Modellers have been using flu-tracking data to tease out incidence of Covid-19 in New Zealand. Photo / NZME
Modellers have been using flu-tracking data to tease out incidence of Covid-19 in New Zealand. Photo / NZME
Piggy-backing a major flu survey has been giving modellers a handy insight into our Covid-19 rates β and all without relying on test results.
In a new study, Kiwi researchers have outlined how they used weekly data from the Flutracking survey to get a fresh perspective on Covid-19 in New Zealand, where another 1103 new cases and 11 virus-linked deaths have been reported today.
For Dr Emily Harvey and her University of Auckland colleagues, the internationally-run survey, in which tens of thousands of Kiwis now self-report any influenza-like symptoms, offered a rich vein of data to mine.
Over the nearly two years where our elimination measures wiped out flu here, it provided a useful way to measure possible coronavirus cases against the number of PCR tests being carried out.
This graph shows the impact that lockdowns and border closures in March and April 2020 had on all influenza-like illnesses in New Zealand, until Omicron arrived in February 2022, followed by the return of flu in May. Source / Emily Harvey
This told them not just how high testing rates truly were, but also how quickly a new community outbreak might be picked up.
"Although we looked into other sources of data, including from GPs and Healthline, we found that these other data sources all had major limitations," Harvey said.
The biggest one: that people typically only reached for those services when their cold or flu symptoms became severe enough.
"This is an issue for monitoring Covid-19 testing, because we know that only around 20 to 25 per cent of Covid-19 cases will have the 'influenza-like illness' symptoms of cough and fever, and this varies with different variants."
Finally got the publication proofs sorted out for our NZ @Flutrack survey data analysis article, so thought it would be time to do a 𧡠open-access link to the publication here: https://t.co/k3Jr7IUqT5
After FluTracking was widened to ask about a range of Covid-like symptoms, including "runny nose" and "sore throat", the team saw a new way to capture infected people who weren't necessarily seeking healthcare.
Their first step was to define two new groupings from the data β those reporting any one or more, or two or more, Covid-like symptoms β alongside those people with cough and fever.
That allowed them to capture a greater proportion of Covid-19 cases, as well as more non-Covid respiratory illnesses that were also circulating but couldn't be picked apart without a Covid test.
Next, they made careful adjustments to correct for two known biases β that participants were more likely to report symptoms, and that certain age groups were under or over-represented β which could distort incidence estimates.
"Once we'd done this, we also used some statistical packages to calculate confidence bands for our weekly estimates," Harvey said.
"This was important to capture how much of the week-to-week variability we were seeing could be coming just from low survey responses, particularly in small regions or under-represented age groups, and how much we could have confidence that the increase or decrease seen was real."
This graph shows how changes in alert levels - marked with arrows - were effective at lowering all respiratory illnesses in New Zealand over 2020 and 2021. Red denotes people with traditional influenza-like illness (ILI) symptoms like fever and cough. Blue shows rates among those with one or more Covid-like symptoms, and yellow, two or more. These two groupings helped modellers tease out likely Covid-19 incidence here alongside flu. Source / Emily Harvey
"Although Flutracker cannot distinguish what is causing the symptoms, the expansion of the symptoms asked about has meant that we've been able to track the rates of respiratory illness in different age groups and in different regions," she said.
"We have seen clear signals of reductions in respiratory illnesses in school-aged groups during school holidays, with resulting drops in adults in the week or two following.
"Similarly, when schools go back, we consistently see an increase in rates of illness in school-aged children, with a rise seen in adults in the week or two after."
Over 2020 and 2021, Harvey found it handy in counter-acting headlines claiming that testing rates were plummeting, when it could clearly be seen the alert level restrictions were also lowering the spread of all respiratory illnesses.
"The number of people actually getting sick was dropping as fast, if not faster, than the testing numbers β so, our estimates of the testing rates were still very high."
This year, they've turned to it for estimating how many rapid antigen tests should be distributed, to ensure that everyone with new symptoms consistent with Covid-19 was getting tested.
"It's also been used in conjunction with other surveillance data, including wastewater testing, to help give confidence that observed case number drops are real, and not just a drop off in reporting."
That was important, Harvey added, pointing out that the FluTracker data could be picking up other illnesses with similar symptoms to Covid-19 β and indeed, it this year captured what was one of New Zealand's biggest flu waves in years.
This graph shows overall incidence of traditional influenza-like illnesses (ILI) like fever and cough in New Zealand. The first peak shown lines up with our first Omicron wave, driven by the BA.1 and BA.2 subvariants, while the second aligns with a surge in influenza, with the BA.4 and BA.5 wave not visible. Source / Emily Harvey
"To get insights into what is causing those symptoms, this data needs to be combined with surveillance testing that ESR and others do on swabs from hospital and other settings, and the incredibly valuable work done as part of the WellKiwis study, where participants with any respiratory illness get swabbed to determine which virus is causing the illness."
As experts have often pointed out, the numbers officially reported each day weren't reflecting the true picture of Covid-19 in New Zealand β something not just down to people not testing, but asymptomatic cases and RATs not picking up infections.
"Estimating the true proportion of people who have had Covid is crucially important for knowing the height of the 'immunity wall' and thus the size and timing of future waves, but also for knowing where the burden of disease if falling - who is being hit hardest by the infection," Harvey said.
"This would give a robust estimate of the proportion of people each week who have Covid-19, as well as valuable information on things like what symptoms people had, and their experiences of the illness," Harvey said.
"With Covid-19 here to stay, the sooner this gets up and running, the better."
