There’s a lot going on. It can be hard to know the right thing to say or do or think. I do hope your weekend provided at least some respite from “doom-scrolling”.

Today, a question about testing which I had planned even before it seemed so topical.

On the topic of reliability, you wrote in your latest post: “The alternative to quarantine is testing, which may or may not be available where you are. If you can be tested before you go, or when you arrive, this might alleviate fears.” I have read about high rates of false negatives in test results–I believe at one point at least one of the types of tests had a 40% false negative rate. Do you think it is helpful to test in light of potential inaccuracies and that one test result can be relied upon in the context of visiting family or having family visit? Are repeated tests more prudent/necessary (to the extent testing is available) (e.g. 2 or 3 tests on consecutive days)?

The question above came in last week, and I wanted to write about it since testing has been on my mind. It dovetails with issues of school reopen and University operations, both of which I think a lot about. And, of course, it’s come up in the context of the current White House outbreak.

I’m going to answer the specific question above, but I’m going to start with a bit of an overview of testing — focusing on the basics and the statistics — before getting into the answer.

Testing Overview & Background

COVID-19 tests can test for current infection or past infection. Tests for past infection look for evidence of antibodies. There was huge attention on these test early on, and we’re still doing some population level screening. But what seemed like an early focus on “immunity passports” has faded a bit. Increasingly, testing discussions have focused on testing for current infection.

Nearly all the tests we run for active infection are what are called “PCR” tests (longer form: polymerase chain reaction). These tests rely on either a nasal swab (most common) or saliva collection. The sample (mucus or saliva) is run through a machine that amplifies the viral DNA, and a positive result is delivered if it detects viral RNA. (Yes, virologists, I know this is a simplification! If you want more details – the RNA to DNA conversion and so on try the COVID-Explained Explainer here).

These test are quite accurate. Early in symptomatic infection, detection rates are very high — in the range of 95%. But they are subject to both false positives and false negatives. Since they rely on detecting viral RNA, they work best when viral load is high (i.e. when infections are symptomatic, and in the first days after symptom onset) and are more likely to miss asymptomatic infections, or infections in the days before symptom onset. These missed infections would be false negatives. False positives can also occur, for various reasons (contamination, presence of other coronaviruses).

Worth saying: many people associate these PCR tests with extremely unpleasant nasal swabs (“they tickle your brain” is a common claim). It is true that some kinds of nasal swabs can be awful, including the kind that were commonly used at the start of the pandemic. But: testing protocols are increasingly using much more minimally invasive test procedures (think: swab a q-tip right inside your nose) without much loss in detection quality. This has allowed, for example, many universities to implement regular, frequent population testing.

These test take some time to process and must be done in a lab. This means the turn-around time is a minimum of a few hours and, realistically, typically a few days with transit time, etc. There are faster, point-of-care testing processes. Several of them use a version of the same PCR technology. The newest entry in this space (not yet available, recently approved and in production) is a less expensive, very fast (15 minutes) test which relies on antigen detection (details here).

All of these faster technologies, at the moment, are less accurate than the standard PCR. In particular, they are more likely to yield false negatives — they miss more infections.

In general, testing technology is likely to continue to improve over time, although manufacturing and availability is clearly still an issue. It’s absolutely insane that that is still true eight months into this but it is.

What is Testing Good For?

I see this as a shortened way to ask the question above. Can we really rely on testing? It’s not perfect. Really, what’s the point?

In fact, I think there is tremendous value to testing. But let me first say what it is not, which is a way to return to exactly the way things were before. Aaron Carroll has a thoughtful piece on this in the context of the White House outbreak. If we had a perfect test which could detect any amount of virus immediately with extremely high sensitivity, then we can imagine a scenario in which you could test people on the way into the Rose Garden and then let them safely cozy up and hug one another (seriously, watch the video). But this is not where we are on tests. A negative test is not a perfect signal of safety (something that the Administration should have known). Wearing masks and distancing when possible and not hugging strangers is also a big part. (As Lizzie O’Leary pointed out on Twitter, this is reminiscent of the “Safety Lasagna” idea).

But this doesn’t mean tests are not incredibly useful. Even an imperfect test provides a lot of information. To think about why, imagine you have a PCR test which detect 90% of infections and going into the test you think you have a 1% chance of infection. With a negative test result, the chance you have the virus falls to (just about) 0.1% — from 1 in 100 to 1 in 1000. This is much lower! It isn’t zero, but it’s a lot less.

A number of Universities have made testing a key part of their reopen plans. If you test your population repeatedly, you’ll pick up a lot of infections. If you can isolate people who are infected and infectious, you lower spread. To zero? No. That’s why Brown (for example) requires everyone on campus to test twice a week and to wear masks and distance. But if you pull out 90 out of every 100 infected people, there are simply less to spread.

If we could do better, more reliable, faster testing in schools we’d have some of the same benefits. If there are infected people, many of them would be pulled out. Yes, a few people with negative tests would come through the cracks. Again, that’s the reason for masks. But there are simply many fewer opportunities for infection.

Ultimately, no matter what we are assuming some risk. Routine testing (or even reliable testing for those with symptoms) can lower that risk.

Weren’t you going to answer the question above?

Right, yes.

There is no question that completely isolating with no outside contact for 14 days is the safest way to ensure that you are not infected before seeing others. And in a situation in which the family member you are seeing is very elderly or immune compromised this may be what it takes.

But: if you find yourself in a lower risk situation (say, visiting older but healthy parents) you may be comfortable with allowing for some risks. There are always some risks (flu, etc) even when we do not think about them. In this situation a shorter quarantine could be appropriate, and testing could be an alternative or supplement to this. A test could yield a false negative, but if you do test negative you know your risk is much reduced.

As for when to test, repeated tests… Repeated tests are probably not realistic unless your resources outstrip the rest of us. Given the timing of viral onset, a test 3 to 5 days after possible exposure (for example, 3 to 5 days after travel) may be the most likely to be effective. As for whether and where to get this, HHS has a site where you can look up your state. Let’s just say some are better than others.