COVID-19 modeling is hard and models are only as good as the data available to you. You need good models but you also need high quality data in large quantities. Improving model predictions is a matter of better data input and that comes mostly from surveillance testing. In New Zealand we have very poor quality data in the public domain.
Interpreting data is fraught with difficulty because pandemics scale exponentially. Variations within and between populations makes it difficult to compare different outbreaks. These natural differences, changes in public health policies, changes in public sentiment, and mutations in the SARS-CoV-2 virus add further complications.
All of this is not to say that models are irrelevant or have no purpose. In fact, the purpose of models is to provide the mathematical basis on which such comparisons can be drawn. The effective reproduction number, Reff, is an epidemiological variable that is useful for this analysis.
Good models have strong theoretical foundations based on fundamental biology, make a minimum number of assumptions, are repeatable, and are widely applicable across time and space.
Rako Science has built such a model. This model has now been running for nearly two years in hundreds of countries and thousands of counties, states, and provinces. In that time, we have learnt a lot.
The model shows that when the effective reproduction number Reff value is above 1 for an extended period it is preceded by a surge in positive daily cases and what follows is stress in the country’s health system – every time no exceptions. Reff is a leading indicator.
In New Zealand our model shows that Reff went positive on or about 8 January 2022. This occurred at a time when we were on summer holidays.
In other countries this would typically have been picked up with surveillance testing as an early warning of what was to come. In this country, we do not have publicly-funded surveillance testing. Testing is rationed to symptomatic individuals and close contacts of known positives. Testing rates were very low throughout the Christmas period. That meant we had no warning of what was to come.
What was to come was that Reff rose rapidly peaking at a value of 3.3 on or about the 15 February 2022. Some commentators have expressed doubts about the absolute value of the peak in our model. What the absolute value was is secondary to the fact that it was much greater than 1 and this was sustained for 6 weeks leading up to the surge in positive cases.
The massive increase in daily cases starting on the 15 February 2022 coincided with the return to school for primary school and secondary school students. This pattern was expected, and many similar surges had been observed in northern hemisphere surges in July-August 2021.
The surge in positive cases led to the predictable collapse of testing in our public health network after labs with pooling dependent capacity were overwhelmed with the high positivity rate, as it had in Australia only two months before.
After that people took voluntary social distancing measures and increased mask wearing, stopping the outbreak and returning R0 to under 1 on or about 11 March 2022.
So that is hindsight. But models are nothing if they can only explain the past.
What can models tell us about the future?
Let’s look at Australia.
Australia has had an Omicron outbreak very similar to us except it started 3 months earlier on or about 8 November 2021 when R0 went greater than 1. They peaked at about R0 = 3.2 on the 29 December 2021. Australia had a rapid drop in the value of R0 after their public health testing system collapsed. Like New Zealand they switched to self-reporting using rapid antigen testing to cope with the high volumes and high positivity rate.
There is a remarkable similarity between the peak and length of the omicron surge in Australia and New Zealand.
Because Australia appears to be 8 weeks ahead of New Zealand, we can look at what is happening there now to help anticipate what could happen here. In Australia R0 hit a minimum and has started rising again. That again is a leading indicator of a rise in positive cases which started in March 2022.
There is great uncertainty in all models. One of the ways that mathematicians model uncertainty is to use confidence intervals.
What is striking about the confidence intervals in our model for New Zealand is how large they are. The value of R0 at the time of writing was 0.8 (CI 95% 0.0-3.4). That is one of the largest confidence intervals we have observed in two years of running the model.
Why is that?
The fact that we do not have any public health surveillance testing is partially responsible. The collapse in public health testing due to high positivity rates is also a contributing factor. The unplanned switch from public health PCR testing to self-reported Rapid Antigen Testing revealed the extent of the community outbreak.
All modelers are reluctant to make predictions based solely on mathematics for the reasons we have outlined. But sound mathematics combined with empirical knowledge can inform our public health experts and policy makers about what comes next and to avoid the mistakes of the past.
If we don’t have adequate, sensitive surveillance testing in place it is certain that our public health leaders and policy officials will be caught out again. And you don’t need to be a modeler to predict that!
 | Stephen Grice is a founder and director of Rako Science. Stephen has a PhD in Physical Chemistry and 30 years of experience working in the high tech, business and science sectors. Stephen's work with University of Illinois Urbana Champaign on new mathematical methods for the epidemiology led to licensing and development of covidSHIELD in New Zealand and the implementation of the saliva test protocol by accredited New Zealand laboratory IGENZ. |
|---|
Rako Science is a pathology company offering saliva based rt-qPCR tests. Rako Science uses the University of Illinois covidSHIELD saliva test that was diagnostically validated and accredited by IANZ and contracted to IGENZ Ltd. Rako Science’s test is the first saliva test accredited in New Zealand and remains the only diagnostically validated test in New Zealand.
Rako Science is also the first test that supports IATA Travel Pass.