COVID-19 data is flawed, but it can be fixed
We’re seeing only the tip of the iceberg on COVID-19, Jack Siemiatycki says.
It is disheartening that, three months into the COVID -19 pandemic, we don’t know such key information as the prevalence of the infection in the population, the trend lines of this prevalence and the proportion of infected people who are asymptomatic. But the actions taken to control the epidemic depend on knowing these factors.
The public-health pilots who are determining policies are flying with faulty instruments.
The irony is that it would not take much to correct this. The key is testing in the general population. And contrary to the fears that this would remove needed testing resources from the clinical testing capacity, it would require perhaps one per cent of the testing capacity of the province to obtain crucial data for measuring this epidemic.
There have been serious shortages of the materials needed for testing; consequently it is mostly only patients with the most serious symptoms who are being tested. Premier Doug Ford has announced his intention to dramatically increase the testing capacity in Ontario. That is good news for people who become infected. But increasing the numbers of tests does little to improve the validity of the epidemiologic data on the evolution of the epidemic.
The daily reports of the numbers of people who have been diagnosed with COVID -19 or hospitalized or died are being used by the authorities and the media to estimate the prevalence of the infection, to track the progress of the epidemic, to assess the case fatality rate, to model and predict future numbers, to plan for the need for hospital resources, and other vital tasks. A lot depends on the validity of the data.
Unfortunately, the data derived from severely symptomatic self-selected patients do not provide good indicators of the prevalence or trends in COVID-19 in the population. We see the tip of the iceberg; we need to know how many infected people are below water line. We do not know if confirmed cases represent 10 per cent, 50 per cent, or 90 per cent of the truly infected cases in the population. We cannot use the numbers of “confirmed cases” to estimate prevalence or to track the progress of the epidemic. Similarly, we cannot compare the numbers across provinces or between Canada and other countries. Nor do we know what the case fatality rate is for COVID-19 because we do not know the denominator for such calculations — namely the number of people with the infection.
The epidemic is being propagated by people who are carrying the virus. We now know that infected cases can spread the disease before they get symptoms, and probably even if they never get symptoms. We need to know how many people in the population are capable of spreading the infection.
Among other things, modelling the future of the epidemic should take into account the fraction of the population that is capable of transmitting the infection. Current modelling efforts are hamstrung by inadequate data.
What would be required to have good data? Testing the entire population would do it. But that is hopelessly ambitious because there will never be enough testing capacity to cover everybody in a large country. But there is a feasible strategy.
Just as we carry out sample surveys on all manner of topics, from pre-election polling to marketing surveys, we can use the magic of statistical sampling to obtain valid information on this topic. We would only need between 100 and 400 people tested each day in Ontario. This, carried out from as soon as possible to the end of the epidemic, would provide adequate statistical power to obtain adequately precise and accurate estimates of the prevalence and the trend in prevalence over time, as well as the proportion of asymptomatic cases.
There are many technical and logistical decisions to make about how such a scheme should be implemented, but the basic principle is that a very small representative sample can provide valid estimates, and this would not require many test kits.
The testing most commonly carried out now in Canada is a genetic test for the presence of COVID-19 virus in the body. Alternative tests measure antibodies to the virus in blood. The two types of tests are complementary, not redundant. As soon as the antibody test becomes available, it should be added to the proposed testing protocol.
Hopefully, this opportunity to track the epidemic will not be lost; time is of the essence. Jack Siemiatycki, PhD, is professor of epidemiology, University of Montreal, Guzzo-Cancer Research Society chair in environment and cancer, and adjunct professor, McGill University.