COVID-19: How to Use Rapid Tests
by Raywat Deonandan, PhD
Epidemiologist & Associate Professor
University of Ottawa
(I add my credentials to these COVID-19 blog posts in case they get shared. I want readers to know that my opinion is supposedly an educated and informed one)
Now, this is a topic that I thought had been covered well over the last few months. I know I’ve talked and written about it a few times. And yet journalists still ask me about it. So today I’m going to go over some quick thoughts and explanations about the uses, limitations, and strategies for using COVID-19 so-called “rapid tests.”
My intent is to make this will be a short post. Let’s see if I have the discipline to hold to that promise.
What is a PCR Test
The gold standard for COVID testing is, of course, the PCR test. It essentially amplifies the minute genetic signal of the invading virus until it is detected and identified as such. PCR technology is powerful, resource and time intensive, and very very sensitive.
This hypersensitivity has led to some people (erroneously) believing that the PCR test results in an overabundance of “false positives”, since (they believe) that old infections and viral fragments are being misidentified as actual active infections. I explained why this is not something we should worry about in this post.
The PCR swab test involves the deep nasopharyngeal insertion of a swab, which is then sent to a lab where the signal is amplified chemically, resulting in a diagnosis in typically several days.
What is a Rapid Test
Rapid tests typically do not look for tiny genetic signals to amplify. Instead, they look for antigens, or proteins on the surface of the virus, which are much bigger. Therefore they are unlikely to find viral fragments and confuse them for an infection. Whereas they are very good at seeing and identifying the antigen in question. Therefore the likelihood of a false positive is very rare.
In other words, if an antigen test says you’re infected, you’re probably infected.
However, the test can often miss the antigen if there are low levels of virus circulating in the person’s system. As a result, the probability of a false negative is surprisingly high.
Rapid tests tend to be nasal or throat swabs that are placed into a magic box onsite, or onto a strip of diagnostic paper, which renders a result –presence or absence of the antigen– in a matter of minutes.
What Are the Advantages of Rapid Tests?
As noted, they are rapid. Instead of days, you get your result on-site in minutes. In theory, a non-expert can apply the test. And the tests are much much less expensive than the PCR tests, so a community could acquire them in vast numbers relatively cheaply.
What’s the Problem With Rapid Tests?
While many rapid tests could, in theory, be applied by amateurs, let’s be honest here. No one is letting an untrained person stick a swab in their nose or throat. So in real life, a health care worker will still have to be the person doing the sticking… and that costs time and money.
As noted, the rate of false negatives is uncomfortably high. I explained sensitivity and specificity in this post, so I don’t want to go over it again. But those two indicators are what we use to measure the quality of a test. The reported sensitivity and specificity of various rapid tests on the market sound high, claiming to be over 90% in each category; but post-market “real world” application has shown vastly reduced numbers.
So we’re back to our conundrum with false positives and false negatives. In short, these tests –particularly a group called “lateral flow assay” (LFA) tests– are great at finding infected people with high viral loads (i.e., a low false positive rate), they are not good at finding infected people with low viral loads (i.e., a high false negative rate).
So the fear is that someone getting a negative result would inappropriately take that to mean that he or she is genuinely not infected, and will go about their business, accepting exposure risks that they would not have otherwise taken. In other words, an infected person with low viral load will get a false negative from an LFA rapid test, then take that result as license to go kiss their grandmother, thus infecting her.
The English city of Liverpool attempted to use LFA tests to do a near-census of residents and weed out COVID cases. But the LFA test missed half of all infections, presumably because it was unable to detect the low viral load cases… rendering a high false negative rate.
What Are the Various Applications of Tests?
Let’s go over the various applications of medical tests. I like to think of them as falling into one or more of four application categories: diagnostic, screening, surveillance, and reassurance.
Diagnostic testing is what happens when you’re sick, you go to the hospital, and you’re given a test to figure out what’s wrong with you. An MRI or CT scan of your head is a good diagnostic test to help determine if you’ve got a brain tumour or physical injury to your brain, for example. Such tests can be expensive, time and labour intensive, but must be highly highly specific and sensitive, because we cannot afford a misdiagnosis. In the COVID domain, the PCR test is the best diagnostic test that we have.
Screening tests are an initial test done to determine if a person qualifies for a follow-up test. A mammogram is a good screening test. We apply them to many qualifying women, including those with no symptoms. And if a mammogram suggests that there’s a lump, those individuals are given a follow-up investigation via biopsy.
Mammograms have a high level of false positives but a low number of false negatives. This is because we want to cast the net widely to make sure we capture all the cases of breast cancer. False positives are okay because we will weed them out with the biopsy, which is exquisitely high sensitivity and specificity. But the false negatives are intolerable because those women will be walking around not only not knowing that they have breast cancer, but actively thinking that they do not have breast cancer, so will never seek a biopsy to set themselves straight. In such a situation, the screening test does actual harm.
Right now, the best COVID screening tool is a symptom check or questionnaire. But it’s not great. The rapid tests are poor screening tools because of their high false negative rate.
Surveillance is the act of looking for cases in the population. Surveillance testing can be active (in which we go out and hunt down cases) or passive (in which we wait for the cases to present themselves). For the most part, COVID surveillance has been passive– we know mostly about the cases who present themselves for testing. But some places, like the city of Ottawa, do daily testing of the wastewater, which is an excellent form of surveillance.
An active surveillance test does not need to be high sensitivity or specificity. A different sample of the population can be randomly selected for a test in various time intervals. This would give us a sense of where the disease is present, and how it is trending. A rapid test can be well used for this purpose.
Reassurance testing is what some people call the need for testing for administrative purposes; for example, the testing of workers who must provide proof of negative status to remain at work.
So Where Can We Use Rapid Tests?
Because rapid LFA tests have a higher than ideal rate of false negatives, three conditions must be applied to their use:
(1) They must only be used for asymptomatic people, and/or people with no reason to think that they are infected. This is critical. This is because of something called the pre-test probability of infection (which I described here).
If you have no reason to think that you’re infected, then your pre-test probability of infection is low. If, on the other hand, you recently French kissed a COVID-positive person or have a fever, the probability of you being infected got higher, so your pre-test probability of infection is high.
So if you take a rapid test and get a negative, whether or not that negative result is indeed a true negative or a false negative depends on whether or not you have a low levels of virus in you. If you do, then clearly you’re infected, and the result was a false negative. And what increases the probability that you have any levels of virus in you? Any exposure, behaviour, or risk that might increase your pre-test probability of infection.
So, to repeat myself: a negative result from a rapid test only really has meaning if you had no prior reason to expect to be infected. Therefore the test should only be applied to asymptomatic people with no known recent exposures.
(2) They must be repeated frequently. Let’s assume that the probability of a false positive or false negative is 10%. That’s high. But guess what? If we repeat the test (preferably within 24 hours) then the probability of a false negative or positive becomes 10% of 10%, or 1%. Repeat the test again and it drops to 0.1%, etc etc. This is called orthogonal or serial testing.
If you get two different results, you should then proceed to a PCR test for confirmation. If you keep getting the same result, the probability that that result is false drops by an order of magnitude each time.
(3) Recipients must be made aware of the limitations of the test, and that a negative result is not a license to attend an orgy or doff your mask.
Where Would I Use Rapid Tests?
Two places come to mind right away: schools and homeless shelters.
In schools, almost all infections will be asymptomatic, given what we know of COVID and children. And since symptom checks are already in place, most people attending schools will not show outward symptoms. The use of rapid tests for this population is therefore ideal, as the rate of false negatives will be low. Either all or a sample of students and workers would be tested repeatedly over the weeks. All negative results could be considered genuine if paired with a negative symptom check. And all positive results could be considered genuine since the rate of false positives is very low indeed.
Homeless shelters are a bit more problematic, since the pre-test probability of infection would be higher in this population. But by layering a symptom check and increasing the frequency of re-tests, the probability of false readings could be reduced.
Reassurance testing for workers is also a good application. In the UK, rapid tests are used for this purpose en masse in several workplaces. Imagine your workplace instituting a rolling random selection of 25% of the workforce getting tested every 48 hours. That’s not so bad, right? It’s not perfect –some cases will be missed. But it might be good enough to get a handle on outbreaks, especially if any detected positive cases are followed up with contact tracing and more intensive testing.
Other good targets would be quarantine-exempt incoming travelers, such as truck drivers. Mandatory rapid tests at all gas stations, truck stops, or weighing stations could help detect COVID carriers where simple symptom checks are insufficient.
I would not recommend rapid testing for long term care residents, for a number of reasons. First, many of these populations have already been immunized. Second, they are already over medicalized and likely would not meet the low pre-test probability criterion. Third, and most importantly, these centres are already stressed in terms of health care worker availability and workload. Requiring repeated rapid test swabbing would just add to the workload unnecessarily.
However, a rapid test would be useful for incoming visitors. Ideally, such individuals would pass a symptom check, come with proof of a recent negative PCR test, and submit to a rapid test on site. That way we would catch that handful who developed testable viral load in between their negative PCR swab and the visit.
The issue with rapid tests is that they can’t be used to replace PCR tests in all their functions. They are best applied as surveillance or reassurance testing tools, where asymptomatic people with no reason to assume they are infected are tested repeatedly to ensure their negative status.
Our inability to use them at scale thus far is due to two things, in my opinion. First is a failure of imagination among some (not all) planners who only see tests as diagnostic tools. Second is a largely paternalistic view of the population as being unable to be trusted with knowledge of likely negative status.
I think that we are far enough along this pandemic, and still in the thick of things, that all tools must be on the table. It’s time to let rapid tests have a go at bringing normality to a few institutions.