A Million-in-One Shot


This Sunday’s edition of my local paper, The Boston Globe, included a couple of op-eds worth analyzing, starting with this piece (reproduced below) on calculating risk when it comes to the Omicron variant of COVID-19.


The Globe story provides an example of how we inhabitants of an infected planet can do something I mentioned last time was in our interest: move from treating COVID-19 as a deadly peril to treating it as a risk which, like many risks humans face daily, we may need to learn to live with, rather than eradicate.


When COVID first hit in 2020, it represented a deadly threat with no known means of protecting against it. Under the circumstances, it made sense to do whatever we could to stop the spread of the disease by isolating ourselves from one another, whatever the cost.

Given how little we knew at the time, it was also reasonable to try different ways to protect ourselves and others, even if some of those measures (such as sanitizing surfaces and wearing cloth masks) later turned out to not be terribly effective.


Nearly two years of experience, however, provides a basis for understanding and calculating risks associated our next set of choices using techniques that have been in place for decades.


For example, until recently nearly all car rental companies refused to rent vehicles to those under the age of 25, despite the fact that adults 24 and under were old enough to drive, vote, go to war, marry, and start businesses and families. Why? Because research indicated that those 24 and younger were so much more likely to get into accidents that it was not worth the risk of renting cars to them.


Note that this decision was based on statistics, meaning that rental companies were not saying anything specific about any particular 18-to-24-year-old. Rather, they were saying that – in aggregate – the reward of serving a young-adult market was not enough to overcome the downside.


Along similar lines, the Globe story introduces readers to a unit called a “micromort” which represents a one-in-a-million chance of dying of something. For instance, 413 skydiving deaths out of 48 million recorded jumps translates to a risk factor of 8 micromorts associated with the sport.


Applying the same calculation to COVID, it turns out that being infected by Omicron while vaccinated translates to .25 micromorts, compared to 6.3 micromorts for the unvaccinated, providing an elegant demonstration of the twenty-five-fold increased level of risk the unvaccinated are taking with their own lives.


In theory, associating our choices with different micromort quantities can give us a way to navigate decisions we need to make in the Age of Omicron, not to mention other threats that might show up in the future. As helpful as it might be to quantify human experience in such a way, however, statistical decision-making has its limitations.


To begin with, statistical prediction – like statistical evidence – must sit alongside other sources of insight, such as anecdotal data. The one-in-four-million odds of a vaccinated person dying of Omicron represented by .25 micromorts might seem like an extreme longshot, but for those of us who knew people who perished from COVID, anecdotal evidence of their horrible deaths generates legitimate fear that can overwhelm arguments based on extreme statistical unlikelihood.


Also, because COVID is so contagious, getting infected increases the risk of spreading dozens or even hundreds of micromorts among loved ones and strangers. Knock-on effects, such as the infected taking up hospital beds that could have gone to those with other serious illnesses, means there is more that needs to be worked into risk calculations beyond risk of our own demise.


There are also all kinds of other risks associated with choices made and not made in response to COVID that need to be worked into any analysis. Shutting down businesses and schools, for example, would very likely lower the risk of increased infections (which can be measured in micromorts), but might also increase the risk of impoverishment, learning loss, and deterioration of mental health, all of which are all much harder to quantify than the binary of living versus dying.


Finally, statistical calculations tend to work best when they are based on a reasonable level of stability within which to make predictions. Needless to say, a new variant that has infected millions of people who did all the right things over the last two years represents an extreme outlier that has already swamped the statistical models we used to understand the spread of COVID since it first appeared in 2020.


Mathematical and algorithmic solutions to human problems can seem alluring, and statistical methods can definitely play a role in helping us understand the world and make good decisions. But they need to be supplemented with other sources of insight, including wisdom and the ability to argue over alternatives using the tools and methods you have been learning about here at LogicCheck.


Omicron gives new meaning to calculated risk (Boston Globe, January 9, 2022)

Forty-odd years ago, Stanford engineering professor Ron Howard coined the term ‘micromort’ to quantify a 1-in-a million chance of death.


For two years, we have been living with the risk of a coronavirus infection, and the attendant chances, in declining order of possibility, of illness, hospitalization, and death.


But there have been almost no reliable tools to help us navigate these risks. It wasn’t so long ago that friends of mine were sanitizing their mail and triple-masking outdoors — precautions later revealed to be excessive, but who knew at the time?


There is a science of risk. Forty-odd years ago, Stanford engineering professor Ron Howard coined the term “micromort” to quantify a 1-in-a-million chance of death. For example, there were 413 skydiving deaths in the United States between 2000 and 2016, out of over 48 million recorded jumps. So each jump had a risk of roughly 8 micromorts.


For comparison, everyday life has its risks. Once you’ve woken up, there is a 22 micromort, i.e., 0.000022, i.e., a 1-in-50,000 chance that you will expire before day’s end. Flying is safer than driving, and so on.


So what about COVID? Using Centers for Disease Control and Prevention data, Dr. Ethan Craig at the University of Pennsylvania calculated coronavirus risks for May 1 to Oct. 18, 2021. Risk of death among the vaccinated was 0.25 micromorts a day. Risk of death among the unvaccinated came to 6.3 micromorts, roughly 25 times higher. “Does the idea of BASE jumping make you nervous?” Craig wrote. “Being unvaccinated sure should, as well.”


Craig admits that there are inaccuracies built into his “back-of-the-envelope” calculations, and that risks rise for the elderly and the immunocompromised. That said, he is far from alone in trying to quantify Life Under COVID. A group of “relentless tabulators” in San Francisco, including a couple of MIT grads and a physician, wanted to figure out how risky it was to go shopping, take a bike ride, or attend a protest during the pandemic.


As a homage to Howard, they created the “microcovid” metric, a 1-in-a-million chance of catching COVID. Just this week, they updated their online Microcovid Calculator to take account of the teeming Omicron infections. I plugged in the details of my Wednesday lunch with my fully vaccinated friend Roger at the deserted Ani’s Cafe in Belmont. It set my risk at about 50 microcovids, or what they deem to be one-quarter of my weekly COVID “risk budget.” That is, a 1 in 20,000 chance of being infected.


Post-Omicron, Craig has rethought his article, somewhat. In November, he noted that “a vaccinated individual engages in many activities that are associated with a higher risk for death than COVID-19 — including driving to work.” Now, “I suspect some of the numbers will prove to be pretty similar,” he told me. “For an otherwise healthy, fully vaccinated and boosted individual, the risk for an adverse outcome remains low.”


But fast-spreading Omicron has changed the collective risk, he explained. “Your individual risk factors into a population risk, where the chance of spread is much greater now.” Craig has adjusted his personal COVID risk budget accordingly, not eating indoors and not inviting guests to his home.


“We are in a much better place in a lot of ways” now, compared with March 2020, he said. “Now we have vaccines that work remarkably well at preventing severe outcomes (despite an immune-evasive variant). We know who is at highest risk for severe illness. When I go to work, I’m not afraid like I was then. At this point, I’d say I respect the virus, and take it seriously, but don’t fear it.”


Everyone, including Craig, knows there is a lot of coronavirus circulating now. The generally sane Dr. Robert Wachter, chairman of the department of medicine at the University of California at San Francisco, says this month will be “awful” for Omicron spread.


For the immediate future, mundane activities such as shopping or chatting with the postal worker give new meaning to the term “calculated risk.” The best advice: Proceed with caution.