We can do better on pre-risk assessment
Welcome to Plugging the Gap (my email newsletter about Covid-19 and its economics). In case you don’t know me, I’m an economist and professor at the University of Toronto. I have written lots of books including, most recently, on Covid-19. You can follow me on twitter (@joshgans) or subscribe to this email newsletter here.
Today’s post is about assessing the pre-risk of people so as to better allocate scarce tests. Put simply, the more likely that someone is infected with Covid-19, the more likely a test will provide a clearer signal of whether they do actually have Covid-19. This is especially the case for point of care tests.
The CDC recognises this. Here is a flow chart of their protocol for POC tests at nursing homes.
Notice how it works. The first question is about symptoms and pre-risk. As I have noted earlier, symptoms associated with Covid-19 do increase the likelihood you have it. But the second question is also associated with it in the sense that it asks a question about the population the candidate is part of and whether there is an outbreak occurring.
As the pre-risk that you have Covid-19 goes up, how the test is used changes. If you have symptoms, the tests are used to clear you. In effect, with symptoms, you are presumptively positive (they don’t say that but their actions speak louder than words) until you can have tests that prove you otherwise. If you are asymptomatic then the test provides a clear signal if there is an outbreak. If there isn’t one, they don’t really know and want more tests. It should say that a test outcome is presumptive in both cases. They don’t say that but again the request for more testing in each case speaks louder than words.
The point is that pre-risk matters. If you have a POC test with a 15 percent false negative rate and a 2 percent false positive rate, if your pre-risk is 1 in 4,000, a positive test only gives you a 1.1% of actually being positive while a negative test hardly gives any additional information. Jump that pre-risk probability up to 5 percent and the positive test gives you a 69% chance of being infected while a negative one drops your risk to less than 1%.
But the CDC guidance, while based on pre-risk, does actually put any numbers there. And the numbers matter. There are outbreaks and there are outbreaks. There is high and low incidence. It seems like we could do better.
In effect, we want more information than just symptoms, population prevalence and close contact outbreaks. There would surely be a big difference between someone who just comes from home to work each day and someone who has multiple points of contact with other people. Surely we need that information.
The irony is that we could have it. The very same contact tracing apps that indicate to us whether you have had contact with an infectious person, could be used to assess people’s pre-risk in terms of coming up with a measure of how many people, in what settings and for how long they have had contact with others. You could then use that to construct a proper estimate with their pre-risk. We could then use that to provide a much more granular and accurate protocol like the CDC one.
And you wouldn’t need to violate privacy. This is information that your app is collecting anyway so they can inform you if someone in your past tests positive in the future. You just have to aggregate it differently and then provide that score to assist in the testing protocol. If tests are supply-constrained, this would be a good way of sorting who needs a test and who doesn’t.
Once again, we have the ability to do so much better in terms of using information to manage the pandemic. I wish we would do it.