By Justin Fox
When White House coronavirus coordinator Deborah Birx went to Bismarck, North Dakota, late last month, she wasn’t thrilled with what she saw.
“Over the last 24 hours, as we were here and we were in your grocery stores and in your restaurants and frankly even in your hotels, this is the least use of masks that we have seen in retail establishments of any place we have been,” she told reporters. “And we find that deeply unfortunate because you don’t know who’s infected and you don’t know if you’re infected yourself.”
Birx is right! Not wearing a mask in the midst of one of the country’s worst Covid-19 outbreaks is dumb. Sitting in a Bismarck restaurant for 90 minutes without a mask on may bring a 5% risk of catching the disease, going by the estimates of the microCOVID project.
But do you know what most Bismarckians were doing when Birx visited? Staying home, from the looks of it. For the past couple of weeks, people in the Bismarck metropolitan area have been restricting their movements relative to normal even more than New Yorkers have.
This measure is calculated by economists at the Federal Reserve Bank of Dallas from mobile-device geolocation data. It combines several different variables such as the fraction of devices leaving home in a day and the average time spent at locations far from home. In the now-months-long debate over whether lockdowns or the pandemic hurt the economy more, this seems to offer a bit of evidence for the latter, as New York still has many more government restrictions on commerce and behavior than North Dakota does, according to the coronavirus stringency index compiled by the Blavatnik School of Government at the University of Oxford.
I’m more interested, though, in what the slowdown in activity implies for what happens next with the outbreak in North Dakota and neighboring states, and the overall resurgence of the disease that has now for the first time put the U.S. at more than 100,000 new confirmed cases a day.
For North Dakota, the Centers for Disease Control and Prevention’s aggregate of 25 independent epidemiological forecasts projects confirmed cases in the state will stop rising this month. This forecast could turn out to be way off, of course. And confirmed cases are an inadequate metric of the disease’s spread when the percentage of tests coming back positive is in the double digits and rising, as it has been in North Dakota (the latest seven-day average is 14%).
But the change in behavior reflected in the above chart should have some impact. Similar modest declines in mobility preceded epidemic peaks in several Sun Belt metropolitan areas hit hard by Covid this summer.
The link between people going out less and the disease spreading less seems obvious enough, but it can be helpful to run through a simple infection model to clarify it. The key variable is the reproduction number, usually represented by the letter R, which is how many other people each person with the disease can be expected to infect. R0, or R-naught, reflects how the disease spreads when (a) nobody has immunity, (b) nobody changes their behavior and (c) everybody mixes with everybody else more or less randomly. For Covid-19, R0 is usually estimated at a bit above 3. That’s a lot lower than the measles (15 or so) but a lot higher than seasonal influenza (around 1.3).
The effective R — how fast the disease is really spreading at any given time — reflects people developing immunity to the disease, changing their behavior and mixing with one another unevenly (that is, some people hang out together in Bismarck bars while others stay home). It can be thought of as the product of four things that London School of Hygiene and Tropical Medicine epidemiologist Adam Kucharski dubs DOTS, for:
- The duration of time someone is infectious
- The opportunities for transmission during this period (such as social interactions)
- The transmission probability during each opportunity (such as whether someone coughs or sneezes)
- The probability that the person on the other end of the interaction is susceptible to that infection.
When people go out less, the opportunities for transmission of Covid-19 decline. When lots of them wear masks, the transmission probability likely declines. And when people get the disease and recover from it, susceptibility declines. Estimates of the effective state reproduction numbers on the rt.live website created by Instagram co-founder Kevin Systrom and statistical modeler Thomas Vladeck ranged from 0.84 in Mississippi to 1.45 in Maine when last I checked. North Dakota’s is 1.17, which doesn’t sound like much. But as German Chancellor Angela Merkel elegantly explained back in April, it only takes a few months at an effective R of 1.1 or 1.2 for a Covid outbreak to grow from tiny to hospital-filling.
Still, that’s a lot slower-moving than a Covid outbreak with an effective R of more than 3, which is what overwhelmed New York City in a matter of weeks early this year. It also means it shouldn’t take all that big a change in behavior, coupled with rising immunity, to get the effective R back down below 1.
That’s the optimistic view currently reflected in most of the epidemiological models tracked by the CDC. The agency’s aggregate forecasts show cases plateauing in coming weeks in currently hard-hit states such as Montana, North Dakota, South Dakota and Wisconsin. On a national level they predict cases will continue to increase but at a slower pace than over the past couple of months. Cases are rising now in almost every state, but most states have already been through at least one Covid wave, which makes a difference.
As Bob Wachter, medicine department chair at the University of California at San Francisco, put it on Twitter this week: “One gets the sense that prior surges are protective — likely a combo of people w/antibodies & fear of God.” Add to that the seeming decline in the deadliness of the disease, and the possible looming arrival of vaccines and, well, maybe there really is a light at the end of this tunnel.
A less-optimistic view is that we have yet to experience Covid through a Northern hemisphere winter, and can’t judge yet what impact much-colder weather and shorter days will have on both the spread of the disease and its severity. There is at least the possible upside that the customary winter scourge of influenza will be much less widespread than usual because the precautionary measures people are taking against the much-more-contagious Covid-19 drive the effective R of the flu well below 1. Who knows, though.
Back at the beginning of July, I took a probably unwarranted stab at epidemiological forecasting, arguing that because of the forces described above, the Covid epidemics in Arizona and elsewhere were about to peak. That turned out to be right, but I don’t get the sense anybody worth taking seriously is truly confident in forecasting just what will happen with Covid in the US over the next three or four months.
We haven’t turned the corner, and it’s not magically going away. Beyond that, in the likely absence of any major pandemic-fighting push from Washington before late January, it’s going to be up to state and local policies and a lot of individual decisions like the ones people are currently making in Bismarck to determine the shape of the curve. So … wear a mask, people.