A Non-Scientist?s Primer on Testing for COVID-19

A Non-Scientist?s Primer on Testing for COVID-19

First some important caveats:

  • I?m not a scientist. I have been science-adjacent, professionally and by marriage, for some time, and have done years of work together with some very accomplished scientists. But nothing I am writing here is qualified by formal medical or scientific training on my part.
  • Nothing here is peer-reviewed, and what we learn and know is changing quickly. I am doing my best to be punctilious, and embrace any suggestions on incorrect or missing or outdated information. I will edit this over time.
  • I work for Crossover Health, but this is not a formal communication on Crossover?s behalf, nor does it necessarily represent any of its views or positions. My motivation in writing this is personal, seeing a lot of confusion about what testing is and needs to be, and I am hoping to help inform the discussions with a little bit of scaffolding. This is my personal effort and there are no commercial interests represented in anything I write here.
  • There is a lot more to know about testing, but this is not meant to be exhaustive. It?s just meant to dispel a lot of the confusion I see online. This was written in a few bursts of periods while on hold with customer service, trying to take care of our household. Everyone has to fill the time.

With that, a non-scientist?s primer on COVID-19 testing.

Headline: There are two types of tests

Broadly, there are two types of tests, doing two very different things in two different ways. There are others, also, but these are the two you?ll be doing, and hearing a lot about.

There?s 1) diagnostic tests (sometimes called ?tests?), and 2) serological tests (sometimes called ?antibody tests.?) They?re easy to confuse if you?re not in this all day. A friend just got an alert on their phone ?free Covid testing at Rite-Aid.? Everyone needs to be able to ask the right questions when they get messages like that. These tests will shape your life in the months and years to come.

Diagnostic tests

An introduction

Diagnostic tests are looking to detect the active presence of the virus. In simple language: are you currently infected, and likely-infectious with Covid? It?s entirely possible you are fully asymptomatic and just walking around with it; a study of routine diagnostic testing of pregnant women coming in for delivery in NYC hospitals revealed that 14% of pregnant women were actively carrying and spreading C19 (to one degree or another) all of whom were broadly asymptomatic. Let that sink in. Anecdotally, I know people who work labor & delivery in hospitals in other cities where they have seen days when it hits 50%.

Diagnostic tests pick this up. In premise. They do that by taking a swab from your nose or throat, and working that up through RT-PCR. The idea of PCR in general is new to many people, but the bottom line is that it needs a lab with some meaningfully fancy equipment in it, or some clever workarounds (which are rolling out, solving some problems and introducing others.) It can generate a result quickly enough, call it 15 minutes, but the equipment and its technician needs to be in direct proximity (as in, at the hospital, or the clinic) or alternately, the swab result needs to get to the equipment. And then the person needs to be free to run the test.

If you hear about ?Abbott? testing in the news, that?s probably diagnostic testing; if you hear about drive-through testing, that?s generally diagnostic (tho increasingly, maybe not!) If it involves swabbing you ? gathering mucous or spit ? it?s diagnostic testing to see if the virus is present.

One significant point about diagnostic testing, speaking generally: it?s a highly specific method, producing very few false positives (that is, it doesn?t frequently say you have it, when you don?t.) But?? as it has been and is being practiced?? it?s a low sensitivity method, so there?s a lot of false negatives (that is, it may frequently tell you that you don?t have it, when you do.) I?ll explain why in a moment.

The latest Abbott testing (one of the workarounds to using conventional PCR) only catches it about 82% of the time. To give a specific example, it means that a friend in Berlin with unambiguous C19 symptoms has tested negative twice in a row? even as she presents unmistakable evidence for a C19 diagnosis in the ICU. In current conditions, this will happen to roughly one out of five people who are walking around with it, or struck down by it.

The diagnostic tests are way, way better than nothing, but they?re not perfect. Many things vary from country to country (duh) but the overall precision of diagnostic testing does not vary substantially from country to country at the moment; there?s no silver bullet in one place that?s missing from another.

Where it happens

Broadly, diagnostic needs to be done either in a clinic (which poses some varied degree of exposure risk to everyone involved) or in a drive-through setting (which has some problematic demographic consequences), or at home (in principle, and see below.) Some tests can be analyzed at the point-of-care; most have to go back to a lab; you can go deep on this if you?re into it.

Clinics may well be environments where infectious C19 carriers are putting the virus directly into the air. Everyone can be cautious, but the bottom line is that transmissibility in closed-air environments is a tricky problem to solve. And most recently, a friend in LA presented symptoms, got a clinical diagnostic test, and it had a 9-day backup in the test queue. Symptomatic or asymptomatic, nine days of Covid is a long time.

Very recently, the FDA approved an ?at-home? diagnostic test. My personal opinion is that this is not a good use of everyone?s time, effort and precious resources as it?s currently presented. All you?re doing is getting a kit sent to you (2 days?) and swabbing yourself and then ? going to the post office? to send your sample in the mail. Introducing a new massive burden into the already-overburdened USPS isn?t an obviously good idea, and the dynamics of getting samples to a post office or even a mailbox may be problematic for many people, especially those with symptoms.

The at-home test gets back to the lab (2 more days?) and then gets analyzed there (perhaps in a queue) and then eventually someone gets you an answer. By then, if you are symptomatic, you?re pretty clear you have Covid. And if you?re asymptomatic, you?ve had 9? 12? or more days to infect everyone around you.

Overall, I am not optimistic about the individual or collective utility of diagnostic at-home testing as it is currently constituted, but perhaps someone can change my mind about that.

In addition, the current means of gathering enough material for analysis require a rather uncomfortable swab deep into the nasal cavity or the throat. There?s now a saliva-based test that seems to be better in a few different ways, but overall I am not optimistic about the capability of American individuals (in particular) to meaningfully follow careful instructions about getting it done right.

That said, in the last 48 hours it has emerged that testing with saliva as opposed to swabs might have been a better idea all along. The problems with nasal and throat swabs might also (as it turns out) have been giving kind of weird results especially two weeks into the infection (namely, in throat swabs it sometimes dips below the diagnostic threshold and then comes back up later. Weird.) That?s how quickly this changes. This whole post will be out of date by the time you read it.

Bottom line, also is that diagnostic tests require machines and people to analyze the results. Today (April 23) Cuomo announced that if New York State used all its machines and all its people 24/7 just running diagnostic tests, it gets you to 40,000 tests/day. Is that a lot? A little? Remember that diagnostic tests are just a moment in time, and that no single test will ever test positive for a person who gets exposed the next day.

What it can mean

Diagnostic testing is very helpful to those who get the right answer, at the right time. But the high number of false negatives in current practice can be a problem for individuals and a different kind of problem for public health modeling and consequences. Symptomatic people who get a negative answer can?t assume that negative answer is correct. Nor can asymptomatic, or pre-symptomatic people.

But those who get a positive answer, especially when pre-symptomatic or asymptomatic, can take additional caution to prevent infecting others, and can also possibly take some personal actions to mitigate the effects of c19 (which I have no advice about.) That said, they would have to get that answer quickly enough to act on it. By some models, that needs to be in the first 48 hours of infection.

The false negative problem may also get better. The abbott tests have an 18% false negative rate but that?s in part because it?s a novel (and proprietary) process. Running ?proper? PCR testing (what the Abbott tests were designed to replace) is a lot more time and resource intensive, but is one part of cleaner results. And the shift to using saliva vs nasal swabs is likely to resolve a lot as well. Again: all these high-leverage details are emerging in real time. Any one of them can put holes in the boat, or be the wind in it?s sails.

Diagnostic testing is crucial for individuals and for populations. But it?s currently far from 100%, and its applicability for individuals and populations needs to take that into account. The bottom line is that, in particular for asymptomatic and presymptomatic individuals, a diagnostic false negative could be consequential.

Then there?s serological testing.

Serological testing


Serological testing is not detecting the presence of the virus; it is looking to detect the antibodies that are (or were!) produced in fighting it. There are a few different immunological responses that can be tracked, but for contemporary purposes, what?s called ?IgG? is what is given priority in detection (after IgM; here?s a paper on how they are produced in SARS, which is not the same, but the use and principle of IgM and IgG is similar).

To get fussy for a moment, if you test positive for IgM antibodies ? as opposed to IgG ? it means you?re actively fighting the infection, and likely to still be an exposure risk. But it?s exactly the point that you don?t produce those antibodies right away. IgM declines as you recover. IgG starts being produced 10?14 days into the infection and when people talk about testing for whether you?ve ?been exposed at some point? it?s probably IgG they?re using to determine that.

Bottom line is that this is a test that tells you whether you have at some point been exposed to C19, and if indeed recovery produces immunity, that?s important to know, especially with the high number of asymptomatic cases. Further, with the incentives to allow people to leave their homes without risking exposure to anyone (or being exposed) then if recovery = immunity this is a valuable piece of individual and collective data.

How it?s done

Serological testing is done with blood. Venous draws are best, and it?s not 100% clear ? yet ? that capillary draws (finger prick) produce enough blood for meaningful analysis, though many well informed scientists believe capillary draws are sufficient. The knowledge around this is evolving real time.

In RDT (rapid deployment testing) ? the more common form of serological testing insofar as we have common forms ? blood is placed in a device like a cartridge (typically microfluidic, which is interesting if you know what that is and doesn?t matter if you don?t) and exposed to a test strip coated with the C19 antigen. Like a pregnancy test, if the test strip looks one way, there?s the presence of antibodies. If it looks the other way, there?s not. No need for a lab, and the results show in about 10?15 mins.

There are other forms of ?non-rapid? serological tests that do require labs, notably ELISA testing, which is familiar to anyone who?s ever dealt with the HIV epidemic. It?s an old methodology that produces reliable data at an individual level.

RDT and lab based serological tests can, in premise, be far more accurate than the diagnostic tests, but we?re still learning how accurate. A recent (April 14) letter from an NYC health official states ?only three serology assays have received FDA Emergency Use Authorization (EUA)?documents for all three tests include disclaimer information regarding possible and demonstrated cross reactivity with common human coronavirus antibodies.? This will all likely change quickly, and is one thing in the world that will likely get better soon.

Where it happens

The blood draw means that like diagnostic tests, it needs to be in a clinic, or a drive-through (some distinguish open air testing from clinical, but for these purposes I will treat them as the same.) The FDA has thus far denied all applications around at-home serological testing even thought the RDT serological tests are simple enough and give results without needing any kind of lab. The rationale is likely that it?s too easy to mess up a blood draw if you?ve never done one before (though many people do them just fine, like diabetics) and it?s too easy to corrupt the test? maybe even cross-contaminate others in the household, who knows. There were 30 cases of people who poisoned themselves with bleach just yesterday, in NYC. This is America.

If you mess up your blood draw for diabetes monitoring, you affect your personal health. But if you mess up your blood draw for C19, it?s a matter of public health. I don?t disagree with the FDA position on this, though other countries (maybe Denmark?) do allow at-home serological testing, and there?s arguments to be made for it, for sure.

There are a lot of bullshitters in the serological testing space in particular, and many get shut down as quickly as they shoot up. The test is easy to do in premise; but the science, manufacturing, verification and logistics are properly difficult. This April 24 NYT article makes clear that not all the serological tests are to be understood as equal, and few should even be considered valid.

That said, results ? when they come through verified antibody tests ? are pretty reliable. As opposed to diagnostic testing, serological testing is generally highly sensitive (very few false negatives) sometimes with tradeoffs on its specificity (it sometimes produces false positives.) The first one approved, Cellex, was around 94% sensitivity and 96% specificity, which is high for both. There is some complexity in interpreting this, around who it?s been tested with, and who it?s being used with now. If you want to go deeper on why this isn?t an absolute value, here?s a tweet from a month ago (when the first antibody test was approved) and a youtube explainer on Positive Predictive Value.

Ideally, a positive result in any individual would be retested to confirm it. In the decades old of use of ELISA for serological testing of HIV, for example, a positive result is routinely confirmed with a second (different) test, precisely because of the false positive problem.

The bottom line is, for an individual who has never been exposed to the virus, a serological false positive is consequential; they may think of themselves as ?recovered? and ?safe? and still be as vulnerable as anyone else.

Again, if recovery from C19 yields immunity, serological testing could be instrumental in reassuring essential workers (for example) that they are not risking exposure to themselves ? or others ? as the world finds strategies to open up responsibly. But doing serological testing in clinics and drive-throughs is fraught, and the FDA isn?t likely to approve at-home serological testing as it?s been presented thus far.

Mass scale serological testing is also valuable just in order to understand the broad epidemiological patterns of C19: at the scale of the world, as well as nations, cities, workplaces, or anywhere on earth with a vulnerable density of humans.

I?m working

I?m working with some brilliant folks on a possible path to solve for a tricky aspect of this, but we?ll see. I hope to be able to talk about it soon. But in the meantime I saw four posts today that were confusing one type of test with another ? fair enough! ? and thought that it might be useful to try and lay it out. Today, as I?m writing this up, Cuomo presented a bunch of stats. It?s hopefully useful to know how to contextualize them.