FC Bayern Forum

MUTU wrote:

aterford wrote:According to Sport Buzzer (Not sure about them tbh), Sule's test was also a false positive, somehow.

Which - in typical Bayern luck - probably means that we'll get the worst of both worlds and he'll miss Salzburg and BVB while being fit and ready to return for international matches.

If it's really 1 in 10k that are false positives, then it's 1 in 100 million to get 2 false positives, albeit that's the chances of getting 2 false positives in a row. We don't know the total number of tests our team had.

PCR is being used as a screening rather than a diagnostic measure here since the players are regurlarly doing it while asymptomatic
Screening tests have SENSITIVITY and SPECIFICITY. What’s the difference? If you place the cut off point for defining a positive test pretty low, the likelihood of a false negative is is also low…ie will probably catch most/all cases. This means that the test is very sensitive at detecting disease. The downside of that is obvious, however. You will also have alot of false positives.
This is also true in reverse. If I use a high cut off point to define a case, I will probably get alot of false negatives as they will not fit my very strict definition of a positive case, but very few false positives ie the test is very specific.
This sounds confusing, but is better understood with an example. The cut off point for defining diabetes using fasting blood sugar is 7mmol/L.
If I decide to lower the cut off point to 6 (ie make the test more sensitive but less specific), I can expect to hardly miss a case. I will catch them all perhaps! But, the downside is I will probably also label alot of nondiabetics as diabetic (have alot of false positives). On the other hand, if I put the cut off point for defining a positive test at 9 (ie make the test more specific but less sensitive), then I can be sure that anyone that comes with a fasting blood sugar of 9 is probably diabetic (false positives are low), but I will also end up missing alot of true diabetics by placing such a high cut off point (false negatives are high).

Ideally, a test would have the best cut off point where you can balance between its sensitivity and specificity.
Unfortunately, this shit is not done yet.
You also have something called the "positive predictive value (PPV)" and something else called the "negative predictive value (NPV)". These answer whether or not you can trust someone’s positive test (trust a pos test if the PPV is high) or negative test (trust a neg test if the NPV is high). The sensitivity and specificity of a test are always fixed, but the PPV and NPV are not and they depend on the person’s likelihood of having the disease ("pretest probability") & also the test’s own sensitivity/specificity. The PPV is high if the pretest probability is high and the specificity of the test is high. The NPV is high if the pretest probability is low and the test sensitivity is high.

Now, back to Süle and PCR
PCR for COVID has a sensitivity and a specificity of >95%.
But the pretest probability of having Covid here is very low. The players are asymptomatic and one can assume that they also have no risk factors for covid (ie they are wearing their masks, not going about in crowded closed areas, Germany as a whole has less cases than the US, etc).
The low pretest probability will lower the PPV. This means that you may not necessarily trust a positive test. It will also make it so that a negative test is very trustworthy.
Some examples and math help demonstrate all these points, but I think that may take things way too far.

Now, you may ask: if we cannot necessarily trust a positive test, then why does Süle have to self isolate etc? BECAUSE IT’S A VERY CONTAGIOUS DISEASE, WE ARE IN A GLOBAL PANDEMIC, AND WE OUGHT NOT TO TAKE ANY RISKS!

If I were to put this in another way….
PCR has about an 0.8% false pos here for covid.
In a general population, if I screen 1000 assympatomatic individuals, the latest figures say about 1 will be a true positive.
With a false positive rate of 0.8%, that works out to 8 false positives in 1000. So I will end up with a total of 9 positive tests, only 1 of which actually has the disease (so 88.88% of my positive tests are false positives!!!!.

That’s also why testing is not recommended for random asymptomatic individuals in the population.
The situation would be totally different if I was testing someone who was a close contact to a known case. In this case, the pretest probability would be high and you can trust a positive test.