In March of 2009, the Mexican government confirmed it: A four-year-old boy in eastern Mexico’s La Gloria village had swine flu, or H1N1. Sixty percent of the village had reported an unknown respiratory illness back in February, and since, it looked like the virus had jumped. A flu case later confirmed to be H1N1 popped up in California. Then, it spanned an ocean. By late April, H1N1 had been reported in Spain, Israel, New Zealand, Austria, Germany, the United Kingdom, and Switzerland.
Within an increasingly globalized and mobile world, the spread of contagion doesn’t work how it used to. But by taking these factors into account, theoretical physicist Dirk Brockmann and his colleagues have a radical new model that could predict the arrival times of the next global pandemic. The model relies on something called “effective distance,” and it destroys a centuries-old way of thinking about maps.
It’s a simple concept. Flight patterns and common aviation hubs are more accurate predictors of the spread of disease than distances traveled on foot, slow boat, or horse. Brockmann had a lightbulb go off when a student, Daniel Grady, biked to his office at Northwestern University and remarked that no matter how he traveled–by subway, bus, or bike–it always took the same amount of time to get there.
“And that is when it started,” Brockmann said. “In the modern world that’s so connected, old school, conventional geographic distance is not so meaningful anymore.”
Over the last decade, Brockmann and his colleagues demonstrated how “effective distance” might work by building theoretical pandemics and models that predicted their theoretical spread. They analyzed three-years-worth of airline data in order to see how these distances interact. But could they come up with a universal equation that would account for the arrival times of any disease across the globe?
A paper published in Science today with co-author Dirk Helbing presents a culmination of that work, and Brockmann’s proudest breakthrough. It shows that the jumbled, erratic spread of disease can be reduced to simple, constant wave patterns, riding on effective distances, rather than geographic ones, through a “global mobility network.” Brockmann was able to prove the concept using data from the 2003 SARS outbreak and the 2009 H1N1 pandemic. When he plotted the epidemics’ actual arrival times in different cities against their effective distances, he found strong positive correlations–stronger than he had anticipated.
“So we don’t need to know anything about the disease,” Brockmann explained. “If you throw a rock into the water, you’ll see a concentric wave. If you throw a big rock into the water, you will also see a concentric wave. If you throw a rock into honey, or some different liquid, you will see a wave that is propagating slower, but it will still be a concentric wave.”
Regardless of the transmission method, Brockmann contends that his global mobility network represents the wave pattern, while the rock and the liquid show the launch and spread of the disease.
All the network requires is a reexamination of the rules by which we move. According to Brockmann, the global mobility network doesn’t just have to apply to disease, either–it could be a social network, for example, and the contagion could be the spread of fads, invasive species carried by unwitting humans, or misinformation.
“As a physicist, we’re always interested in the things that different phenomena have in common,” he said. “We’re after the, you know, quote-unquote, ‘universal features of phenomena.’ So for this disease stuff, it was just redrawing the network in the right way.”
Using the global mobility network, we can see how a disease launched in one city might spread to the rest of the world. Brockmann developed visualizations of this phenomenon in Mexico City, Chicago, and London. Unfortunately, they also show that an outbreak in a global city is unlikely to be contained, or quarantined.
Brockmann was also able to use the same method to map an outbreak of E. coli in Germany. He came up with an accurate global mobility network based on the flow of food distribution trucks throughout the country.
But human nature also makes for kinks in the system. “One of the problems that we have in this business of predicting diseases is we rarely go to human decision processes,” Brockmann said. If Brockmann and his colleagues were better able to predict human responses to outbreaks, they might be able to build that into the model. It might turn out, for example, that large groups of people running away from a disease could contribute to a faster spread.
Hopefully, considering the spread of disease through a reorganized sense of time and space for the modern era could help cut off critical links in transmission. And now that researchers understand how the underlying process works, the model could be applied to just about anything.
“It’s almost like for 10 years we’d been poking around in the dark. We had models that could compute forecasts, but we didn’t really know why,” Brockmann said. “Now we finally understand what’s going on.”