Can the US increase its capacity for COVID-19 testing fast enough to accomplish simultaneously:

1) slowing the spread of disease to allow hospitals to provide care for the infected, and

2) minimizing the impact on the economy by allowing social distancing measures to be relaxed (e.g. reopening schools and restaurants)?

Is there a testing strategy that can accomplish this and what kind of resources would it require?

See a related question here.

EDIT: The ACLA expects US labs to be able to do 280,000 tests per week by April 1st (acla.com). What if the full weight of the federal government was behind this ramp-up in capability?

EDIT2: Tom Frieden recently wrote (vox.com) that "testing is not a panacea," but he goes on to say that

"testing is absolutely, crucially important ...in areas with few or no cases, to inform containment and isolation strategies and facilitate contact tracing. Seattle would likely have had a much smaller outbreak if testing there had been widely available sooner."

The containment strategy right now seems to be evolving toward one in which the entire country is on lockdown. Could we do more targeted (but equally effective) containment if we had a clearer picture of where the virus was? Note: Conversely, Frieden warns against widespread testing in areas of confirmed community transmission.

  • 2
    better in medical sciences stackexchange, though the economics half doesnt belong there either.
    – Italian Philosophers 4 Monica
    Mar 18 '20 at 16:48
  • 1
    It's a good question, but much to speculative for us to be able to answer here, unfortunately
    – divibisan
    Mar 18 '20 at 17:19
  • 1
    This isn't really a politics question. It depends on the science of developing tests, the technology of producing & running them, the economics of their cost - obviously you can run a lot more tests if each costs $10 rather than $10K.
    – jamesqf
    Mar 18 '20 at 18:16
  • 1
    I agree that the title question is a good one, however the text in the discussion actually diminishes the value of the question. The author, by adding the elements of what he may think would flow from an increase in COVID-19 testing, begins to enter the realm of rank speculation. I'd suggest that the author edit the discussion to simply ask if there are any reliable sources (CDC, manufacturers, suppliers, medical labs) that have suggested that testing ( specimen procurement and analysis) can be increased and by what rate.
    – BobE
    Mar 18 '20 at 18:21
  • 1
    Testing in areas with few or no cases requires screening tests. AFAIK, the current tests are not meant for screening and may not perform sufficiently well to use them in situations with few cases - they are meant to be used on people with a substantial risk of actually being infected. If you are interested, you can read some background in my answers medicalsciences.stackexchange.com/a/21379/11479 and medicalsciences.stackexchange.com/a/21399/11479 Mar 18 '20 at 20:47

Well, I'm no epidemiologist and have neither the required data nor their simulation tools at hand. However:

While testing is of course needed, I doubt that more testing can compensate for lack of distancing now.

Also, there is Japan with social distancing but low numbers of tests with slow curve, and South Korea does not only test but also practice social distancing. There are also countries with lots of tests / inhabitant and heavy increases, Italy for example.

How are screening tests different from what is done now.

(Slight disclaimer: I'm in Germany, so I know most details about the strategy here. AFAIK, it isn't that different in the US, though. At least, also here people are complaining that they get tests refused and people are talking that test capacities are not unlimited - but it's not clear to me whether people are sent home who should be tested according to a doctor ordering that test or only those who go/drive to the test centers because they think they should be tested.)

What is done right now in most countries (including Italy, Germany, South Korea) is testing populations that are at a non-negligible risk of having contracted the virus. Looking at the current (Mar 19) numbers of positive tests among all tests (Germany) and using LR+ of better than 20 (below in parentheses)and expected 61 and LR- of better than 1/20 (below in parentheses), expected 1/61 (see long explanation here)

            | pos tests/tests | PPV         | 1-NPV
USA         |          11   % | 88 % (72 %) |  0.2  % (0.6  %) 
Germany     |           3.9 % |
South Korea |           2.8 % | 64 % (36 %) |  0.05 % (0.01 %) 

PPV = probability of having SARS-CoV-2 after the test was positive
1-NPV = probability of having SARS-CoV-2 after the test was negative

(There's the implicit assumption that everyone is tested with that same test - which is not true, the South Koreans by now may have better validation data, but I don't know this)

Screening means testing the normal population, which (still) has much lower prevalence of SARS-CoV-2. Assuming the dark number is 99 x the known case count (i.e. we have 100x as many infected as positive test results),

            | 100*cases/inhabitants | PPV            | 1-NPV
low risk    | 1 in 10000 =   0.01 % | 0.6 % ( 0.2 %) | <0.01 %
USA         |                0.36 % | 18  % ( 6.7 %) |  0.2  % (0.6  %) 
South Korea |                1.6  % | 50  % (25   %) |  0.05 % (0.01 %) 
Germany     |                2.0  % | 54    (28   %) |  0.03 % (0.1  %)
Germany*    |                0.4  % | 20  % ( 7.4 %) | <0.01 % (0.02 %) 

With that assumed large dark numbers, Germany and South Korea would both have prevalences by now where using the existing test on the general population would start to make sense, as that would mean only 1 false positive case per true positive case. On the other hand, with prevalence in the single digit percent we'd be in the region where Frieden argues that testing of the general population doesn't make sense because it makes more sense to tell people to stay at home and away from contacts (which is the case for Germany - and since the numbers aren't growing as fast in South Korea, their dark number is probably much lower).

In the low risk population (where Frieden says testing would make sense), we'd have only correct 1 out of about 200 positive cases. The PPV is for many practical considerations too low: screening needs very good tests, with very high sensitivity and specificity. Sensitivity, because it is crucial here not to overlook truly positive cases, and specificity because the true positives are otherwise hidden in army of false positives.

To some extent, it is often possible to trade in some sensitivity for speificity and vice versa, e.g. with HIV test for blood donors, we do exactly this: we choose a very high sensitivity (so we can be sure not to infect anyone), and accept low specificity. So low, that a blood donor with positive HIV test isn't even told that they were tested positive. Instead, another test is applied (preferably by another manufacturer, because that adds more information than duplicate testing with the same kit), and if that is still positive, yet another one (a more elaborate one). After that IIRC it is still more probable that the blood donor does not have HIV, but they's be looking into it and taking anothe blood sample. (And the decision for the donated blood is anyways to throw it away just to be on the safe side).

In principle we could do the same for SARS-CoV-2. Chaining lab tests would mean an enormous burden to the health system of dealing with false positives (and possibly re-testing people) at a time where we don't have unlimited health resources and where the probability that the false positive anyways catches the virus next day or so is non-negligible. This makes this strategy that starts from testing low risk population a complete waste of ressources (even if we had unlimited capacity for testing). This may change in the future when we have tests where proper full validation has shown that they have very good specificity and a useful sensitivity.

What we do currently, isn't very different though in principle: We do currently chain two independent tests, just that the lab test comes 2nd: first a doctor comes and tests whether you are high risk group because you had contact with confirmed cases (maybe even 2nd order contacts) or you show symptoms. Then comes the lab test.

Now the Germany* line in the table above is an interesting additional point. The prevalence there is from the influenza sentinel data from week 11 (last week). These are samples collected from people who were not sent to SARS-CoV-2 tests, but who went to a doctor with flu-like symptoms. 1 out of 251 was SARS-CoV-2 positive (and also influenza positive). The practically important point is: what did we miss because that SARS-CoV-2 case was overlooked (or would have been, had they not been in the sentinel sampling)? Actually, nothing: someone with the flu is told to stay at home and not infect anyone else pretty much like someone with SARS-CoV-2 (particularly now since it would also relieve the health system if the seasonal influenza would be overcome a bit faster).

Testing in areas with few infections

Frieden does not propose screening, though: he proposes testing people who have been in contact with infected ones. Which apparently did not work out in Seattle, from what I read due to a lack of tests - but it did work out in Germany where it allowed to completely stop the first (small) outbreak in January.

Here, the "had (2nd order) contact to infected person" is the first test in the chain. This is a containment strategy, not a delay strategy.

In order for this to work, we need a two "ingredients": we must be able to trace the contacts (and better also 2nd order contacts), and people sent to quarantine must obey this quarantine.

Comparison Germany - South Korea (mainly)

... the US is left as excercise to the informed reader

  • South Korea apparently makes extensive use of autamated tracing of people in a fashion that would be completely unconstitutional in Germany.

  • South Korea does use social distancing: according to Wikipedia they also shut down kindergardens, schools, and universities postponed their semester start. It did not have quite the same impact on overall economy that we see in Germany now. Their shutdown apparently was not all over the country, but earlier: they did that when they had 1000 cases.

  • Apparently the population did a whole lot of that volountarily: the wiki page says the streets of Daegu were empty on Feb 20 (when they had a total of 100 cases).
    In contrast, Germany now has 15 k cases, and pretty much an official shutdown since the weekend in terms of kindergardens/schools/universities, pubs/clubs/... and apparently there is a sizeable sub-population that gives in to the urgent need of having their own big (!) parties since pubs are closed.
    (I can only speak from what I read in the news, I live in rural Germany where my sports clubs canceled everything - but the neighbors had a 4 generation party (not that many people, but...) for their new house yesterday. (OTOH, we have 8.5 confirmed cases per 100k inhabitants here, so even with that factor 100 for dark number, we're still below 1 % prevalence now)

  • What seems to be similar: in South Korea young adults are the ones who spread the virus. Similar in Germany (more middle-aged initially [those who can afford to make skiing holidays in the Alps], now the Corona-Parties are a young adult thing)

  • I believe East Asia is in general more on the alert than we were. MERS and SARS are cited as reasons, but of course overall high population density also means a permanently increased risk of infectious disesases.

  • The Wiki page about COVID-19 in South Korea indicates similar testing capacity for South Korea (10k tests/day; 51 mio inhabitants) and Germany (16k tests/day; 80 mio inhabitants)

Some countries with more or less flat curves and more or less testing (per mio inhabitants):

  • South Korea so far done about 6000 tests / mio inhabitants
  • Italy 7300 / mio (maybe a bit late, but the main thing that was late in Italy is that for a while they didn't realize they had SARS-CoV-2)
  • Japan only 117/mio - but they have a very slow growth, still less than 1000 cases and their curve is decidedly flattening. Japan distances socially since Feb 25th (incl. e.g. to stay at home and not go to a doctor with only mild cold-like symptoms) at 167 cases in the whole country.
  • Singapore practices certain measures of social distancing since late January when they had < 10 cases. (Didn't find testing numbers, though)
  • Norway has 6440 tests/mio inhabitants. Hopeful look at curve suggests that a slight flattening may take place. They started their lockdown at ≈700 cases, a bit more than a week ago.

  • Germany has ≈2100 tests/mio. 100 k tests were done in week 11, current capacity is about 30 k tests/day (is currently ramping up).
    (substantial flattening at the earliest expected in about 2 weeks)

  • USA: according to Wikipedia 300 / mio inhabitants, but the German newspaper cites 3.5x as many tests, that would be 1100 / mio inhabitants.
    (The difference may be whether private lab tests are counted or not?)

The Dutch test

... is about antibodies whereas the other tests look for virus RNA. These tests again good for different things:

  • The Dutch test tells whether there has been sufficient exposure to the virus (and sufficient time thereafter) to build antibodies. (Once we know more about this, it may also tell whether one is immune against SARS-CoV-2)
    As they say, this is also important for getting an idea about the dark numbers of people who have been infected, but had so mild symptoms they didn't realize it.
  • The other tests tell whether you currently have the virus (and are contagious). Screening for active cases needs this test - the more so as the antibody test may become positive only a while after one is already contagious.


The Washington Post has simulations of the spread of a (easier, more basic) disease called simulitis with 4 scenarios, ranging from no measures at all over a not quite perfect quarantine to moderate and extensive social distancing (25 % and 10 % of the population run around as always, rest stays where they are). I highly recommend reading, but here are some runs I just did:


Brown are active cases, violet cured/immune and blue susceptible persons.

Important findig for this question: in the first part of the quarantine simulation, the quarantine worked an noone got out (1st hump). Note that this hump has roughly the same height as the moderate social distancing hump, and leaky quarantine is worse (though the simulation doesn't get any newly infected into a new quarantine - I'd say the quarantine situation may be less like real quarantine and more like totally closed borders vs. not-quite-free movement between two countries).


This is basically what South Korea is doing to limit the spread of the disease, and as the graph of the new-infection rate shows, it's working.

The problem is that it requires far more testing capability than the United States currently has; by April 1, it's likely the disease will be widespread enough that a quarter-million tests a week will be insufficient for trace-and-test containment (each potentially-infected person requires multiple tests over an extended period). Instead, the US is going for Chinese-style lockdown, but doing so far less effectively.

  • 2
    The US doesn't have enough public health personnel to do trace-and-test containment for that many, regardless.
    – Charles
    Mar 19 '20 at 20:16
  • 1
    South Korea has also what we now call social distancing, and they started that very early on (according to wikipedia). They apparently also had a "smallish" lockdown (including schools, kindergardens, universities, but regional). Sorry to -1, but one cannot conclude from tests + flat curve => works without considering what else they do that may cause the "works". Mar 20 '20 at 1:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.