Post #723: Updated: Aerosols, drinking, and bars. Some clear policy advice, for a change.

Posted on June 19, 2020

Edit:  OK, this one now seems to be out in mainstream news outlets.  In Mississippi, they’re blaming fraternity rush.  In Florida, it’s described as a “radical shift” in demographic of new COVID-19 cases. “They also tell us that younger South Carolinians are not taking social distancing seriously.”  “Our average age last week of people that were positive was age 30, the average age of people getting tested was 47,”   And so on.  So it looks like the smarter people have now figured this out, at least as far as age goes.

Source:  Notes in red were added for this posting.  Original graph taken from: Xie X, Li Y, Sun H, Liu L. Exhaled droplets due to talking and coughing. J R Soc Interface. 2009;6 Suppl 6(Suppl 6):S703-S714. doi:10.1098/rsif.2009.0388.focus.

What you see above is a decade-old footnote in the scholarly literature on the production of droplets during human speech.  Yet, I think it may be relevant to the current spike of coronavirus cases in several states.

Because, as I read it, standing shoulder-to-shoulder in a crowded bar and talking at the top of your lungs is NOT the stupidest possible thing you could do.  In the current era, I mean, in order to spread COVID-19.  You need to be sipping a sweet drink at that same time.  That’s the stupidest thing you could do.

Detail follows.

Drinking liquids increases droplet production, drinking sweet liquids also increases aerosol-sized droplet production.

Source:  Screenshot taken from video on

You may or may not have seen the gross-yet-hilarious Mythbusters episode where they tested ways to stifle a sneeze.  Briefly, they drank a little bit of liquid containing dye, sneezed, then tracked where the droplets landed.  While filming the sneezes with a high-speed camera.

That approach — dye — was more-or-less the standard scientific method for studying droplet transmission for many years.  Only recently have other methods (lasers and scanners) become available as an alternative.

Interestingly, there was one scholarly article that not only studied the droplets, they tested the methodology.  They measured droplet emissions from speech and from coughing, with and without dye, and with and without dye sweetened with sugar.

As it turns out, drinking a little bit of liquid — and in particular, sweetened liquid — greatly increased droplet emissions and aerosol (under-5-micron-droplet) emissions from speech.

Here’s that research.  In essence, it’s Mythbusters, minus the fun, with greater precision.  So it’s a bit of dry reading.  But what matters are the results (shown above, and here):

Source:  Cited at top of posting.

I’ve flagged the key bar on that graph with a red circle. Based on this one piece of research, drinking something sweet, just before you talk, greatly increases the number of aerosol-sized droplets that you produce.  And it’s those aerosol-sized droplets that can hang in the air, travel well beyond “social distancing” distances, and result in a single individual infecting many, in a crowded situation.

Not shown here, but also in that research (bottom of Table 2), drinking anything at all roughly triples the number of droplets produced.  For their particular speech test, they give average droplet counts of 760 with no fluids, and over 2000 after drinking a bit of fluid dye.  (That overstates the total mass (amount) of additional droplets, but total mass did increase materially when liquids were ingested before talking.)

Now let’s see if you can put that together the way I do.

By now, I think we all realize that standing around in a crowded room, talking loudly, is an invitation to spread COVID-19.  For that, it doesn’t matter if the room is a church, indoor arena, restaurant, or bar.  All it takes is one infected individual who is a “super-emitter” of aerosols, plus-or-minus some poor ventilation, and you can end up with a cluster of infections.

I also think that most of you have figured out that masks work, some, to prevent spread of disease.  And so, doing the above, without wearing masks, would be even stupider from the standpoint of spreading COVID-19.  But at that point, the church (and hopefully the indoor arena) drop out of the competition for places you want to avoid.  Presumably, every sensible mainstream church is going to require masks for attending services in person.

Now stand around drinking, as you talk loudly.  At that point, the total number of droplets emitted by the talkers triples, based on the research above.  So whatever dose of droplets you were getting before, you’re now getting three times that.  So a maskless, noisy bar is yet another notch higher on the stupidity scale.

But if you really want to max it out, you need to have a sweet drink in your hand.  That not only increases total droplet production, but amps up the rate at which the attendees produce aerosol-sized droplets, the ones with the long hang time.  At that point, short of close bodily contact, you’ve come about as close as you can get to maximizing the likelihood of spread of COVID-19 infection.

An important limitation of this research is that it only tested those two substances.  For all we know, other common substances could have that effect as well.  In other words, we just happen, by chance, to know about sugar, due to this one study.  Nobody has ever actually studied this, as a topic, i.e., how much does intake of various foods or liquids increase aerosol emissions from speech.  So don’t fall into the trap of thinking that sugar is the only problem.  It’s the only one we know about.

Why does this matter?

Briefly, and informally, it sure looks to me like the new cases are skewing very young in the areas where new COVID-19 cases are spiking.  This is totally armchair epidemiology, without a shred of hard work to back it up.  But I read one news article that stated that the bulk of new hospitalizations were individuals under age 30.

And I’m not the only one with this impression.  The Governor of Texas just kind of flat-out said, in effect, never trust anyone under 30.  Well, no, what he actually was quoted as saying was:

 "certain counties where a majority of the people who are testing positive ... are under the age of 30, and this typically results from people going to bar-type settings."

It’s hard to say just how unusual it would be to have the majority of cases be under age 30.  In Virginia, adults under 30 account for just 16% of all cases.

And, no surprise, the Governor of Florida says that new cases there are concentrated among younger individuals.  Those people are more likely to survive than the elderly (so the gist was that the spike in cases in Florida is not as serious as it might at first seem.)  He doesn’t go on to blame bars, but hey, it’s Florida.

And hey, what a coincidence, the new cases in California are much younger than average as well.  (Or, read the same thing here.)  They cite 44% of new cases being under the age of 35.  That is again hugely different from the norms to this point.

For Arizona, I have no such solid evidence, just a video of a bar packed with maskless young people.

Anyway, I think I’m seeing a trend.  With all due respect for the uncertainties of this, and for the shallowness of my analysis, I think I’m willing to call this one.  The current spike of cases, in a handful of states, is from young people getting frisky and going out to bars.  And similar.


Why this matters, right here, right now

In all seriousness, I think there’s a real lesson here for the Commonwealth, and for colleges and universities planning to re-open throughout Virginia.  If what I said in the section just above is in fact true, then you have to treat bars (frat parties, mixers, and so on) as a venue that is completely different from, and far riskier than, other places that seem superficially similar.  If you are properly focused on minimizing infection risks, then you need to clamp down on those locations.

So, let me sum up.  I’ve already said that going to church is not the same as going to (e.g.) Home Depot.  That gets back to the Japanese concept of avoiding the three Cs (Post #720).  And by now, we’re all aware that talking loudly increases transmission risks, because it increases aerosol emissions.  Most of us, I think, are well aware that masks help prevent transmission of disease.

And now we can add one more little tidbit:   Drinking while talking increases droplet emissions three-fold, and drinking something sweet while talking increases dangerous aerosol droplets enormously.

And so, step by step, we’re figuring out a kind of hierarchy of risk in indoor spaces.  Not formally, but possibly usefully. 

At one end, we have the great outdoors.  It may be possible to transmit COVID-19 in that setting, but all the epidemiology suggests that’s a pretty rare occurrence.  (Though that may have been tested by the recent protests — we’ll eventually find out, or not.)

Then, indoors, we have something like the following hierarchy, from least to most risky:

  • Home Depot:  Cavernous warehouse space, broad aisles, widely-spaced customers, mandatory mask use, brief interaction with cashier, if at all.  I guess we can put Costco and similar in that category.
  • Other retail, including grocery:  Smaller spaces, more customers.
  • Church:  This depends on the church, but no congregation responses, and no singing, and mask use, all would indicate a relatively safer situation.  Drop any of those restrictions, and it gets proportionately more risky.
  • Restaurants, indoor:  By its nature, you can’t wear a mask.  And people talk during dinner.  And drink.  So this appears to up the transmission risk considerably.
  • Bars, indoors:  Loud speech.  Much of the point is to drink.  Masks likely will not be worn.  They tend to be noisy, so people talk loudly.  Based on all the criteria I know about, for how aerosol transmission of COVID-19 works, this has to be the riskiest situation in terms of disease transmission.

I had a request to rate coffee shops as a venue.  I’m guessing, for the typical coffee shop, I’d rate them as being safer than indoor restaurants.   The reason is that they tend to be quiet, and patrons can stay masked except when sipping their coffee.  And while the same is true of mask use while sipping a beer, ah, how can I say this?  For many reasons, I expect the average bar patron to be far less compliant with rules about mask use.  Among other things, you don’t typically go to a coffee shop to (ahem) see and be seen (i.e., try to hook up with somebody).  It’s just a generally calmer scene than a typical bar would be.  And I think the ability to stay masked most of the time makes it somewhat safer than a restaurant.

I noted more than a month ago that in all of the various state re-opening plans, everybody seemed to hate bars.  In most places, they were the last venues allowed to open.  I now think there was a lot of wisdom in that, intentional or otherwise.

Addendum:  When the data talk, you need to listen.

I’ve harped on the fact that I can’t make head or tail out of the patterns of spread of this disease (Post #716).  It’s just too willy-nilly, in most cases.  And I explained why that was, in that post.  Further, a) I’m not alone and b) I have decades of experience in analyzing disease incidence (new cases) and prevalence (all cases) data.

But my experience has also been that sometimes, if you just keep looking at it, a clear pattern will emerge.  If the data are cooperative, and you are lucky enough. You just have to pick up on the hints that the data offer you.  In short, when the data talk, you need to listen.  Even if they only whisper, at first.

I think this is one of those times.  Because what we see, in the skewing of diagnosed cases toward younger people, must be just a whisper of the total cases.  Why?  This disease has vastly more impact, on average, on older people.  By contrast, younger people more typically just shrug it off, not much worse than common flu.  And so, for every 100 old people infected, maybe 50 will show up at the doctor’s office or the ER and get diagnosed.  But for every 100 20-somethings infected, maybe 5 or 10 might have to do that.

Again, with my sloppy armchair epidemiology, if half the diagnosed new cases are persons under 30, then … 90% of total new cases must be persons under 30.  I.e., just taking my made-up diagnosis rates above, a mix of 50 young and 50 old, diagnosed, means that the underlying mix of infections was 1000 young and 100 old.

And so, while the data are whispering to me about bars, somewhere in the background, I can faintly hear some data shouting about the overall infection rate.  The age skewing of the diagnosed cases has to be just a faint echo of the true picture.  And that matters because even if the younger individuals themselves don’t get very sick, they can spread it to those who might.  So you may see a second, larger wave, with worse consequences, and a greater mix of ages, if this first post-re-opening wave of younger, infected individuals is as large as it would have to be, to generate the number of diagnosed cases being observed.

Presumably, an epidemiologist with access to the data could do the proper calculation, and start to tell use just how widespread COVID-19 infections are now, among younger individuals, in these states.  I have a feeling that’s going to be a really unpleasant reality, and I hope that someone with access to the detailed data can do that calculation sooner rather than later.

Canary in a coal mine?  Tripwire?  I’m not sure what the metaphor is.  But if half the diagnosed cases are under 30, I can’t even guess what fraction of the actual infections are in the under-30 population.

Finally, this points out one of the shortcomings of epidemiology that is rooted in traditional academic research.  I can come right out and point the finger at bars.  But I’m nobody.  By contrast, if I worked for the Commonwealth, and had an official position, I probably could not make a simple declarative statement that bars appear to be the problem.  I would be held to a “research” standard of proof, which, frankly, you might have a few weeks or months from now ( as contact tracing progresses), or you might not ever have.  In the course of a pandemic, those weeks or months matter greatly.  And so, requiring such hard proof, before the government acts, probably is not an optimal approach from the standpoint of protecting the public health.  But it’s the slow and deliberate way that we typically require our government entities to act.