Anti-maskers and COVID-hoaxers are claiming that a small CDC study published in September proves masks don’t work. The study very nearly finds the opposite.

On September 11, 2020, the Centers for Disease Control (CDC) published a small study in which they attempted to assess the risk of certain types of contacts among people who tested positive for COVID-19 vs. people who had symptoms similar to COVID-19 but did not test positive. The researchers were trying to understand the relative risk of various types of activities such as going to a restaurant, getting your hair cut, going to the gym to exercise, etc. The study itself was conducted by interviewing 314 people gathered from 11 different hospitals around the country.

On October 13, Tucker Carlson falsely claimed on his Fox show that this same CDC study said that “almost everyone — 85% — who got the coronavirus in July was wearing a mask, and they were infected anyway. So clearly (wearing a mask) doesn’t work the way they tell us it works.” And two days later, Trump falsely claimed during his town hall that 85% of people wearing masks catch COVID, clearly a reference to what Carlson said on Fox.

Over the last week, this has turned into the latest disinformation attack on mask wearing. But to anyone who has read the study in question, it is very clearly dishonest, cherry picked, and quote mined disinformation.

Let’s look closer at the CDC study to understand why.

We can start with the title of the study: “Community and Close Contact Exposures Associated with COVID-19 Among Symptomatic Adults ≥18 Years in 11 Outpatient Health Care Facilities — United States, July 2020.” Right off the bat we notice that the study is small – just 11 outpatient facilities, and from one of the footnotes we see that all of the eleven facilities were in 10 states – California, Colorado, Maryland, Massachusetts, Minnesota, NorthCarolina, Ohio, Tennessee, Utah, and Washington. And the title doesn’t mention masks at all, so it doesn’t appear that the effectiveness (or lack thereof) is something of particular interest to the study authors.

The study itself is a comparison of 154 COVID patients who tested positive using RT-PCR testing to 160 patients who did not test positive but who had COVID symptoms nonetheless. This means that the total population of the study is only 314 patients, which is a very small number of patients given the over eight million Americans who have had the disease so far.

This raises the question – what limitations on their results did the authors identify? Thankfully, they wrote five down explicitly while acknowledging that there may be more (for example, the possibility that people lied in their answers).

1. There is some self-selection of the population because the patients actively sought out tests at the 11 study sites. The people who didn’t respond or who refused to participate in the study might be significantly different than the study group, and the study group and control groups were not controlled for age and sex simply because there weren’t enough total participants to do so.
2. The questions that were asked didn’t include all potentially important details, such as whether the study participants who ate at restaurants ate indoors out outside (COVID transmission outdoors is lower than indoors) or whether coffee shop visits were indoors or only via drive-up.
3. The patients from the 11 participating facilities may not be representative of all COVID patients or of the US population at large.
4. The patients were aware of their COVID test status and so that could have influenced their answers to the authors’ questions during interviews.
5. The control group (symptomatic but not testing positive) could be subject to the false negative effect (insufficient sensitivity) and so some of them could have turned out positive.

Based on this list of limitations, we can see that the study’s authors would not agree that 85% of COVID patients wore masks. The study clearly says that 85% of the study group (the 154 patients with positive test results) wore masks, but given the significant limitations of the study listed above, we can’t extend the study’s results to everyone who has a positive test.

So claims like those pictured above and at right are clearly not supported by this study. Maybe the claims are accurate, maybe they’re not – this study can’t say one way or another because of its limitations and the fact that answering this question is not what the study was designed to do.

Another false claim I’ve seen made is that this study supposedly shows that masks don’t work (and Carlson made that claim explicitly on October 13). After all, if most people who wear masks still get sick, then clearly masks are useless, right? Again, no. And the reason is that this claim is built on the claim that 85% of all COVID sufferers wore masks. As I just showed, the study doesn’t say that at all and, because the study wasn’t designed to answer that question, it can’t answer it. So this claim is flat out wrong.

As an aside, let’s not forget that this study is not meaningfully different from a public opinion poll. There are a lot of people who claim that polling firms got the 2016 election wrong (which is incorrect, but that doesn’t change the impression) and who thus reject polling. Yet a lot of those same people are now trumpeting this poll because they think it supports their false “masks don’t work” narrative. It’s opportunism and hypocrisy based in ignorance and cognitive biases, nothing more. But I digress.

So what was the study designed to do and what does it actually tell us?

The purpose of the study was to try and determine which activities were most likely to result in a positive COVID test as compared to a control group that was COVID-negative. In other words, the study wanted to determine the relative level of risk involved in the various activities. The activities that the authors asked about were hosting more than 10 persons in a home, shopping, dining at a restaurant, going to the office, visiting a salon, working out in a gym, going to a bar or coffee shop (indoors), attending church or a religious gathering, and using public transportation.

What the authors found was that the 154 COVID-positive study group were more likely to have visited a restaurant (p=0.01) or a bar/coffeeshop indoors (p = 0.22) than the control group. All the other activities – shopping, visiting the gym, going to a salon or barber, and so on – showed little to no statistical difference between the study group and the control group.

The authors didn’t mention this, but if you look at the green highlighted region in the screen capture from the study you’ll see that visiting the office (assuming the patient works in an office, of course) was one of the safer activities people could have done outside their home, although the statistical significance isn’t particularly high (p=0.27). The study doesn’t look into the details of why, but given many office-style employers have developed COVID plans that include mask wearing, distancing, minimizing meetings, work-from-home, and the like, this finding makes logical sense.

When the authors combined their findings about restaurants with followup questions about whether or not other patrons or employees at those establishments were wearing masks and/or social distancing, the authors found a statistically significant difference. Specifically, the 154 COVID-positive study participants were more likely to have been at restaurants or bars with less masking and/or social distancing than control participants (p=0.03 and p=0.01 respectively).

Put another way, the authors found that going to restaurants and bars appeared to significantly increase the risk of getting sick enough with COVID-19 to test positive for the virus, and the risk was greatest at restaurants and bars where patrons were not observing social distancing or masking. In fact, the only riskier activity was having close contact (longer than 15 minutes) with someone who was also COVID positive, and this risk was statistically significant (p<0.01) regardless of whether the person was family, a friend, or a coworker.

Essentially, wearing masks and distancing was safer than not wearing masks and distancing, and going to restaurants and bars is risky. Which is nearly the exact opposite of the "masks don't work" disinformation being spread by Tucker Carlson and so many others on social media.