The true number of COVID-19 infections is probably much higher than what's being reported in many high-income nations around the world.
A newly modelled estimate from the United States, Australia, Canada, South Korea and 11 countries in Europe suggests official figures could be struggling to capture the full scale of the outbreak.
The new model from scientists in Australia employs a 'backcasting' method, which projects the number of new daily fatalities in reverse, from the time of death to the time of infection. This allows scientists to avoid using epidemiological and serological data, which comes with testing limitations.
Comparing the new estimates with official confirmed cases, the team was able to predict the 'true' infection rate for each country. According to their results, at the end of August the population infection rate was, on average, six times higher than reported cases.
"Unlike reported infections based on RNA tests, backcasting is not dependent on the coverage or efficacy of testing regimes, which can be very different across jurisdictions and over time," the authors write.
That means it's much easier to use on a regional, national or even international level than other methods. What's more, because it doesn't rely on a nation having widespread testing, it can help public health experts prepare in areas that have limited testing capacity.
"Simply put, we analysed statistics on how many people had died from COVID-19 in a given country and then worked backwards to see how many people would have to have been infected to arrive at that number of deaths," says data scientist Steven Phipps from Ikigai Research in Australia.
"Our method is a novel and easy-to-use method for estimating the true infection rate wherever there is reliable data on the number of fatalities attributable to COVID-19."
Some nations were better at reporting these infections than others. In South Korea, the actual number of infections was found to be 2.6 times higher than reported figures, whereas in Italy, the 'backcasted' number of cases was a startling 17.5 times higher.
In general, since March, countries around the world have gotten better at rolling out COVID-19 testing, educating the public on symptoms, and coming up with more and more accurate ways to detect and track the infection.
Despite that improvement, international numbers continue to lag behind the likely reality. Even in Australia, which has one of the best levels of detection among all 15 countries studied, researchers say the rate of infection could still be nearly five times higher than what's being reported.
"We found COVID-19 infections are much higher than confirmed cases across many countries, and this has important implications for both control and the probability of infection," says economist Quentin Grafton from the Australian National University.
"These findings raise serious questions about how we deal with all facets of the coronavirus pandemic, including ongoing morbidity and life-long health impacts for people who have been infected, how we implement and manage lockdowns, and how we make sure we are on top of this pandemic more broadly."
This isn't the first time scientists have found a discrepancy between actual COVID-19 cases and reported infections. Pretty much since the beginning, experts have warned we are likely underestimating the true extent of viral spread.
Determining a cause of death from the novel coronavirus is no easy matter when testing is limited, symptoms often overlap with other illnesses, and those who are most vulnerable have pre-existing medical conditions.
Many estimates to date have compared the total death rate in 2020 to what it would usually be in any other given year, or they've used antibody testing to go back and identify individuals who were not included in initial case figures, possibly because they showed little to no symptoms.
Most epidemiological models agree that actual infections far outnumber confirmed cases, but exactly to what extent and how that changes over time is less clear.
Epidemiological data are limited by the level of a nation's testing, and antibody testing comes with some false positives and false negatives, which means that if the number of cases is low, on a population level even a handful of false results can skew the data.
An estimate by a different study in the US found the number of infections in April was 3 to 20 times higher than the number of confirmed cases, and most of that was due to incomplete testing and, to a lesser extent, imperfect test accuracy.
The new model is only based on high-income countries that have relatively widespread testing regimes. Most nations, however, have taken far fewer tests among their populations, which suggests the number of people globally who have been infected with COVID-19 is likely several times greater than official figures.
Some countries like Belgium, France, Italy and the United Kingdom were found to have very low true detection rates. As of 31 August 2020, official figures in these nations represented only 10 percent of all actual COVID-19 cases, according to the new analysis.
At this point, however, no estimate is perfect, and this new method shouldn't replace existing ones, it should merely complement them.
Epidemiological models are still much better at predicting future hospitalisations than backcasting methods, and the authors wholly admit this.
It's also important to note that for backcasting to be accurate, the age distribution across those infected with COVID-19 has to be broadly similar, because older people have a higher chance of death once infected. This may skew the results in places like Australia where around 75 percent of the deaths have occurred in aged care.
Finding the best way to estimate past, current and future COVID-19 cases will take time, and to a certain point, it might be impossible to ever truly know exactly how many people will be sickened by the current pandemic.
That said, accurate estimates of the real COVID-19 burden will be crucial in determining how to respond to the global tragedy on our hands.
The study was published in the Royal Society Open Science.