New research published today shows that the number of cases of COVID-19 are much greater than reported, and this has important implications about how we respond to the pandemic crisis, Quentin Grafton, Steven Phipps, and Tom Kompas write.
A peer reviewed manuscript published in the respected journal Royal Society Open Science today shows that in some countries, COVID-19 case numbers may be up to 17 times greater than those confirmed by testing.
According to John Hopkins University and the World Health Organization (WHO), as of 17 November, the global number of total cases – that is, all recovered, active, and fatal cases combined – was 55 million. Total deaths were 1.3 million, and the number of confirmed cases was increasing by about half a million per day.
Confirmed cases are the number of persons who have tested positive for RNA material of SARS-CoV-2 (the virus that causes COVID-19) present in their nasal secretions or sputum. As a complement to RNA testing, and to estimate the true infection rate, researchers have also undertaken seroprevalence studies that test for antibodies in blood samples. Identification of antibodies in the blood can indicate that a person either has or has recovered from COVID-19.
Seroprevalence studies need to be repeated regularly and be based on an appropriately stratified random sample of the population to obtain a reasonable estimate of the true (population) infection rate.
A challenge with seroprevalence studies is that, if the true rate of infection is relatively low (say one per cent or less), then the number of false positives or false negatives may make the sero-surveys unreliable as a means of estimating true infection rates.
This is true even if the sero-test has a high proportion of true positives and true negatives.
To combat this difficulty, we employed a statistical method called backcasting that allows us to obtain estimates of the true infection rate based on a range of infection fatality rates (0.37 to 1.15 per cent) and the time from infection to symptoms (4.1 to 7.0 days) and time from symptoms to death (12.8 to 19.2 days).
Using our method, we generated multiple random values based on the range of possible values of these three parameters (infection fatality rate, time to symptoms and time to death) to provide a 95 per cent confidence interval around our estimates of the true infection rate.
Our analysis covered 15 developed countries with a combined population of over 800 million people. Our backcasting method generated similar results to national seroprevalence studies. Importantly, we found that COVID-19 is far more prevalent than is suggested by reported statistics of confirmed cases identified by RNA tests.
We found that the true number of infections across our sample of 15 developed countries is 6.2 times greater (95 per cent confidence interval: 4.3–10.9) than the number of cases confirmed by RNA testing. We also found a strong negative relationship between the proportion of people testing positive for the virus from RNA testing and the detection rate of COVID-19 in the population.
Our method is novel and easy-to-use and especially valuable wherever there is reliable data on the number of fatalities attributable to COVID-19. Unlike reported infections based on RNA tests, backcasting is not dependent on the coverage or efficacy of testing regimes.
Backcasting is also scalable to a local, regional or national level, can be readily updated on a daily basis using data that has already been reported, and makes no assumptions with regard to how the number of COVID-19 infections has evolved over time.
Our approach is particularly advantageous in locations where there is little testing or limited capacity to forecast rates of infection but where there is a need, for the purposes of public health planning, for a more accurate population-level measure of COVID-19 infection.
While our method is robust, we highlight three limitations when comparing infection rates across countries and over time.
First, the age distribution across different populations needs to be broadly similar, because the infection fatality rate from COVID-19 is highly dependent on age.
Second, the level of medical care across countries should also be comparable, because COVID-19 fatalities depend on access to medical services, such as the use of ventilators.
Third, the infection fatality rate should be broadly constant over time, as any substantial change may introduce biases into the estimated population infection rates.
Most countries in the world have undertaken fewer tests per 1,000 people than the 15 countries considered in this research, and also have a lower capacity to test for COVID-19. This suggests that globally the number of people who are infected with, or who have recovered from COVID-19, is many times greater than the reported number of cases from viral testing.
Even within our sample, the countries with the lowest detection rates (Belgium, France, Italy, and the United Kingdom) appear to have a population infection rate that could be up to 17 times greater (Italy) than reported by confirmed cases via John Hopkins University or the WHO.
While there are many ways to respond to COVID-19, our statistical measures of the true infection rate should promote better public health decision-making. This is important because if governments do not know how many people have been infected in a population, it becomes very hard to plan a pandemic response effectively.