In an ideal world, the best way to determine who in the country has the new coronavirus would be to simply test everyone. So far, though, the United States does not have the capability to do that and has been mostly focused on testing people who are symptomatic or find themselves in specific risky situations.
Two Dartmouth professors are suggesting that is actually not the best way to get a sense of who's infected, and are instead proposing that random testing be carried out of the general population. Michael Herron, a political scientist, and Dan Rockmore, an applied mathematician, make the argument in their cowritten online piece.
VPR's Mitch Wertlieb spoke to Dan Rockmore about the proposal. Their interview is below. It has been edited and condensed for clarity.
Mitch Wertlieb: How many people do you think would have to be tested randomly to come up with a reliable sampling for how severe the spread of this virus really is?
Dan Rockmore: Without getting deep into the weeds and the mathematics: there's a tried and true theory based on probability and statistics that the margin of error is one over the square root of the number of samples. So when people say, for example, that the margin of error was 3%, 1 over the square root of a thousand is roughly 1 over 30, and that's roughly 0.03. So you get 3% out of that. The interesting thing is it doesn't actually depend on the population size. That's what's kind of counterintuitive to people. So if you had a billion people in your population, if you still wanted 3%, if you could choose people randomly, you'd still only need a thousand people to talk to or to test.
I mentioned we don't have the capacity to test everybody in this country right now. Do we have the capacity to do this random testing?
Well, capacity is a matter of both actually having the tests, and then, how would you go about deploying such a test? The government does have all sorts of information on us, as people know, and it couldbe a matter of more or less aggregating that information, getting a big chunk of the population, and simply assigning numbers to them like our Social Security numbers, and then more or less putting those in a big computer barrel, so to speak, and choosing them at random. I mean, we we do have plenty of computational techniques as well as the actual records that would enable the process to go forward. And then, there is the question of whether or not governments are willing to do that and want to do that. But again, it just depends on whether they have the tests and what their will to do it is.
Well, Professor Rockmore, I hope this is not an unfair question because you are a mathematician. But my question is, you know, from a treatment perspective, why do you think it's important that we have a good sense of the infection and fatality rate across the whole population?
Mitch, thank you for pointing out that I'm not a doctor! Honestly, all I'm saying here is that treatment is tricky, right? People are symptomatic or they aren't symptomatic. And the way in which one treats them depends on the symptoms, often more than the disease, assuming that you don't actually have a cure.
This is really about baseline statistics regarding the prevalence of the disease. Without those kinds of baselines, it's hard to put policy in place with respect to activities and social actions. It's really about getting people back to work. The fact that so many people are out of work is leading to all kinds of other hardships, other kinds of illnesses. So what we've focused on in this piece is really about public health and then how society can function in the face of a virus, which presumably is going to be with us for sometime.
If we had a baseline infection rate, then you could actually say with some quantitative assurance, you know, that things are better here and they are worse here. We have a little sign just outside of Hanover as you enter Vermont: "If you've come to stay, quarantine for 14 days." There are concerns that some states are more at-risk than other states. To make these kinds of decisions in an informed way, we need to begin to get some better statistics about prevalence and mortality.
