How Data Reveal Covid - Mask-Use Correlation: More Masks, Less Covid

The chart of real time data about the correlation between mask wearing percentages in a state with the prevalence of Covid reveals that mask wearing does reduce Covid risk. JL

Christopher Ingraham reports in the Washington Post:

Data from Carnegie Mellon’s CovidCast, a project tracking real-time coronavirus statistics, illustrates how mask usage influences the prevalence of covid-19 symptoms in a given area. State residents at the bottom for mask rates have the highest levels of covid-19 symptoms, 40 and 50%. As mask use increases, the frequency of observed covid-19 symptoms decreases.The R-squared of mask and symptom data is 0.73, meaning that you can predict 73% of the variability in state-level covid-19 symptom prevalence by knowing how often people wear masks.

Despite the clear opposition to masks within the Trump White House and among its allies, Americans of all political stripes overwhelmingly support their use as a public health measure and say they wear them whenever they’re in public.

Still, there are significant differences in mask-use rates at the state level. And data from Carnegie Mellon’s CovidCast, an academic project tracking real-time coronavirus statistics, yields a particularly vivid illustration of how mask usage influences the prevalence of covid-19 symptoms in a given area. Take a look.

For all 50 states plus D.C., this chart plots the percentage of state residents who say they wear a mask in public all or most of the time (on the horizontal axis) and the percentage who say they know someone in their community with virus symptoms (on the vertical axis). If you’re curious about the exact numbers for your state, there’s a table at the bottom of this article.

Take Wyoming and South Dakota, for instance, in the upper left-hand corner of the chart. Roughly 60 to 70 percent of state residents report frequent mask use, as shown on the bottom axis, which puts them at the bottom for mask rates. They also have some of the highest levels of observed covid-19 symptoms, approaching 40 and 50 percent.

Now, note what happens as you move across the chart. States farther to the right have higher rates of mask use. And as mask use increases, the frequency of observed covid-19 symptoms decreases: More masks, less covid-19.

This relationship is called a correlation, and it’s a strikingly tight one. Often in these types of plots you have to squint really hard to suss out such a relationship, and researchers occasionally go to comical lengths to divine the presence of a correlation where none really exists.

But there’s no need for that here. There’s a simple statistical measure of correlation intensity called “R-squared,” which goes from zero (absolutely no relationship between the two variables) to 1 (the variables move perfectly in tandem). The R-squared of CovidCast’s mask and symptom data is 0.73, meaning that you can predict about 73 percent of the variability in state-level covid-19 symptom prevalence simply by knowing how often people wear their masks.

One other observation to note is that almost without exception, the states with the highest rates of mask-wearing were won by Hillary Clinton in 2016. In other words, more Democrats, more masking — a vivid reflection of how partisanship has been a factor in much of the response to the pandemic.

Let’s pause a minute to talk about where exactly this data comes from. Ideally you would want it to be from something like a random-digit-dial survey, the type typically used in public opinion polling, which with enough participants would produce a sample of each state that’s representative of its population and demographics. But the cost of running one such survey for all 50 states plus D.C. would be enormously prohibitive — to say nothing of doing so on a daily basis, which is necessary to produce the kind of real-time data of interest to epidemiologists.

So the CovidCast team partnered with Facebook, which is used by 70 percent of U.S. adults and has the ability to survey tens of thousands of them every day at relatively low cost. While the resulting state-level samples aren’t perfect representations of the general population, the researchers weight the responses using Census Bureau demographic data to ensure they’re a good approximation.

“If Facebook’s users are different from the U.S. population generally in a way that the survey weighting process doesn’t account for, then our estimates could be biased,” cautioned Alex Reinhart, a Carnegie Mellon professor of statistics and data science who works on CovidCast and wrote a book on statistical methods. “But if that bias doesn’t change much over time, then we can still use the survey to detect trends and changes.”

He also cautioned that the old saw of “correlation doesn’t equal causation” applies here as well.

“There could be other explanations for the correlation,” he said. “For example, states that had worse outbreaks earlier in the pandemic both have higher mask usage now and more immunity.”

And, he added, “if people say they’re not wearing masks, they may not be taking other protective measures either. So perhaps what we see is a combination of mask usage, other social distancing behaviors and perhaps other factors we haven’t measured.”

Nevertheless, the chart is particularly useful in the context of all the other high-quality evidence showing that masks reduce the transmission of the coronavirus and other respiratory diseases. There’s good reason to suspect, in other words, that rates of mask use are driving at least part of the relationship seen in the chart above, even if the data can’t prove that definitively.

For people living in states that are driving the latest spike in coronavirus cases, the takeaway is clear: Wear a mask when you go out in public.

Chart data