Public health experts and politicians are suggesting that we can see a point when pandemic-related restrictions could relax. But any relaxation must be careful and measured, quickly reversible if COVID-19 reappears.

Hey there, time traveller!
This article was published 22/5/2020 (521 days ago), so information in it may no longer be current.

Winnipeg Free Press

Delivering Crucial Information.
Right Here.

Support this work for just $3.92/week


Public health experts and politicians are suggesting that we can see a point when pandemic-related restrictions could relax. But any relaxation must be careful and measured, quickly reversible if COVID-19 reappears.

Everyone maintains that widescale testing is a key to controlling the reopening of the economy. 

Most of the proposed approaches, such as one advocated by Harvard University, involve testing of those with symptoms and in critical positions, detailed contact tracing of those testing positive, and isolation of those testing positive. But is this enough? South Africa’s experience with mass testing offers an important lesson. 

In the 1950s, as the Cold War intensified, the Canadian and American military staffed about 60 radar stations in the Arctic to detect Soviet bombers. Known as the DEW Line (Distant Early Warning), the intent was to provide timely warning of an impending nuclear attack. With COVID-19 we need a "DEW" line to alert us before cases appear in hospital and force us to play catch-up again.

A COVID-19 DEW line has two components: testing widely in the general population, both for the presence of the virus and for the presence of antibodies; and a strategy for efficiently deploying a testing program across the population. On both fronts, we have some considerable distance to go.

Testing — or, more correctly, the lack of testing — is a signal failure in managing the response to this disease. On that point a consensus exists. However, to be of any value, diagnostic tests for the presence of disease or antibodies against a disease must be accurate

Accuracy in a diagnostic test, medical and otherwise, has two components. A false positive for COVID-19 is a test result that indicates someone has COVID-19 when, in fact, they do not. The result of this error is inconvenience; since if the person is not showing serious symptoms, they will needlessly self-quarantine.

Far more problematic is the false negative, which shows that someone does not have COVID-19, yet they do. If symptom free, that individual would be al liberty to go and spread disease.

The problem is that the false-negative rate on the COVID-19 diagnostic test is not known; the suggestion that it could be three in 10 is probably wrong. It is likely much lower than that, and getting better all the time, but unless the error rate is lower than five per cent, the test may not be useful that for mass screening.

If someone has the classic symptoms of COVID-19 but the test was negative, most physicians would override the test result and presume the patient was ill. Further, the turnaround for test results must be rapid

What about the antibody test? Here, the reported errors could also high, but with different consequences. A false negative means the test shows you do not have antibodies when in fact you do. The consequence is that your social life continues to tank. The false positive is clearly more dangerous to the person and society, because someone behaves as if they have immunity when they do not.

Both tests will get better, and probably quite quickly. The recent recall of the rapid test is a setback; however, another challenge is even the best tests are not ready for wide deployment. Manufacturers will require time to produce the hundreds of millions of test kits.

What process should we use for general population testing? We could replicate the random-sampling designs used by market researchers, pollsters and official statistics agencies, where a small sample supports inference about the population. Logistically, we could intercept people much as we test for drunk drivers. Police stop tens of thousands of motorists in a six-week winter holiday period.

How many people would we need to test, and how often? That is where another problem arises: if we assume the current infection rate is one per cent, then a random sample of 5,000 in a biweekly period might reveal 50 with the virus.

The true infection rate may be lower, simply because we are currently testing only those with symptoms and those in risky occupations. A random sample of 5,000 in any province or city could reveal as few as five with the virus, far too low to serve as a baseline for guiding public-health policy.

Increasing the sample size is one approach, but 20,000 or even 50,000 intercepts repeated every two weeks would increase logistical complexity and cost.

An alternative course uses a stratified random sample, which targets the intercepted population to optimize estimates of the infection rate. Organizations such as Statistics Canada routinely develop these sampling designs; however, our understanding of COVID-19 must increase before we can confidentially target the sample and estimate how many people we need to intercept in each biweekly period. But it is feasible.

Certainly, returning travellers at airports, high density urban areas and nursing homes are locations where we may need to test more intensively.

We have a three-month window to create a COVID-19 DEW line before school starts in the fall and people want to travel for the upcoming holiday season — all parts of "normal" life. Refining processes for mass screening, ensuring rapid-test accuracy, and accelerating the manufacturing the test kits, to create a COVID-19 DEW line is a very tall order in that short time. But meeting this challenge will define this generation.

Gregory Mason is an associate professor of economics at the University of Manitoba.