Not Wearing a Mask for One Day is Like Drunk Driving Five Times
How important is it to wear a mask during the Covid-19 pandemic? How important is it to get the vaccine as soon as possible? These kinds of reasonable questions have been hard for authorities to answer in a way that makes sense to normal people. The official response has amounted to declaring, over and over, “Trust us and do what we recommend.”
For various reasons, ranging from perfectly sensible to cartoonishly paranoid, many Americans have balked at this response. Here I propose a different approach to communicating risk that might work better in some cases. I suggest authorities do some basic calculations to compare “unsafe” behaviors with more familiar risks we all agree should be prohibited as excessively reckless, such as drunk driving.
My own rough calculations suggest that, over the last year, every single day a person has skipped wearing a mask in public has created about as many additional deaths as drunk driving five times. If you work and shop and socialize without a mask on Monday, that one act is like drunk driving home from the bar on Monday, Tuesday, Wednesday, Thursday, and Friday that week. Not wearing a mask exposes ourselves and others to more risk than a behavior that has been outlawed in all fifty states on a bipartisan basis.
Getting a vaccine is like wearing a super-effective mask every day. So postponing a vaccine by an additional day also amounts to drunk driving over and over again, at least for people who would otherwise go about their daily lives without masks or social distancing.
I provide the calculations and sources for these statements below. My calculations are surely somewhat wrong, but probably not terribly wrong. I hope people with more epidemiological expertise invest in clarifying, formalizing, and popularizing these kinds of comparisons. This approach can reassure a skeptical public that recommended behaviors make basic intuitive sense, and are not some neurotic bureaucrat’s hyper-conservative personal preference.
Why haven’t epidemiologists and other experts already provided these comparisons more routinely? Perhaps many will object that the risk of not wearing a mask varies dramatically by time and place due to prevailing infection rates, hospital capacity, age and health of community members and the focal person making the decision, local prevalence of mask-wearing and social distancing, weather, etc. They’re right, but they’re missing the point, and suffocating clarity with nuance. The same variation also applies to drunk driving — it’s often safe to drunk drive if you’re not tanked, the roads are clear, the weather is good, you’re an experienced driver, etc. Despite all that, we go with a rule of thumb: “don’t drive a multi-ton hunk of metal around your community if you’re not perfectly sober.”
Helping people build a bridge between unfamiliar risks and familiar risks is perhaps the simplest way to empower informed decisions. Without that bridge, people struggle to reason clearly. Without clear reasoning, and without trust in authorities, they are lost. In a recent article by Atul Gawande, a mask skeptic does not dispute the fact that Covid has a death rate close to 1%. But without any bridge to more familiar risks like drunk driving, he doesn’t know how to process that risk — to him it sounds safe, because 1% is small in so many other contexts. “We’re living in fear,” he protests, “we’re instilling that fear — fear for a virus that has a cumulative survival rate of over ninety-nine per cent.” In the article, it’s clear the person is likely a pleasant, responsible adult in many other ways. He surely would not endorse drunk driving. In a way, his mistake reflects a failure of public authorities to help people like him understand small probabilities in more familiar terms.
The same approach can also work in reverse — we can help people accept that something is safe by comparing it to a familiar risk such as sober driving. Some vaccines and medicines have extremely rare side effects that can result in death. What does a 0.000001 % risk of death mean to a regular person? Nothing. One option is for a physician to look you the eye with confidence and say, “It’s a small risk, I wouldn’t worry about it if I were you.” That works in a lot of cases. But sometimes, it might be useful to tell patients that a 0.000001% risk of death is the same risk associated with driving one mile in their car. In other words, driving to visit the doctor is many times more dangerous than the side-effects of this hypothetical medicine. People can understand that; they know driving one mile is perfectly fine. If they trust the number itself, they don’t necessarily need to trust the judgment of the authority figure making the recommendation. They can take the number and make up their own minds.
Details for Calculation
Deaths per drunk drive:
1 death per 100m driver-miles — https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/811701
8 miles per typical drive: https://www.statista.com/statistics/1004111/us-average-person-trip-length/
Drunk driving about 10x more deadly than sober driving — https://www.verywellmind.com/drunk-driving-the-dangers-63002 citing Romano et al 2014, also citing this summarizing older research: https://pubs.niaaa.nih.gov/publications/aa31.htm, so 10x about right for just-over-threshold, higher for worse offenders — I use 10x since that is the law.
Calculation:
10x death risk multiplier for drunk-driving * 1/100m deaths/mile for regular driving * 8 miles
= 1/1.25m more deaths per drunk drive
COVID-19 deaths per day of mask avoidance
Mortality rate around 1% — https://www.who.int/bulletin/volumes/99/1/20-265892/en/ and also https://www.webmd.com/lung/news/20201030/covid-19-infection-fatality-ratio-is-about-one-point-15-percent
Ambient infection rate — tricky business:
Total infection rate over prior year was around 20%: https://www.news-medical.net/news/20210208/New-machine-learning-algorithm-estimates-number-of-COVID-19-cases-in-the-US.aspx#:~:text=Of%20those%2071%20million%20Americans,4%2C%20according%20to%20the%20algorithm. I divide this by 365 to get a daily infection rate — that is not quite right. But the data on daily flows of new cases over the last year suggest about 0.5% average ambient infection rates assuming each case is sick for 14 days, and that would be in the ballpark of 20%/365.
Another sanity check—there are 25 two-week periods in a year, and under a 0.5% ambient infection rate the odds you’re not infected in any of those 25 draws is .995²⁵ = .88 suggesting a 12% total infection rate, which is in the ballpark of that 20% infection rate estimated independently cited above. So that reassures me that taking the daily cases and multiplying them by 14 to get a 0.5% daily ambient infection rate is not crazy.
Transmission rate — say mask reduces R (the average number of additional infections spawned by each new infected person) from 1 to 0.2, so 4/5, based on finding that masks are 80% effective: https://www.pnas.org/content/118/4/e2014564118 and R without masks has typically been close to 1 over the last year in the US: https://covid19-projections.com/infections-tracker/.
Calculation:
1/5 ambient infection rate over one year of Covid * 1/100 mortality rate * 4/5 mask efficacy impact on transmission rate (“R0”) / 365 days/year
= 1/228k new deaths from skipping a mask for one day
Comparison: how many deaths does one day of mask-avoidance entail, in units of “drunk drives”?
1.25m/228k = 5.4 drunk drives per day of mask-avoidance