How accurate are the rapid at-home COVID tests?

Now that 15-minute home tests are available, millions of people can get a quick reading on whether their symptoms are due to Covid-19 or something else.

The at-home tests, although very fast and convenient, are less sensitive than the tests available at most Covid-19 testing centers. The home tests are “antigen” tests, while the gold-standard tests available at medical facilities use another technology, RT-PCR. (I’ll explain a bit more about the technical differences at the end of this article.)

An at-home test is far preferable to leaving your home, possibly exposing others to the virus, and then waiting hours or overnight for test results. The question is, how accurate is it?

Very accurate, for the most part. The chance that you’ll get an incorrect reading from a rapid antigen test is less than 1%. But (there’s always a “but”) it depends on what you mean by accurate. By another measure, they are not quite so reliable.

Let’s dig into the numbers from two new studies, which looked at thousands of cases, and see what they tell us.

In the first study, published recently in JAMA, Joshua Gans and colleagues from the University of Toronto looked at over 900,000 rapid-antigen test results. As expected, the vast majority of the tests were negative, and only 1,322 (0.15% of the total) were positive. They collected results using more-sensitive PCR testing for 1,103 of the positive tests, which allowed them to check whether or not the rapid antigen tests were correct. (This assumes the PCR test is always right, which isn’t quite true either, but it’s a good approximation.)

Surprisingly, 462 (42%) of the positive results were negative when double-checked with PCR. In other words, these were false positives. The investigators tracked these tests back to the source, and they discovered that about 60% of the false positives all came from a single batch of the Abbott Panbio Covid-19 rapid tests. So apparently there was a lower-quality batch that yielded more false positives.

However, another way to look at this data is that the overall rate of false positives was still very low. Out of more than 900,000 tests, only 462 were false positives, which yields a false positive rate of just 0.05%.

Another way to explain this is: if you just walked in off the street, your chance of getting a false positive test was about half of 1%. But once you saw that your results were positive, the chance that the positive result was correct was only about 60%, at least in that study.

One thing lacking in the Toronto study is that the investigators didn’t test everyone with PCR. They only used PCR to double-check the positive tests, so that study doesn’t answer the question of how many infections might have been missed.

Fortunately, the second study answers that question.

In this study, released in late January as a preprint on medRxiv by a group of my Hopkins colleagues led by Zishan Siddiqui, the investigators looked at 1054 participants, and tested all of them with both a rapid antigen test and RT-PCR.

Even though this study looked at far fewer subjects (1000 versus 900,000), they checked everyone with PCR, which allowed them to measure both sensitivity and specificity; i.e., they could count how many infections the antigen test missed.

So how good was the rapid antigen test in this study? First, its sensitivity was 92.7%, meaning that it correctly identified 92.7% of people who had Covid-19, whether or not they had symptoms.

What about those false positives? Here the news was better than the Toronto study, but still far from perfect: about 28% of the positive results from the antigen test were false. That’s better than the 42% found in the Toronto study, but it still means that many positive results from the rapid tests turn out to be incorrect.

What’s the take-home message from these studies? Well, I’d summarize it in three points:

1. The rapid antigen test is generally very accurate, and certainly worth taking if you have any reason to think you might have Covid-19. The chance that you’ll have an incorrect reading, either positive or negative, is very small, less than 1%.
2. If you get a negative test, you can relax: over 99.5% of negative results are correct, meaning you truly don’t have Covid-19.
3. If you do get a positive test, you probably have Covid-19, but there’s still a roughly 30% chance that you don’t. If possible, you should immediately get a followup test using RT-PCR, which is more accurate.

Addendum: for those who want to understand the difference between the rapid antigen test and RT-PCR, here’s a bit more on those.

The rapid antigen test contains molecules called antibodies that bind to a specific molecule, the nucleoprotein, which is present on the surface of the SARS-CoV-2 virus. The antibodies in the test kit are designed to bind to other molecules that create a small band of color on a test strip, so you can see the results as a colored band in just 15 minutes. The main drawback of these tests is that they sometimes fail to detect the virus. For more details, a good description can be found here.

The RT-PCR test detects the RNA that is the genetic code of the SARS-CoV-2 virus. Every virus particle contains this RNA, which is a sequence of about 30,000 “letters” or nucleotides. In RT-PCR tests, we first convert the virus’s RNA to DNA, and then amplify it to make millions of copies. Because this test uses an amplification step, it can detect tiny amounts of virus, which is why it is more sensitive than rapid antigen tests. However, it takes at least a few hours to run this test, so it’s not as fast as the antigen test. For more details about how these tests work, check out the NIH explanation here.