COVID Risk Calculations
There are too many variables to make generalized calculations.
A front-page story in yesterday’s NYT, “Is Covid More Dangerous Than Driving? How Scientists Are Parsing Covid Risks.” demonstrates how challenging public health policy can be.
Like it or not, the choose-your-own-adventure period of the pandemic is upon us.
Mask mandates have fallen. Some free testing sites have closed. Whatever parts of the United States were still trying to collectively quell the pandemic have largely turned their focus away from community-wide advice.
Now, even as case numbers begin to climb again and more infections go unreported, the onus has fallen on individual Americans to decide how much risk they and their neighbors face from the coronavirus — and what, if anything, to do about it.
For many people, the threats posed by Covid have eased dramatically over the two years of the pandemic. Vaccines slash the risk of being hospitalized or dying. Powerful new antiviral pills can help keep vulnerable people from deteriorating.
But not all Americans can count on the same protection. Millions of people with weakened immune systems do not benefit fully from vaccines. Two-thirds of Americans, and more than a third of those 65 and older, have not received the critical security of a booster shot, with the most worrisome rates among Black and Hispanic people. And patients who are poorer or live farther from doctors and pharmacies face steep barriers to getting antiviral pills.
We’re long past the point where I have a lot of sympathy for people who haven’t gotten the free vaccines that the government has been begging them to get for more than a year. But, yes, there are a whole lot of people with weakened immune systems, mostly through no fault of their own. Morever, the effects of the virus differ with age and other factors.
These vulnerabilities have made calculating the risks posed by the virus a fraught exercise. Federal health officials’ recent suggestion that most Americans could stop wearing masks because hospitalization numbers were low has created confusion in some quarters about whether the likelihood of being infected had changed, scientists said.
“We’re doing a really terrible job of communicating risk,” said Katelyn Jetelina, an epidemiologist at the University of Texas Health Science Center at Houston. “I think that’s also why people are throwing their hands up in the air and saying, ‘Screw it.’ They’re desperate for some sort of guidance.”
While I agree that we’ve done a poor job of breaking down risk calculations, I’m skeptical that it much matters. We couldn’t get people to wear masks when the disease was at its peak and before vaccinations were available. And many experts have argued that this very exercise—looking at risk from the standpoint of the individual rather than the collective—is the biggest challenge in getting people to act responsibly.
To fill that void, scientists are thinking anew about how to discuss Covid risks. Some have studied when people could unmask indoors if the goal was not only to keep hospitals from being overrun but also to protect immunocompromised people.
Scientists might be doing that but public health leaders aren’t. We made the switch from tracking spread to tracking hospital capacity precisely because we were never going to get spread levels down to levels that made it “safe” for immunocompromised people to go around maskless and people simply were no longer willing to put up with endless masking.
Others are working on tools to compare infection risks to the dangers of a wide range of activities, finding, for instance, that an average unvaccinated person 65 and older is roughly as likely to die from an Omicron infection as someone is to die from using heroin for a year-and-a-half.
But how people perceive risk is subjective; no two people have the same sense of the chances of dying from a year-and-a-half of heroin use (about 3 percent, by one estimate).
Indeed, I would have guessed the risk from sustained heroin use higher than that. But, then again, I’m skeptical that the rate for heroin addicts as a group tells us all that much; presumably, one’s risk varies based on frequency of use, dosage, the purity of the product, and all manner of other factors.
And beyond that, many scientists said they also worried about this latest phase of the pandemic heaping too much of the burden on individuals to make choices about keeping themselves and others safe, especially while the tools for fighting Covid remained beyond some Americans’ reach.
“As much as we wouldn’t like to believe it,” said Anne Sosin, who studies health equity at Dartmouth, “we still need a society-wide approach to the pandemic, especially to protect those who can’t benefit fully from vaccination.”
Again, we tried that for two years. People simply aren’t going to put up with restrictions based on worse-case scenarios. Regulations based on the effects on an 87-year-old with a lung transplant are going to be very hard, indeed, to enforce on healthy 23-year-olds.
Still, with far more immunity in the population than there once was, some epidemiologists have sought to make risk calculations more accessible by comparing the virus to everyday dangers.
The comparisons are particularly knotty in the United States: The country does not conduct the random swabbing studies necessary to estimate infection levels, making it difficult to know what share of infected people are dying.
Dr. Jetelina, who has published a set of comparisons in her newsletter, Your Local Epidemiologist, said that the exercise highlighted how tricky risk calculations remained for everybody, epidemiologists included.
For example, she estimated that the average vaccinated and boosted person who was at least 65 years old had a risk of dying after a Covid infection slightly higher than the risk of dying during a year of military service in Afghanistan in 2011. She used a standard unit of risk known as a micromort, which represents a one-in-a-million chance of dying.
Here’s a chart based on her calculations:
But, again, this is trickier than it looks.
Presumably, the deaths from running a marathon numbers are based on deaths during marathons. It would be much higher, indeed, if we forced people who hadn’t trained and qualified to run a marathon to run 26 miles. The risks from driving, motorcycling, or skydiving depend on one’s behavior, equipment, and environment. The risk of dying in Afghanistan surely depended on one’s occupational specialty, location, and time period. Of the activities in the chart, then, I’d guess that flying commercial and receiving anesthesia are the only ones that can truly be normalized
And, of course, the scientists trying to figure all of this out are well aware of that.
But her calculations, rough as they were, included only recorded cases, rather than unreported and generally milder infections. And she did not account for the lag between cases and deaths, looking at data from a single week in January. Each of those variables could have swung estimates of risk.
“All of these nuances underline how difficult it is for individuals to calculate risk,” she said. “Epidemiologists are having a challenge with it as well.”
For children under 5, she found, the risk of dying after a Covid infection was about the same as the risk of mothers dying in childbirth in the United States. That comparison, though, highlights other difficulties in describing risk: Average numbers can hide large differences between groups. Black women, for example, are almost three times as likely as white women to die in childbirth, a reflection in part of differences in the quality of medical care and of racial bias within the health system.
And even that vastly understates the challenge. We might be able to calculate a generic “risk of a child under 5 dying from a Covid infection” statistic but it would be meaningless. The risk of any given 5-year-old dying from Covid would presumably be a function of where they live, what activities they engage in, what precautions they’re required to take, their baseline health, their access to medical care, and a host of other factors that don’t immediately spring to my non-expert mind.
Cameron Byerley, an assistant professor in mathematics education at the University of Georgia, built an online tool called Covid-Taser, allowing people to adjust age, vaccine status and health background to predict the risks of the virus. Her team used estimates from earlier in the pandemic of the proportion of infections that led to bad outcomes.
Her research has shown that people have trouble interpreting percentages, Dr. Byerley said. She recalled her 69-year-old mother-in-law being unsure whether to worry earlier in the pandemic after a news program said people her age had a 10 percent risk of dying from an infection.
Dr. Byerley suggested her mother-in-law imagine if, once out of every 10 times she used the restroom in a given day, she died. “Oh, 10 percent is terrible,” she recalled her mother-in-law saying.
While that was a perfectly effective way of explaining how often “10 percent” is, it was a mind-bogglingly poor way of explaining what a 10 percent risk is. If there’s a 10 percent risk of 69-year-olds dying from Covid, it means 9 in 10 won’t die from Covid. The bathroom analogy makes death sound inevitable because it’s a ten percent risk taken ten times a day every day.
Dr. Byerley’s estimates showed, for instance, that an average 40-year-old vaccinated over six months ago faced roughly the same chance of being hospitalized after an infection as someone did of dying in a car crash in the course of 170 cross-country road trips. (More recent vaccine shots provide better protection than older ones, complicating these predictions.)
That’s much more helpful. Why, we have millions of long-haul truckers who make their living taking precisely that sort of risk. And, while truck driving is one of our most hazardous vocations, we don’t tend think of the risks they take as unmanageable. We do attempt to make the job safer by regulating training, safety equipment, sleep deprivation, intoxicating substances, and the like. But, crucially, we don’t excessively burden non-truck drivers as a means of mitigating the risks of that line of work.
For immunocompromised people, the risks are higher. An unvaccinated 61-year-old with an organ transplant, Dr. Byerley estimated, is three times as likely to die after an infection as someone is to die within five years of receiving a diagnosis of stage one breast cancer. And that transplant recipient is twice as likely to die from Covid as someone is to die while scaling Mount Everest.
Yet roughly 900 people chose to try to climb Everest every year. And, indeed, while the submitting success rate has steadily increased, the death rate has held remarkably steady at 1 percent—mostly from unpredicted severe weather episodes.
And yet some are calculating as though we can eliminate risk.
With the most vulnerable people in mind, Dr. Jeremy Faust, an emergency physician at Brigham and Women’s Hospital in Boston, set out last month to determine how low cases would have to fall for people to stop indoor masking without endangering those with extremely weakened immune systems.
He imagined a hypothetical person who derived no benefit from vaccines, wore a good mask, took hard-to-get prophylactic medication, attended occasional gatherings and shopped but did not work in person. He set his sights on keeping vulnerable people’s chances of being infected below 1 percent over a four-month period.
To achieve that threshold, he found, the country would have to keep masking indoors until transmission fell below 50 weekly cases per 100,000 people — a stricter limit than the C.D.C. is currently using, but one that he said nevertheless offered a benchmark to aim for.
It’s quite probable that we’ll never achieve that rate. And, again, healthy, vaccinated people simply aren’t going to put up with restrictions based on immunocompromised geriatrics indefinitely.