Gaming the Metric and Unintended Consequences

 In the UK, a "tiered" lockdown system has been implemented in response to the alleged COVID-19 pandemic. The state of New York in the US has a similar "microcluster" system. While the details and the labels might vary between the two, both are driven by a similar metric - the ratio of positive test results to the number of test results taken for a given time slice. 

At first blush, this would seem a reasonable proposal. But the devil is always is the details, and it's the hidden leak on the O-ring that will kill you and your crew. The "percent positive" metric makes sense only if the sample of people is normalized and the test is fairly accurate. Any biases in the input to the statistic will show up in the analysis. 

So, who is getting COVID tests? While some number of people are getting regularly tested occupationally or if they are high-risk, most people will not get tested unless there is a reason to. Reasons such as

• Travel, and the government or your employer requires it
• Symptoms are shown
• Close contact with an infected individual

And many others - all of which increase your probability of becoming infected with the virus. So, in the pool of individuals being tested, there will be an over-representation of people who have a somewhat higher probability than the "baseline" of being infected. So, that is the population that the "percent positive" ratio is being derived from. 

When researching the reliability of the most popular form of COVID-19 rapid tests, what I found was as clear as split pea soup. The most wild range I found in one article was "63-94%" - a statistic itself which is completely useless. Another source claimed 91-94%. In fact, every source claimed a different range. The most accurate answer and analysis I can provide is "it depends". So I will call the reliability "questionable" in my analytical calculus - without bias as to which way the uncertainty would break. 

To review, that leaves us with a slightly skewed sample (in favor of finding infection) with a test of questionable reliability. In the case the test were overly sensitive, we would see an increase in the "percent positive" cases that was unexpectedly large, and very hard to "fight". We would also see an IFR that would have a negative correlation with the percent infected metric - that is, the lethality of the infection would seemingly decline while the contagion exploded in the population. 

Note that viruses will also do this in human populations shortly before herd immunity is achieved. 

If the testing were biased to come back negative, what we would see would be a "slow burn" percent infection ratio - perhaps dropping away to the point of some politician declaring victory against the virus! But - there'd be a stubborn problem with Flu-related deaths all winter. Perhaps it would be correlated with people who vape THC. 

 So what is one to do? That's simple - if you are less than 60 years old with no complicating factors, skip the testing unless you absolutely have to. The likelihood of a false positive is nonzero in a population that already has a skew towards infection. If that ratio is being used to deny people work, money and freedom, the only sane and rational thing to do would be to avoid any and all COVID-19 testing. This would deny the governments the data they need to justify the draconian intrusions on your liberty.  

The phrase "PCR should never be used for diagnosis" is one that you'll hear quite a bit if you listen for a little bit. 

Why? There's a lot to digest, but a good start is here. Cycle thresholds, false-positives, oh my! That link is a little redpilled!

This means that if a person gets a “positive” PCR test result at a cycle threshold of 35 or higher (as applied in most US labs and many European labs), the chance that the person is infectious is less than 3%. The chance that the person received a “false positive” result is 97% or higher.

(Note that the exact figures depend on the test and lab in question, and that if a sample was already positive at a lower cycle threshold (e.g. 20), chances of infectiousness are much higher.)

Why do labs use such high ct values? From a lab perspective, it is safer to produce a “false positive” result that puts a healthy, non-infectious person into quarantine, than to produce a “false negative” result and be responsible if someone infects his or her grandmother.

However, a negative result at a cycle threshold above 35 still does not exclude a covid infection, as a false negative result may arise if the sample is taken improperly or too early. More recently, US researchers found that single-gene tests were false-negative due to new virus mutations.

That's nuanced, and a bit to digest. But, you can game the metric by tuning the test to be more sensitive, and claim you're only protecting the Most Vulnerable Among Us©  

Be skeptical when they tell you you're not allowed to leave your house.