As Omicron spreads rapidly across Australia, states’ testing regimes have failed to keep up with surging demand and a lack of rapid antigen tests, meaning they are no longer capturing the true number of infections.
As a result, data experts are concerned at the reliability of the figures being used to inform our decisions, with one saying that announcing daily case numbers at this point “feels completely arbitrary”.
They also warn incomplete information can hamper public health tactics as Australia enters the third year of the pandemic.
How high are the actual case numbers?
With many people saying over the last couple of weeks that they tested positive to Covid on a rapid antigen test or had Covid-like symptoms but could not get a PCR test, it is difficult to say how many Australians have had the infection.
There are estimates case loads could be up to 10 times higher than the daily figures announced.
State health authorities have acknowledged their testing systems have not been capable of detecting the extent of transmission.
The New South Wales chief health officer, Kerry Chant, reiterated on Friday that “the numbers of tests positive that we report everyday will be an underestimate”.
“Across Sydney, we have got clearly a very high case burden,” Chant said.
Earlier in the week, national cabinet announced an agreement to transition to accept rapid antigen test results as confirmation of Covid status, following weeks of so-called tourism testers and close contacts crippling the capacity of PCR labs.
On Friday, NSW announced it would follow Victoria in introducing an online portal to report positive RAT results to improve accuracy.
Why is it still important to capture all infections?
Aside from not capturing a true figure for new infections, the lowering of standards for collecting Australia’s Covid data will have flow-on effects on the country’s pandemic experience, says Juliette O’Brien, the data journalist who created covid19data.com.au.
“There’s a giant asterisk next to any daily case number now, unless it’s from a state where it’s relatively under control. But on Friday, typing 38,000 for NSW sounds completely arbitrary,” she says.
“There’s this huge caveat that we just don’t know what the real case numbers are, and an immediate issue is that we can’t calculate the true positive rate and reproduction rate. These are key metrics that show us when outbreaks are starting and when they slow down.”
She says this changes the reliability of the data Australia contributes to international disease reporting efforts, and limits how much we can understand about the nature of the Omicron variant and future strains.
Who are we missing?
O’Brien adds that while states are moving to record positive RAT results, the serious problems with cost and accessibility that have persisted over recent weeks and continue now means we have a less clear picture of the true growth rate.
“Even though we see more cases reported each day, for all we know the rate of growth could be slowing. We are blind to that now.”
Additionally, the emphasis on symptomatic people getting tested means we have stopped capturing information about asymptomatic cases, which can be important for our understanding of the severity of the strain.
This is a departure from earlier in the pandemic, when asymptomatic cases were detected due to a thorough system of contact tracing and isolation.
“When the country has abandoned any attempt to capture anything close to the real number of cases, that is the definition of out of control,” O’Brien says.
What does this mean for modelling?
Australian governments have relied on modelling to make public health decisions throughout the pandemic.
However, the accuracy of much of the modelling released by health authorities has been mixed.
During NSW’s Delta outbreak, hospitalisation predictions failed to materialise.
Meanwhile, a worst-case scenario prediction made in mid-December that the state would reach 25,000 new daily cases by the end of January proved overly optimistic.
Dr Michael Lydeamore, a Monash University infectious diseases modeller who is part of the Doherty Institute modelling consortium that advises Australia’s national cabinet, believes deteriorating quality of case number data will make the task “very difficult”.
This is because without capturing the full picture of cases, modellers need to “calibrate” their models to other, related information where the data is complete.
For example, death and hospital data continues to be “very well reported by the states”, says Lydeamore, and this can be compared with previous points in the pandemic when the accuracy of cases reported was better. That can help calibrate current case data to deliver a more likely number of true infections.
Other methods of calibration will help to better understand the true caseload, such as surveying a sample of the population at a point in time, testing them for antibodies – an indication they’ve been infected – and recording their vaccine status.
This enables the level of less severe infections to be better understood.
However, these surveys have not been conducted in Australia. Lydeamore believes this type of surveying will be necessary in the future for Covid modelling to continue to be accurate and useful for the public health response.
In the meantime, modelling will be most useful on near-term forecasting, with Lydeamore warning it will become “too difficult to build good models”.
As a result, health experts who provide health advice to governments will likely be more cautious in their recommendations, given the lack of clarity of the true transmission that is occurring.
Whether governments adopt those more cautious recommendations is another question, says Lydeamore.
“It’s clearly not a great situation.”
Will this limit how much we can know about vaccine effectiveness?
Lydeamore says that health experts primarily measure vaccine effectiveness via randomised control groups. Participants receive the vaccine and their antibodies are measured at different stages to gauge how they have responded.
While large-scale outbreak data is better understood when looked at with vaccine rates, Lydeamore says this is not the method used to determine vaccine effectiveness.
However, information about the spread in specific areas where not everyone is vaccinated will be less clear, and “become a bit of a guessing game”.
