Considering they claim we exhale the virus to infect other people, I wonder why they cannot diagnose it off that.
This is a bit complicated. The short version is what
evildice just said, but the actual, technical details depend a bit on just what test you're talking about despite distinct parallels.
The first-line (and still
best, going just by accuracy) tests are/were based on PCR --
polymerase chain reaction -- methodology, which is a method used to rapidly make millions or billions of copies of a piece of DNA (or, indirectly, RNA). The problem with them is that PCR is a hilariously finicky method that requires strict laboratory conditions to run due to its ridiculous levels of vulnerability to random contamination (which is a problem for reasons that you can probably figure out from the above -- just think about what happens when random bacteria get into the sample you're running through the "make millions of copies of this" reaction, then multiply it by everything *else* that could get in there). Even
with said laboratory conditions and the strictest guards against contamination possible, it's not at all uncommon for random bits of fluff -- including whatever you're looking for -- to get in there
anyway.
This is -- assuming a competent lab, and I know this stuff because I had to spend several months dealing with lawsuits involving
incompetent ones, several years writing about the deliberate use of said incompetent labs by scientific frauds in various forms, and close to a fucking
decade after that explaining said writing's contents over and over again, usually to the same people as they ignored me and instead listened to hucksters -- mostly fine; labs have a very simple way to deal with this. Simply put, there's usually going to be a shitton more DNA in that sample from legitimate sources than from contamination.
Thus, you run the sample through the process a strict number of times (with the DNA in the sample theoretically doubling each time). While you don't really (directly) look at the quantity curves in methods used for a COVID test, there's a related method called
quantitative PCR which
does, and it produces quantity curves like this:
It's worth noting that DNA
other than the target DNA (transcribed SARS-CoV-2 RNA in this case) can set off the detection instruments if there's enough of it... but let's not get ahead of ourselves.
Anyway, now that I've probably lost my mental place due to taking the time to find the above image (which was a bit harder than you'd think), let's get back to what happens more realistically thanks to the fact that there are almost certainly multiple types of DNA in that sample (due to it being taken from somewhere outside a laboratory). Specifically, you have multiple curves developing, producing a figure that looks something like this:
Where the "earlier" curves are the things we originally had a lot of in the sample, and the "later" curves are things that we originally had less of in the sample.
(And yes, that's a real test curve.
Here is another one, that's far more realistic for an actual sample in this context, despite being of something else entirely.)
All of this means that the main trick when using PCR to search for viral D- or RNA is to run the process through enough cycles that the thing you're searching for will show up if it's meaningfully there, but
not enough that random contaminants will throw out false positives if it's not
.
But... levels of viral RNA vary a good bit between samples, even from people with the disease. They vary between individuals, between sample types (a saliva sample might have virus particles, but far fewer of them than you'd get from a nasal swab), and even between individual samples where the above are identical. Heck, they even vary depending on how the samples are
stored.
And
that means that there's a trade-off: The more cycles of the chain reaction are run, the more likely it is that you'll find the disease in a sample has it... but more cycles
also mean the test is more likely to throw up a false positive.
(This trade-off between
sensitivity and specificity is not exactly going to be new to anyone who's familiar with test design... but I'm focusing on PCR-based testing here.)
Someone who wants an accurate test generally wants to run as few cycles as possible without sacrificing accuracy... which means that you want as virus-dense a sample as possible from the people you're looking for.
Breathing on a collector or the like simply wouldn't be dense enough, especially since the collection device itself wouldn't be taken or stored in a laboratory. You
could, I suppose, sterilize a clean room before collection, drop the patient off in there, have them open the collection device (alone in there), have them seal the thing themselves in an environmentally-controlled container meant to preserve as much virus as possible, and test things that way... but the practical and logistical issues with that should be obvious, and the results wouldn't be nearly as good.
And yeah, the other test methods have the same sort of issue. I just don't know them as well.
COVID transmission, meanwhile, follows what I jokingly like to call the "rule of sperm": Each particle is hilariously unlikely to do much on its own... but it really only takes one to make you into a virus baby factory.
Edit: On reflection, after posting this, a lot of the detail above isn't really needed. I thought of a better, and simpler, way to phrase things shortly after posting it. Still, the above details should be interesting/useful to the people reading, so I'm leaving the post intact.
