t have a randomized trial of the proposed source effect of cloth masks yet. It is hard to do, but probably could be done in a controlled environment like the study in the Marine recruits.

In general in biomedical clinical questions like this, when there are questions and mixed evidence, in weighing the evidence one rank evidence in categories like randomized clinical trials > observational studies > in-vitro studies. The reason for that is that living biological beings have a lot of things going on inside them and engage in a lot of behaviors which can undo the observed lab effects.

 

[PeterNSteinmetz]

"The risk from airborne SARS-CoV-2 exposure does not appear to be strongly correlated with the distance, and many of the peak exposures were observed outside of physical-distancing guidelines. This indicates that mask mandates, well designed HVAC systems, and the combination of exposure time with number of occupants are of increased importance compared to physical distancing."

This is extremely selective quotation and title choice. This study did not contrast the mask and non-mask case so they cannot make any claim about the efficacy of masks, even in this simulation, or what component of the combination was efficacious
They studied this combination and then are making a conclusion regarding the effect of the combination as a whole in silico.

They also made a rather optimistic assumption about the effect of surgical masks, namely a 44% reduction in the likelihood of infection of the wearer.

 

[PeterNSteinmetz]

You guys know the joke that concludes with “consider a spherical cow”?

I love these types of physical models, but they do have their limits.

 

[PeterNSteinmetz]

There are still an average of 784 deaths per day. Remember when there was an over-under poll and the 50/50 was 25,000 total deaths? We are still at nearly that number every month.

The thing is, COVID-19 appears to have displaced the normal seasonal flu. Typically there were 30-50k per year. So about 109 per day when averaged over the whole year.

It looks like we are going to have had about 500-600k excess deaths this last year.

 

[PeterNSteinmetz]

I'm sorry that you feel my quotation and title were not "extremely selective" in the same manner that yours would have been.

Well, Mattias Fridgen, I think it would be better if the titles reflect what can actually be concluded from the results in the article, rather than from a selective quotation of the discussion alone, where the authors are often summarizing in a somewhat looser way.

Your assertion is completely incorrect and not even close to what is actually written in the paper. First, the 44% is the reduction of inhaled and exhaled particles, not the likelihood of infection of the person wearing the mask. Second, the 44% is a pessimistic number based on wearing a mask with a gap, as sourced from the paper "Aerosol Filtration Efficiency of Common Fabrics Used in Respiratory Cloth Masks." There are a few materials and configurations less effective, but most are more effective than 44%. Therefore it does not seem proper to characterize that as overly optimistic.

Here you are being prosecutorial again in your interpretation of language to try and support some strained attack and distract from the actual scientific problems with this article. I do owe you an apology though, since I thought you had said you were an attorney, but you are not. So I guess that doesn't explain these constant hostile nit picks on language. You do raise a good point about what should be assumed about mask efficiency in real life.

The important question in using 44% in this type of model is how well the mask fit works to prevent the exhalation of virions in an actual real-life, not an in-vitro, setting. Konda et al (cited by Foster & Kinzel as #23) used an in-vitro model to measure filtration efficiency and what was only a very rough approximation of gaps, so their numbers are very loose approximations of what might happen in the real world. Their own statement is "Our findings indicate that leakages around the mask area can degrade efficiencies by ∼50% or more, pointing out the importance of “fit”.

So this is no doubt an important issue. As you note, in-vitro these masks do pretty well, but in real life the leaks may be greater than anticipated. We have some experience with randomized trials with surgical masks and whether they protect the wearer. The DANMASK-19 study found there was no significant effect and considered as a point estimate, would suggest that they may create a 15% reduction in the likelihood of infection. Since the likelihood of being infected is probably proportional to the likelihood of having inhaled or ingested a sufficient dose, it is very consistent to assume that a reduction to 15% virions exhaled would correspond to a 15% reduction in the likelihood of others being infected. From this perspective, a 44% for a reduction in the actual virions exhaled into the environment is inconsistent with the randomized studies which are available.

Arguably, the authors should have explored the parameter space around this fraction more broadly given the uncertainty of that 44%. And I certainly think that given the available data it is not far off to say they were perhaps being a bit optimistic in their use of this, but it was not a bad first shot I suppose. That was not the main focus of their paper.

Overall then, Foster & Kindle is a fairly basic model which failed to properly explore the parameter space and so even as an in-vitro study is quite limited in its power to tell us about the relative effects of distancing and a combination of masks and additional ventilation. And it absolutely cannot say anything about the contribution of masks versus ventilation and their possible effect on the spread of Covid-19.

 

[PeterNSteinmetz]

My take is that social distancing, hygiene, increased flu vaccine adoption, and mask wearing displaced the seasonal flu, not COVID viruses themselves. It is not like COVID and Influenza are fighting an epic battle on a microscopic scale. The only difference between this flu season and the previous ones are the specific safety measures that people are taking.

Influenza is ~2 times less virulent than COVID-19, so these safety measures are even more effective at preventing it. Kinda also proves that the safety measures are also an important tool in preventing COVID transmission as they are both primarily airborne viruses (as finally acknowledged for COVID).

That's one take. I don't think the available data support this conclusion really, but it is not something we know with a high degree of confidence at this point. The biggest difference is that a more effective pathogen, SARS-COV-2, is in the mix now. Likely it simply infected all the susceptible hosts.

As we have discussed at length, very mixed data on whether the public wearing cloth masks decreases the spread of Covid-19. I think the evidence for social distancing working, as in the WHO meta-analysis, is fairly strong. The vaccines pretty clearly appear to be ~90-95% effective though they have been deployed fairly late in the pandemic relative to the drop in cases observed near the start of the year.

OTOH, we know with fair confidence that the coercive lockdowns had no detectable effect of mortality. And curiously, 330 million x a 0.25% IFR x 65% for herd immunity is just about 550,000 deaths, which is where we are now roughly as the pandemic seems to be dying off.

My overall conclusion would be that the pandemic has run and will be finishing up its course. These things usually last a season or two. We will likely know more in a few years when the various papers and analyses have been completed.

 

[PeterNSteinmetz]

No, we don't know that. New Zealand had very few deaths. Same with Australia- they had what, maybe 1000 deaths? VietNam had what- maybe 1 death? China is completely open now, their cases seem to be from people crossing the border. All my friends there tell me things are nearly back to normal there, enough that their vaccine roll-out is slow because people just don't see the need.

As I noted above, those type of observational studies, with a large number of potential confounds, are fairly weak evidence. In this case, the cross-national comparison of an island nation in the middle of the Pacific and a small population, is not a very good comparison with other major Western nations with large land borders. If you are willing to be a tyrannical regime like China, you can potentially decrease all sorts of behaviors, like drug usage, but at what cost?

The reason I say a fair degree of confidence is that the evidence from the beginning of the pandemic in the US, comparing between states, is very clear there was no statistically detectable effect of lockdowns decreasing Covid-19 mortality. I have analyzed the data myself and can show it to you if you like.

There actually was an effect, but not in the direction one would expect, when the development rate of mortality is considered over time. The states with lockdowns actually had a higher rate of development of mortality, not less. One could try and argue on that basis that the lockdowns caused excess mortality, but I am inclined to think there is another confounding factor. In any case, no evidence at all in that state by state comparison that lockdowns decreased Covid-19 mortality.

Wearing cloth masks also help, they certainly help catch drops like that guy below.

Yes, they do likely decrease those large type of droplets. But that does not imply that the general public wearing cloth masks decreases the spread of Covid-19 because there are a lot of other factors. The only randomized trials available say that such usage does not significantly decrease the likelihood of the wearer of catching Covid-19. Prior randomized studies found the same thing with the seasonal flu. And one randomized study of healthcare workers wearing cloth masks suggested they actually had a higher rate of influenza like illnesses.

Those type of strongest data, the randomized clinical trials, have to be weighed appropriately against the in-vitro testing, which certainly suggests masking should work, and observational studies, which end up on both sides.

 

[PeterNSteinmetz]

"The effective reproductive number, Re of New Zealand’s largest cluster decreased from 7 to 0.2 within the first week of lockdown."
https://www.acpjournals.org/doi/10.7326/M20-6817

Interestingly, the experience in the US with estimated reproduction numbers as a function of dates from lockdown, was not at all the same. Thus one of the difficulties with trans-national comparisons.

Of course, these numbers are considerably less reliable than mortality data, given the variations in testing levels and changing reporting requirements.

 

[PeterNSteinmetz]

You do know that you can be infected with more than one virus at a time, right? It is not like you get COVID-19 and now there are no more cells to infect with influenza. And you also know that once you get COVID-19, you are not immune to the seasonal flu, right? So people who got COVID-19 could still catch the flu before or after their COVID infection.

So how has it "Likely it simply infected all the susceptible hosts." and that somehow prevent nearly all of the seasonal influenza infections? Please explain.

I am not aware of any data regarding the co-infection of susceptible individuals with both COVID-19 and seasonal flus. Do you?

It is very curious how much the apparent rate of seasonal flu infection has dropped. Given that in the the prior randomized studies, wearing surgical masks did not decrease the likelihood of being infected, it seems to me unlikely that masks suddenly had some large effect this season.

I can see several ways that Covid-19 infection in susceptible people would decrease the likelihood of seasonal flu being reported. For one thing, once sick with Covid-19 one is likely more isolated from others who have the seasonal flu. It might also be that people who get Covid-19 and shortly in time have the seasonal flu, simply don't end up showing up at the doctor to report it.

I really don't know, which is why I would say this is a likely supposition, but I would be happy to see real data on the subject.

 

[PeterNSteinmetz]

So you think the title should [sic] come from the conclusion section of the paper? That is where I got it. Did you read it?

Mattias, isn't that obvious, come on, give me a break, ok?

I think the title best reflects the actual conclusion that can be drawn from an objective view of the results of the paper. As I have discussed before, the conclusions are often a bit looser. In this case, the post title is not strictly incorrect, I would agree, because of the conclusion of the 'and'. What the authors contrasted in their model does not allow them to say anything about masks or ventilation separately versus distancing.

If we want to be strict and prosecutorial in our interpretation of titles, as you are wont to do, I note that the use of 'and' by the authors and in your title is ambiguous. It could mean that both masks and ventilation, considered separately, are better than distancing. Or it could mean that the combination of masks and ventilation is better than distancing. The authors statement clearly had the second meaning in the context of the whole paper, but does not in the separated title which you chose. Thus it could be improved. Now, I am not normally a big fan of this sort of prosecutorial inquisition into people's wording. I prefer to focus on the actual meaning of the data and analysis. Let's focus on that, shall we?

And in this study, the masks were assumed to be leaky. So I don't know how you think this is a flaw in the paper.

The degree of leakiness of the masks likely has a large effect on the magnitude of the effects observed, as stated by the paper cited by the authors. Yet the authors make no attempt to vary this parameter and see what effect changing it would have on their conclusions. That is normally a standard part of testing the validity of models, it is called studying parameter sensitivity. That is a flaw which limits the strength of conclusions which can be drawn from this paper. Not to say it is pointless or worthless, just rather limited.

The DANMASK-19 study concluded masks did not decrease infection rate of wearers by at least 50%, and was inconclusive for a lesser decrease. Since leaky surgical masks reduce particle transmission by 44%, and 44 is less than 50, I would call 44% "consistent" rather than "inconsistent."

I am sorry but we have discussed this before and it really does represent a fundamental error in statistical interpretation. The fact that a study failed to find a significant effect at a p<0.05 level with an 80% power to detect a 50% of greater effect does NOT imply what you say here, that is was "inconclusive for a lesser decrease".

If you want to interpret the data of the DANMASK-19 study despite the fact that any observed difference was 95% likely due to chance, the proper interpretation as a point estimate of the effect is that there was a 15% reduction in the rate of infection. And that there was a roughly 25% chance there was actually an increase. You can't just take the extreme end of a 95% confidence interval like that and try to claim something was "consistent". If anything, the confidence interval says that 44% is rather unlikely. We can compute the likelihood of a 44% reduction given the data from DANMASK-19 if you like, but it is most certainly a lot less than 15%.

 

[PeterNSteinmetz]

They used the data for the "leaky" variant of surgical masks. Do you think using the filtration rate of non-leaky masks would weaken their conclusion? Again, I don't see how your criticism even makes sense when the study authors chose a more pessimistic filtration rate.

In brief answer to your question, no. Let's think about the effects of that, interesting point. I would assume that more leaky masks would cause the effects of ventilation to strengthen, in other words, more virions in the environment imply that ventilation is a more important thing. So using less leaky masks as an assumption would decrease the apparent effectiveness of ventilation. So in that sense, I guess I would agree with you that using less leaky masks strengthens their conclusion relative to the combination being more effective than distancing. But I still think 44% is an optimistic estimate, in the sense of more effective than masks are likely to be. I do see how you meant 'pessimistic' in a different sense than I was using its normal antonym; I didn't catch that earlier so sorry for any confusion.

But my primary criticism is that the authors did not study parameter variation of this likely important parameter at all. They just presented results for one value. This is the great strength of computer models, one can run them ad infinitum with variations. In modeling papers one normally has graphs of the effects as a function of the different parameters and maybe even statistical testing against assumed prior distributions of these parameters based on other data. If they had done so, we would understand directly the effects of leakiness on the strength of their conclusions. It is possible it is the other way around, but I don't think so.

This could be a much stronger and rather interesting study. It addresses an interesting question of the interplay of masking, distancing, and ventilation. Thanks for pointing it out in any case.