A Kiwi company able to predict future outbreaks of Covid-19 and its speed of spread is playing a key role in the global fight against the pandemic.
The free online forecasting model, built by Wellington-based CloseAssociate and being used by governments and organisations here and around the world, has been described as the "gold standard" for global monitoring of the virus.
The technology - which correctly predicted last year's virus spike in Victoria, Australia and has helped the University of Illinois in the US to safely bring back 46,000 students to its campus - uses artificial intelligence technology hosted in Microsoft Azure cloud to predict future infection rates up to 14 days ahead.
CloseAssociate co-founder Leon Grice says the tool is available for any international organisation or government to trace the likely spread of the pandemic in their own countries, giving them time to limit its impact.
"From Norway to Nigeria to Nevada, we are running the algorithm on a daily basis for almost every nation around the world to provide an at-a-glance snapshot of how fast the virus is spreading."
CloseAssociate has been awarded a Microsoft AI for Good grant for its work in developing the tool. Microsoft New Zealand's national technology officer Russell Craig says the aim of the award is to empower clever thinkers everywhere to help solve the most pressing issues.
"CloseAssociate's model is the gold standard for global monitoring of the virus; we hope they'll continue helping countries plan ahead, manage the virus and save lives. Leon and Stephen Grice are doing amazing things that governments and organisations across the globe are benefitting from."
Leon Grice says for those without degrees in epidemiology, the spread of the virus can seem like a roll of the dice: "Why are some areas worse affected than others? Why, when case numbers are falling can there be sudden, terrifying flare-ups?
"The ability to stop these flare-ups and to allow citizens, governments and businesses to plan ahead is crucial to rebuilding our economies and keeping us all safe," he says.
Work on the model became something of a Grice family affair, arising after Leon was invited last year to join a team of business experts helping the government solve urgent supply chain issues during lockdown. It was an exercise also involving New Zealand Defence Force logisticians and a team at the University of Auckland led by Professor Shaun Hendy, an expert in data analytics.
Grice immediately brought in brother Stephen – one of New Zealand's first licensed Microsoft engineers, a man with serious digital know-how, a passion for data and the power of applied mathematics.
Using Microsoft's Azure cloud, web app Blazor and AI technology, Stephen developed a model giving logisticians and planners access to undertake their own "what-if" scenarios: "We thought we could take the methods he (Professor Hendy) developed and make them more accessible through a website that could be updated with new data," Leon Grice says.
Then, he says, came a "eureka" moment. His son Patrick, a recent engineering graduate from the University of Auckland, was helping the team when he saw it was possible to use a numerical method to calculate a daily reproduction number – showing the rate people are being infected by the virus.
"Also known as the RD number, it gives you a reliable short-term prediction for the rate the disease is spreading," Grice says.
After sharing this breakthrough with an old school friend, Professor Richard Laugesen at the University of Illinois, the three men formed a team to refine Patrick's method. The result was a world-first numerical method – the Grices call it SIR+B – for calculating the rate of the virus's reproduction.
"If you know the daily reproduction value with a high degree of accuracy, you can reliably predict the direction of daily cases the next 10-14 days," Grice says.
One of the model's first tests was when it revealed, in mid-2020, that the reproductive number in Victoria "was going through the roof".
"We contacted senior members of the CSIRO in Melbourne (Commonwealth Scientific and Industrial Research Organisation) to provide our analysis and to report they had a crisis brewing," he says. "You could map the outbreaks at their managed isolation facilities and the reproduction rates were skyrocketing.
"We predicted there would be a spike in July, which unfortunately was the case; Victorian authorities progressively moved to lock down the state."
The CloseAssociate model also helped the University of Illinois re-open its Urbana campus for 46,000 students last August at the height of the spike in US cases. The model indicated students would bring widespread infection onto the campus without other intervention. University leaders realised they needed large scale, rapid, repeatable, non-invasive surveillance testing if they wanted to stay open.
Grice says this provided the evidence the university needed to roll out a Covid-19 saliva diagnostic test to all students - alongside mask wearing, online classes and rapid isolation in the event of any positive cases.
To date, the university has conducted almost one million tests on campus and has managed to keep operating despite surges of infection across the US.
To read more about CloseAssociate and AI for Good, click here