In all of the times that we thought about a pandemic, in some hazy future that would probably never happen, the one thing that we mostly never foresaw was the truly epic amount of nonsense that would be produced. It seems like every second minute that a new myth about COVID-19 springs up that needs to be debunked, whether it’s the idea that drinking hot water can kill the virus (remember that from way back in April?) to the newer absurdity spreading around that most coronavirus deaths were actually caused by buses (spoiler: they weren’t).
And so, when the newest misconception sprang up, instead of feeling elated that my purpose as a science communicator and blogger was being fulfilled, I instead felt the bone-crushing weariness that comes from total nonsense going very every single day of the year. And because this nonsense is about a pandemic disease, and policy makers do sometimes base their decisions on trash analyses published online, I felt like it’s important to address this inaccuracy as soon as possible.
Which brings us here, to the claim: that 90% of all positive COVID-19 tests are false positives. This has serious implications, the story goes, because it means that there are actually virtually no cases of the disease in the country, and the government is lying to us all*.
Which is, to be blunt, totally wrong.
Let me explain.
Testing For COVID-19
The main test used to determine if someone has been infected with SARS-CoV-2, the virus that causes COVID-19, is called a polymerase chain reaction test or a PCR. PCR testing basically takes tiny amounts of viral DNA or RNA and amplifies it until you can detect it easily in a sample, and is one of the main ways we test for specific infections.
PCR testing has a number of advantages, as well as some weaknesses. The test has a problem with identifying people who are in the early stages of infection. This is because it can be hard to get enough virus on a swab to replicate it in the test, which means that people who have only been infected 1 or 2 days ago may not be positive when they are tested. In fact, the sensitivity of the test can be as low as 40% in people who have only been infected a few days, meaning that 60% of tests are false negatives in this population. This is why many places will retest you if you are still having symptoms after a few days, even if your initial test was negative.
But in this blog, we’re not focusing on false negatives, we’re worried about false positives, because a few people having their COVID-19 missed is only worrisome if we care about public health rather than strange conspiracies.
So, false positives.
With the PCR tests used to detect COVID-19, the average number of false positives has been extremely well-calculated. We have been using these tests for decades, after all, and we’ve had the better part of a year to perfect them for this specific disease.
While there were some early validations that showed slightly higher numbers of false positives, we’ve now got enormous samples run by researchers from across the world that show the rate of false positives in PCR tests for COVID-19 to be around 1 in 1,000 or lower.
In other words, it is incredibly rare for a COVID-19 test that comes back positive to be a false positive. Almost unheard of. It happens, but very rarely, particularly because many tests are confirmed, i.e. run twice.
We can actually use two numbers, called specificity and sensitivity, to see the exact rate of tests that are ‘right’ in a population. The specificity of COVID-19 PCR tests is the ratio of true negatives to false positives+true negatives, which works out to about 99.9%. In other words, for every 1,000 people you test who truly don’t have the disease, you get 1 test that is falsely positive. Sensitivity is the ratio of true positives to false negatives+true positives, and while this is also high it’s not quite as high as specificity, averaging out around 98% or thereabouts. This means that for every 1,000 people you test who do have a coronavirus infection, you’ll miss about 20 true cases.
Let’s imagine a population where about 50 in 1,000 people actually have COVID-19. You test all of them with PCR tests. Of the 950 people without the disease, you’ll get an average of one false positive test. Of the 50 people who do in fact have the disease, you’ll miss 1 person, which means 49 true positives. In this case, 98% of all positives are true positives, with just 1/49 tests being a false positive. You’ll also have missed one true positive, so your estimated prevalence — the proportion of people who you think have COVID-19 — is actually correct at 50/1,000.
Even with very low numbers of infections, the spectacularly high specificity of PCR tests means that you almost never have more false positives than true positives. If we repeat the calculation for a situation where only 50 in 10,000 people have COVID-19 in our sample, we still find 49 true positives and 10 false positives, which means our ‘correct’ percentage is now 83% — still very high, given the low population prevalence in the people we’re testing. In real life, it’s rare that so few of the people we test actually have the disease, because we target testing at those with symptoms and who think that they might be getting sick.
And we can see this incredible accuracy happening in real life. In Australia, despite hundreds of thousands of tests conducted every week, there are vanishingly few positive results. In New South Wales, the state that I live in, we conduct more than 115,000 tests every week with <40 positive results. Even if every one of those were a false positive, the false positive rate would still be less than a fraction of 0.1%.
More broadly, it is simply wrong to suggest that most tests coming back positive for COVID-19 are false positives. In reality, the simple truth is that most positive tests are just that: positive. The number of people who are sick with COVID-19 is increasing in many places in the world — particularly the UK — despite what certain inaccurate blogs might have you believe.
This is the sad truth, and nothing more.
*Note: It is always good to be a bit skeptical when people tell you the government is lying to you, because this presupposes a government that is coherent enough to settle on a single lie rather than the usual reality of so many garbled responses that even people working in the department have no idea what the message is.