Sunday, 3 May 2020

Day 48 of self-isolation - how good are the tests?

 How good are the tests?

First, some good news. There was some concern that catching Covid-19 might not give you immunity in future - this was based on 292 reinfections in South Korea. It has now emerged that this was a testing fault. After you're infected, bits of virus can linger in the body for months - but you aren't reinfected, it just means that your body hasn't completely flushed the dead virus bit out yet.

The UK government is now testing 100,000 per day. But my first question was, how accurate are those tests? Specifically, how many false positives, and how many false negatives? Because the word "accuracy" doesn't tell me what I want to know. I want to know two numbers; false positives and false negatives. It is exactly the same as a computer antivirus.

It's easy to make a test that has no false positives - you just say "You're clean" to everyone. And it's equally easy to make a test with no false negatives - you just say "You're infected" to everyone - but then you get a lot of false positives. What is difficult, is to make a test with a high accuracy - with a low percentage of false negatives, and a low percentage of false positives. So when the government announced mass testing, the press concentrated n the number of tests, which is easy to understand, and totally failed to ask about accuracy.

So I did a bit of research.

First, understand that there isn't just one test. Anyone can make a test. From the AA Acme Company to Zygote Incorporated.And you can be sure that some tests are better than others - and some might be from complete cowboys (we saw that in the early days of computer antivirus software)..

The FDA is silent on test accuracy, so in the USA some independent scientists looked into it.

Three tests showed zero or 1% false positives. But those tests gave 10% or more false negatives. A false negative means that you say someone is clear when he isn't.

Four of the tests produced false-positive rates ranging from 11 percent to 16 percent; many of the rest hovered around 5 percent.

In Europe, the Abbott antibody test (testing whether you've had Covid-19 in the near past) has been given a CE mark, but that only means it's safe, it says nothing about accuracy. Abbott (who are a long-established diagnostics company) claim that it's 99% accurate (meaning, at most 1% false negatives, and also at most 1% false negatives). If it really is that good, then I'm very pleased, but I'd like to see independent confirmation of that..


  1. We need S&S International to the rescue !

  2. Wrong sort of virus :-(

    But I can say that some of my understanding of computer virus has translated nicely to medical viruses. For example, the epidemiology post that I did a couple of months ago, is very like one that I did in 1989.