Covid-19: We don't need better tests
Updated: Oct 8
The Covid testing strategy in the White House failed to prevent an outbreak in our highest levels of government. Critics quickly noted that Abbott's rapid point-of-care tests like the BinaxNOW are not accurate enough in individuals without symptoms, and so should not be used as a screening method. But it's tempting to think that if we just had more accurate tests, this strategy would work. I could test myself and everyone around me regularly, and if we all test negative, we won't need to worry about getting the virus. Even if I've been exposed to someone with Covid, I shouldn't have to quarantine if I test myself frequently and it's always negative. We just need accurate tests.
Let's do a thought experiment to see how this would work. Let's imagine a patient named Chris. On day 0, Chris is exposed to Covid. (Exposed means being in close proximity to someone with the virus for more than 15 minutes, without protective gear like masks.) During her exposure, SARS-CoV-2 virus particles enter her respiratory tract and bind to cells along that tract. After infecting the cells, the virus forces the cells to make baby virus particles - each individual cell can make millions of viruses. While this is happening, Chris feels fine. This is the incubation phase.
The image below shows a timeline of Chris's viral load - the amount of virus present in her lungs and airways. It starts around zero when she is first infected, seems to rise slowly, and then after a few days rises rapidly. When there are not many virus particles, and they are mostly deep in her lungs, she is not able to infect others. But at some point, the viral load gets high enough for her to be contagious. This is represented by the black dotted line. Note that she still feels fine. Most people seem to be contagious about two days prior to developing symptoms.
If Chris is tested on Day 2, it will most likely be negative, because her viral load is not high enough for the test to give an accurate result. The threshold for a positive test is at the red dotted line. No test is perfect, but for this thought experiment let's assume that ours is near perfect at detecting levels above the line. That is, if the viral load is above the red dotted line, it will give a positive result 100% of the time, and if the viral load is below that, it will give a negative result 100% of the time.
Why is there a threshold for being able to detect the virus? Our COVID tests rely on picking up pieces of the virus. PCR tests, the most common test performed, rely on finding pieces of the virus’s genetic material (RNA). Other tests, such as the rapid Abbot BinaxNOW, rely on picking up an antigen*, a protein on the virus. With both tests, the more virus that is present, the more likely the test will “find” the RNA or antigen and give a positive result.
Even with our extremely accurate test, there is still a gap between when Chris can spread the virus and when she will test positive.
What if we had a more sensitive test? If there are enough viral particles to infect someone else, we should be able to detect those. Here is what happens when our test gets a little more sensitive, and can pick up lower levels of virus:
This allows us to catch the virus a little earlier, but we still have a period in which our patient is able to infect others and have a negative test. As long as there is a difference between these thresholds, there is a weakness in our testing strategy.
Even if we have a test that is so sensitive, it will give a positive result when the viral load is just at the point of infecting other people, there is a problem. Our test provides an accurate snapshot of the viral load at the instant the nasal swab is inserted. A few hours later, it is no longer accurate.
More than just positive and negative
Our example above assumes that our test works perfectly at the viral loads indicated. But of course this is not true. We can have false positives (someone testing positive even though he doesn't have the virus) and false negatives (someone testing negative even though she does have the virus). A test that is very sensitive will have more false positives, while a test that is not so sensitive will have more false negatives.
And though we think of Covid testing as giving an either/or result - you either have it or you don't - it is better to think of these tests in terms of probability. That is, if I test positive, what are the chances that I really have been infected with SARS-CoV-2? And if I test negative, what are the chances that I really don't have the infection? The answers are never 100%. And interestingly, the answer depends on more than just how good the test is. The accuracy of the test depends partly on the person being tested and the chances that this person starts with.
This has a solid statistical basis (Bayes' theorem), but it can be understood intuitively. If no one in my town has Covid, then my odds of having it are very low - regardless of what my test shows. If my roommates were all diagnosed with Covid in the last week, my chances of having it are high - again, regardless of what my test shows. If my town has a high rate of Covid infection but I haven't personally had contact with anyone that I know was positive, then my chances of having it are between the first and second scenarios. We may not know the exact prevalence of Covid in our population, but even rough estimates will get us closer to understanding how to interpret a positive or negative result.
If Chris knew she was exposed to someone with the virus, she already has a fairly good chance of having Covid. A negative test two days later is not reassuring - she likely just isn't making enough virus to be positive. A negative test 7 days later is still not reassuring, since the incubation period can last up to two weeks. Even with a negative result, she still has a good chance of having Covid. Her chances of having Covid are high enough before testing that she has to act as if she is positive (quarantine) until the incubation period is over, regardless of her test results.
We don't need better tests. We need a better understanding of the tests and their limitations. Imperfect tests are still essential for identifying cases and controlling the virus.
For those who have had a recent confirmed exposure to Covid, the tests are not helpful. They still need to quarantine for the incubation time of the virus.
For a more technical explanation of Bayes' theorem in Covid testing, read this article by Jeffrey Schnipper and Paul Sax.
*Not antibody. The antigen is made by the virus. Antibodies are made by you. Tests that look for antibodies are trying to see whether you had the disease in the past. Antigen tests tell you whether the virus (or part of it) is there now. These are two completely different tests.