Wednesday 10 June 2020 2:28pm
Dr Matthew Parry led a team that reviewed the modelling work driving the Government's COVID-19 decisions.
An epidemic in Wuhan will not behave in the same way as an epidemic in Whanganui.
Dr Matthew Parry from the Department of Mathematics and Statistics is an Associate Investigator for Te Pūnaha Matatini (TPM), a Centre of Research Excellence that has been instrumental in providing modelling to support COVID-19 decisions. During lockdown Dr Parry chaired a panel to provide a quality assurance process for these TPM modelling papers. Just like the rest of the country, he tuned into the Government’s daily 1pm briefings; however, Dr Parry was very aware that behind those figures was a maze of noise, uncertainty and assumptions.
Sciences Communications Adviser Guy Frederick caught up with Dr Parry for an insight into the modelling that informed the Government’s world-leading response to tackling the virus.
What did your role as chair of the panel for TPM involve?
The panel was made up of seven specialists that included experts in modelling, epidemiology and public health. They were drawn from inside and outside TPM and also included Anja Mizdrak and Melissa McLeod at the University of Otago, Wellington. We reviewed papers prepared and written by TPM modellers, who would then incorporate our feedback prior to the papers being officially released. Modelling for COVID-19 was new territory, so at every stage there were improvements and we were pretty tough with our reviews. My job was to compile all the reviews and feedback into a single document for the papers’ authors to respond to.
Can you give an insight into how models work?
Each round of modelling is tailored to respond to the question being asked. At the start of the crisis we were interested in what could happen if no action was taken. The TPM models very clearly indicated the epidemic would take off, and there was enough detail to give a general idea of ‘potential futures’ under different courses of action. It also helped that we could see how the pandemic was playing out in other countries, which provided feedback about the decisions that were made in order to avoid similar things happening here. Modelling will never be 100% accurate, but it doesn’t attempt to be, and it doesn’t have to be in order to be useful.
Have you had any direct input into the modelling used during COVID-19?
Mathematical and statistical modelling are quite different and it was good to be able to link these things. Mathematical modelling often looks at parameters as quantities that can be tweaked; statistical modelling hones in on how to estimate parameters from data and with what certainty. I enjoy both aspects. I worked with the TPM modellers on the effective reproduction number pre- and post-lockdown and was able to add statistical insight into estimating the parameter and its possible ranges. I was also involved in modelling with Michael Baker, Nick Wilson and Ayesha Verrall, where we worked on probabilities of a so-called epidemic extinction in New Zealand. Our model estimated that at 27–33 days of no new cases, we would be 95 per cent certain of having reached this state. Now we are in Alert Level 1, the business of elimination is to be on the lookout for imported cases and cases connected to asymptomatic transmission. This means we need to maintain our testing and tracing processes, and maybe keep the modelling going!
Do the nuances of New Zealand’s situation alter how modelling is undertaken?
You can’t expect an epidemic in Wuhan to be the same as one in Whanganui. Whether people live in high rises or detached houses, for example, will change the characteristics of an epidemic. So it’s for this reason that we can’t necessarily apply parameter values from different countries. Modelling is both an art and a science, and we are constantly making decisions about what level of detail is required based on what questions are being asked. As an example, right from early on, the TPM modellers wanted to incorporate the fact that inequity goes hand-in-hand with poorer health outcomes. I learned a lot from the panel members on this and I think the panel was able to make a real difference and focus the modelling work that was subsequently done.
For a mathematician, has the pandemic presented a Pandora’s box?
I don’t have all the answers but this was an excellent opportunity to use my skills in the mix alongside public health experts, epidemiologists and other modellers. Before this crisis we didn’t have any work completed on pandemic modelling response but now we do. Parameter estimation for modelling takes a lot of work, so it’s really great to now have some robust methods in place to be able to address pandemic-related questions, both now and in the future.
Now that we have reached Alert Level 1, what does this mean for the models?
Modellers are certainly very happy with the current situation in New Zealand, but we also take a precautionary approach – which may say something about the psychology of modellers! We know from modelling that something with a very small probability can still have huge consequences, and the probability of COVID-19 in the community now may be tiny but the costs of it still being out there could be huge. Elimination is a public health question not a mathematical one, as it’s based on having no transmission of the virus for a period of time as well as maintaining a high level of testing and surveillance.
What other mahi do you have going on?
COVID-19 work has certainly dominated my life during New Zealand’s response to the pandemic. I’m looking forward to giving some attention back to my research that ranges from modelling invasive pests, fake news, and epigenetic factors in disease. I am also the Associate Dean International for the Science Division and there is a tremendous amount of work going on at present both to support international students and to rethink how the University engages internationally while our borders remain closed. COVID-19 modelling and my role with international students are both directly linked to the pandemic, so they both highlight just how much everything is connected.