Computer models that tell us how internet memes go viral may also better predict the spread of Covid-19 - prompting a call for more of this intriguing work in New Zealand.

In a new study, scientists showed how models currently used to predict social trends can also forecast the spread of contagious diseases.

"It's a funny tool," said the paper's lead author Samuel Scarpino, an assistant professor at Northeastern University in the US.

"We know the model is wrong, but we can still use it early in an outbreak to learn something important about the progression of an epidemic."


The key to stopping the spread of misinformation about Covid-19 was better math.

In other words, even though the model was created to show how, for example, memes spread, it still worked when applied to contagious diseases.

As it stood, Scarpino explained, epidemiologists usually looked only at diseases in isolation without examining how they might interact with other pathogens or change depending on people's behaviour.

Hypothetically, the size of an outbreak should be proportional to the rate of transmission.

That was called a simple model.

For biological contagions, it might pan out that someone who is in contact with a person who had an illness might have a 50-50 chance of getting it.

That meant if they encountered two infected people, the risk doubled.

But Scarpino proposed that epidemiologists should instead adopt a complex model used primarily to track social trends.


According to this model, multiple exposures are required for a person to become affected by a contagion, whether that's a hashtag or a disease.

"One way to think about it is, for example, I'm online, and if I see an idea five or six times, I become much more likely to engage with that meme than if I only see it from one or two people," he said.

To apply this analogy to biological contagions, "it's not increasingly likely you'll become infected as you come into contact with one, two, three, four people" like the simple models suggested.

"It's that you almost have no chance of getting infected until you come into contact simultaneously with 10 people and then you get infected with close to 100 per cent probability," he said, with the caveat that this particular example was only hypothetical.

"We know that's not how biological contagions work in real life. They're much more complicated than social contagions. But mathematically, they're the same."

Big data in NZ

Here in New Zealand, researchers at University of Auckland-based Te Pūnaha Matatini have explored how complex systems can tell us how flu spreads, how our legal system compares with other countries, how ideas move from city to city, and how we can use ecological data better to conserve nature.

Auckland lecturer Dr Dion O'Neale said virus-focused models here can work by combining a series of "pots" mixing different attributes of a community - such as how many people worked more than one job, and thus could spread it further - to give a detailed picture of its overall vulnerability.

"If you've got information spreading about how to prevent the disease, like hand-washing ... you can imagine that will have a big impact on how the disease spreads later on." Photo / File

But not enough of this research was being carried out here.

"In New Zealand, I feel like we are a long way behind what's being done overseas," he said.

"We are trying to quickly scramble and turn the work of the PhD student who has been looking at employment networks into something at could be used to inform models for the spread of Covid-19."

Before the crisis, O'Neale himself was interested in looking at the spread of misinformation about vaccines, and how people make the choice about whether to vaccinate or not.

"And those choices aren't made in a vacuum. You make them based on things like what you see in the news, what your friends say and own experiences of flu-like diseases."

This information about vaccination attitudes, for example, spreads between people on one network at the same time as influenza spreads between the same people on another network.

"So in the case of Covid-19, if you've got information spreading about how to prevent the disease, like hand-washing or what hand sanitisers to use, you can imagine that will have a big impact on how the disease spreads later on. All of these things influence one another."

O'Neale also pointed out the influence of other diseases that were spreading at the same time.

"People talk about complex or interacting models of contagion. So your likelihood of being infected by Covid-19 can actually depend on all of the other diseases that are circulating.

"It might be that if you've had a cold or seasonal influenza in advance, you can dramatically increase your susceptibility to something like Covid-19.

"Or, if people have had a cold and stayed at home that can decrease their exposure at the same time that it's potentially increasing their susceptibility, were they to be exposed.

"Some of the work these researchers in the US did showed you can't distinguish between a complex interaction model or interacting diseases. You need to model everything."

He said the difficulty of doing this work in New Zealand wasn't a case not having enough data - Stats NZ's Integrated Data Infrastructure had enough rich information to build sophisticated models of likely contagion networks - but not having enough resources.

"It's just about having more funding, time and people to do it."