Effectiveness of One-way Masking in a High-risk Setting

Ingu Yun
12 min readNov 5, 2022

November 5, 2022 — by Ingu Yun, MD

Country-western dancing at the Sundance Stompede
The 2022 Sundance Stompede country-western dance in San Francisco

How effective is one-way masking against Covid, really?

Mask requirements have all but disappeared in the United States (and in many places never took hold), but Covid is not done with us. While many folks have ditched their masks in favor of accepting the risks, some of us have not, making it feel a bit treacherous to venture out.

Recent articles in the New York Times and The Atlantic suggest that one-way masking is effective, but cite studies that are mostly based on modelling or simulations that may not reflect real-life situations. What I want to know is this: as someone who wants to stay safe, exactly how much does my mask protect me in the real world, when no one else is wearing one?

In May 2022, I had an opportunity to investigate this question, when I helped organize an event that carried a high risk of Covid transmission. The Sundance Stompede is an annual four-day country-western dance convention in San Francisco that attracts hundreds of people from all over the U.S., Canada, and beyond. [video] The event caters to members of the LGBT community but is welcoming to and attended by all. A large amount of time is spent partner dancing at close range with many different people in a ballroom with hundreds of other people doing the same thing. The timing of this year’s event couldn’t have been worse: right at the peak of the BA.5 surge in San Francisco, which wastewater data suggest was possibly even worse than the first Omicron surge in January.

One week following the Sundance Stompede, I sent a survey to the full-weekend registrants. I was curious to see whether any patterns were evident with mask behavior and other factors. I’ll point out from the start that this was not intended to be a rigorous study — and there are enough limitations already inherent in analyzing a retrospective survey — so I’ll be the first to say that any conclusions need to be framed within this context. However, the results seemed so compelling, that I felt they were important enough to share, if only to add to our body of knowledge and stimulate further study.

The findings that were most interesting:

(1) Overall, 37 out of 265 (14.0%) attendees reported a positive Covid test during the week following the Stompede.

(2) Out of 70 attendees who reported wearing a mask less than 10% of the time, 11 (15.7%) tested positive for Covid. This group was considered the control group in the subsequent analysis.

(3) Out of 66 attendees who wore a high-quality mask (N95, KN95, or KF94) greater than 90% of the time, only one (1.5%) tested positive for Covid. [P=0.01]

(4) Curiously, out of 15 attendees who wore a cloth or fabric mask greater than 90% of the time, 9 (60%) tested positive for Covid. [P=0.001]

METHODS

The Sundance Stompede took place in San Francisco, May 26 to 29, 2022. Each night featured dancing (two-step, waltz, swing, and line dancing) in different venues, with attendance each night ranging from 300 to 450. During the day educational dance workshops were offered, spread over five different rooms. There were also other social events and meetings, including a Sunday brunch. 339 attendees were registered for the entire weekend, with the remaining attendees paying at the door for individual events.

All attendees were required to be fully vaccinated against Covid-19. Face masks were not required, but strongly advised. All full-weekend registrants were required to show photographic evidence of a same-day negative Covid-19 antigen test at check-in. They were also encouraged, but not required, to test again half-way through the weekend, and during the week following the event. The Thursday and Sunday night venues were relatively well-ventilated (CO2 < 900 ppm), but the Friday and Saturday night venues were not (CO2 > 2000 ppm). We had five commercial high-powered HEPA air purifiers (CADR 559 cfm) and 12 Corsi-Rosenthal cubes (estimated CADR 240 cfm) that we moved from venue to venue.

One week after the Stompede, an eight-question online survey was sent to each full-weekend registrant who had checked in and attended at least one of the weekend’s events. A unique PIN code was provided to each registrant so responses could be anonymous, while ensuring we received just one response from each attendee. Survey questions asked about which events were attended, vaccination booster status, mask behavior and type, and Covid symptoms and test results during the week following the Stompede. We did not ask for any demographic information. (I would estimate 70% men, 25% women, and 5% non-binary; and a mean age of 50, ranging from 20s to 80s resembling a normally-distributed bell curve, perhaps slightly skewed younger.)

Attendees were designated as contracting Covid-19 only if a positive antigen or PCR test was reported. Nine attendees reported symptoms but did not test positive for Covid-19, including some who followed up with PCR testing. I chose to designate these nine attendees as Covid-negative in my primary analysis. Some of these attendees offered other explanations for their symptoms, such as allergies. And interestingly, wastewater data suggest that influenza was circulating in San Francisco at this time.

RESULTS

Out of 339 registrants who attended at least one event and were invited to complete the survey, we received 269 responses. Four surveys were rejected due to irreconcilable irregularities, leaving 265 (78.2% out of 339) valid submissions. All respondents were fully vaccinated, and 261 (98.5% out of 265) had received at least one booster. (The LGBT community leans liberal and health-aware!)

Out of the 265 respondents,

• 70 (26.4%) “rarely or never” wore a mask (defined as <10% of the time). This represents our control group in the subsequent analysis.

• 67 (25.2%) “sometimes” wore a mask (10–50% of the time),

• 24 (9.1%) wore a mask “most” of the time (50–90% of the time), and

• 104 (39.2%) “nearly always” wore a mask (>90% of the time).

Positive Covid-19 test results within seven days of the conclusion of the convention were reported by 37 attendees (14.0%). Compared to the control group, the proportion of attendees testing positive was not statistically significantly different for any of the categories of mask-wearers [P=0.89; Fisher’s exact test], including the “nearly always” group:

• 11/70 (15.7%) in the “rarely or never” control group tested positive.

• 9/67 (13.4%) in the “sometimes” group tested positive.

• 4/24 (16.7%) in the “mostly” group tested positive.

• 13/104 (12.5%) in the “nearly always” group tested positive.

However, when breaking down the “nearly always” group by mask type, the results become much more interesting:

• Out of 66 attendees who nearly always wore a high-quality mask (N95, KN95, or KF94), only one (1.5%) tested positive [P=0.01].

• Out of 23 attendees who nearly always wore a surgical mask, 3 (13%) tested positive [P=0.75].

• Out of 15 attendees who nearly always wore a cloth or fabric mask, 9 [60%] tested positive [P=0.001].

[Need a review of “P-value?” Here’s a quick primer written for non-statisticians. Briefly, the smaller the P-value, the less likely the result would be from chance only. A threshold of <0.05 is often used to indicate statistical significance.]

I refer to these groups respectively as the HQ mask group, surgical mask group, and cloth mask group, respectively, in the subsequent discussion.

In sum: high-quality masks seemingly offered highly significant protection from Covid-19. There was no significant protection found from surgical masks. And most unexpectedly, wearing a cloth mask nearly all the time appears to have significantly increased the likelihood of contracting Covid.

DISCUSSION

Studies that look at masking in the real world are scarce, and the few that I have found have generally reviewed events where masks were required (i.e. two-way masking), have taken place in a specific setting that may not be as relevant to me (e.g. hospitals), or have had other limitations. Because of ethical and logistical concerns, randomized controlled studies are difficult, if not impossible, to create. Observational studies, while less compelling, may be the best we can do. (Two recent interesting survey-based studies on mask efficacy include one of attendees of a national surgical society meeting and another of California residents who received a PCR test in 2021.)

It is generally accepted that high-quality masks are quite effective at protecting the wearer from contracting Covid-19. The results from our survey suggest that this protection is extremely high and nearly complete in a high-risk one-way masking environment. Now, it’s quite possible — even likely — that those attendees who wore high-quality masks are also the types of people who would take other measures to reduce their risk of exposure. For example, they might choose to dance only with other masked attendees or limit the number of their dance partners. They might even choose to avoid certain higher-risk events. I will note, however, that the Sunday brunch, which was one of the highest-risk events (because everyone had to remove their mask to eat), was attended by 38/70 (54%) of folks in the no-mask control group, and 31/66 (47%) of folks in the HQ mask group, which shows a slight tendency towards avoidance, but is not statistically significant [P=0.39].

The numbers in the surgical mask and cloth mask groups may be too small to draw any firm conclusions, but it does appear the surgical masks were not nearly so protective, if they were protective at all. I believe that this is one area where modelling studies and real-life studies may diverge, because I observe that surgical masks are almost always worn with a poor fit, typically with a large gap on either side. It’s not surprising to me that surgical masks wouldn’t offer much protection in the real world, compared to what might be evident in the laboratory.

I am quite intrigued by the large proportion of attendees wearing a cloth mask “most” of the time who contracted Covid. With only 15 attendees in this group, I am somewhat inclined to dismiss the finding and chalk it up to a statistical fluke. But on the other hand, this result is just so striking, with 9 out of the 15 testing positive! Was the cloth mask group’s behavior simply much riskier than the no-mask group? That just doesn’t seem likely to me. I think that generally the no-mask group was not taking any precautions to begin with, and so it would be difficult for behavior to get much riskier than that. I’ve been wracking my brain trying to think of any other possible explanation for a negative protective effect of cloth masks. Here’s the best hypothesis I’ve come up with. With a room full of dancers, our venues can get quite hot, and, combined with the aerobic activity, cloth masks can get soaking wet. Could it be that a wet cloth mask becomes “stickier” for the virus, and the mask ends up actually collecting and harboring Covid-19, leading to increased exposure and a higher risk of transmission? OK — I know that’s wild and probably incorrect. But at the very least, this unexpected finding gives me pause about cloth masks, at least when doing an aerobic activity in a sweaty environment. (I feel it important to point out that even if this finding turns out to be valid, it says nothing about the ability of a cloth mask to trap virus exhaled by someone with Covid. You’ll recall that the original purpose of wearing cloth masks was a public health measure intended to protect everyone except the wearer. )

The findings reported here are subject to multiple limitations, many of which have been referenced already. The survey was not designed as a study or intended to be analyzed statistically, and is not prospective, random, blind, or rigorously controlled. • Surveys are inherently subject to inaccurate responses, through social desirability and recall biases, and other factors. • The survey response rate, while quite high at 78%, is not an indication of the validity of the results, which is more dependent on selection bias, which we are not able to assess. • We don’t know what other preventative behaviors attendees may have exhibited to reduce their exposure to Covid. • We did not ask about any recent diagnosis of (and recovery from) Covid, which may have prompted some attendees to forgo masking. • For reasons given earlier, I elected to exclude attendees with positive symptoms but negative tests as having post-event Covid. However, if these subjects had been included as having contracted Covid, the results would not have been substantially different: no-mask control 17/70 (24.2%); HQ mask group 2/66 (3.0%, P=0.001); surgical mask group 4/23 (17.3%, P=0.49); cloth mask group 9/15 (60.0%, P=0.019). • 35 attendees did not test themselves during the week following the Stompede. None of these attendees reported symptoms, but it is possible we missed an asymptomatic case. • Even among those who did test during the week following the Stompede, self-administered antigen tests may have missed some positive cases. (74 attendees tested themselves once during this period, 106 tested 2–3 times, and 50 tested 4 or more times.) • Some reported cases of Covid may not have been contracted during the event, but in other settings, such as when dining out or traveling to and from the Stompede. We did not ask about mask use outside of the event. • Stratification by mask type resulted in small subgroups, making unexpected findings difficult to interpret. • The findings are specific to a particular population and environment, and may not generalize to other situations.

As an aside, this survey turns out to be an excellent example of how different conclusions can be derived from the same data. If I had only analyzed the major groups of mask-wearers based on the amount of time they wore their masks, we might have concluded that masks aren’t significantly helpful. But by breaking the data down into subgroups by mask type, we show that high-quality masks are extremely effective. This is a cautionary example of data that can be manipulated to “prove” two completely different outcomes. Most of us don’t have time or knowledge to delve into the details of published studies. At the very least we should remember that it’s important not to draw firm conclusions on the basis of any single study, and to look for the preponderance of evidence across many studies. In this age of social media, immediate gratification, and confirmation bias, this basic concept often gets completely lost.

WHAT DOES IT ALL MEAN? A PERSONAL PERSPECTIVE.

While I would caution anyone from drawing firm conclusions based on a single report that has many limitations, I am still encouraged by these results that seem to corroborate other evidence that high-quality masks are highly effective in a one-way masking environment, even in high-risk settings.

My personal take: The prevalence of Covid in San Francisco is currently much better than it was at its peak, but as of October 2022, it is still not low enough for me to relax. Unfortunately we’re not able to rely on official Covid case numbers to assess how we are doing, because self-testing results generally go unreported. It’s likely that actual case counts are two to five times higher than the official numbers.

We all have our different relationships with Covid. Many folks are done with worrying about Covid, while others remain in high avoidance mode. I personally have really enjoyed not getting sick over the past few years! At this point I would prefer to avoid even a mild case of Covid, or any other respiratory illness for that matter, with the resulting interruption of my busy life. We are always weighing the risks and benefits of our behaviors. For me, wearing a mask is not a significant burden, and the benefit of wearing a mask continues to outweigh any inconvenience.

Since seeing the results of this survey, I have had much more confidence to feel essentially safe in previously uncomfortable situations, such as riding in public transportation or attending the theater, as long as I have my high-quality mask on. I have also decided that there is no point in wearing a cloth or surgical mask, especially because I have the easy option to wear a comfortable high-quality mask. (And if I were to wear a cloth mask, it would probably be in conjunction with a second higher-quality mask.)

So I continue to dine outdoors at restaurants. When venturing indoors, in a low-risk setting such as an uncrowded grocery store, I wear a KF94 mask, which I find extremely comfortable. In a higher-risk setting, such as a crowded theater or subway, I wear an N95 mask, which is less comfortable, but I believe is the gold standard for mask efficacy. (And for folks who are looking for a recommendation: while I find that many N95 masks are unbearably uncomfortable, the 3M Aura mask is actually quite tolerable. 3M also makes an N95 mask with an exhalation valve that makes it easier to breathe, but it won’t protect others from you, if you have Covid. You can find these in home improvement stores and online. I’m sure there are good KN95 masks too, but I’ve avoided these because of reports of counterfeits. Here’s an article all about high-quality masks, with recommendations.)

And finally, a few words about my choice of medium (Medium!) to share these results. Because I did not set out with the intention of the survey to be a scientific study, I don’t know that the quality of these findings is sufficient to merit being published in a peer-reviewed scientific journal. And even if they were, I wouldn’t know the first thing about how to go about it! I wanted to present the findings in a way that was directed at and understandable by the general public, while still giving useful details to those who understand the statistics — essentially part news article, part scientific analysis, and part opinion piece — and this seemed like the easiest way to do that.

ADDENDUM

I have posted the raw data here:
https://tinyurl.com/Stompede-Covid-Survey-2022

Except where noted, P-values in this article are derived using a two-tailed Z-test, with a Bonferroni correction applied for values <0.05. Thanks to Dave Goldberg for your guidance.

--

--