Creating Connections: How Transportation Data can Predict Pandemics

A historical look at the bubonic plague’s trail of terror across Europe in the 1300s reveals interesting epidemiological data when compared to modern-day pandemics, such as SARS or the 2009 swine flu. The pattern of the plague’s spread was quite simple. It first struck in Sicily, from where it spread outward into neighboring countries in a wave-like pattern. Visually, this is similar to the ripple effect caused by throwing a pebble into a pond.

In the 21st century however, with the introduction of ever-more complicated transportation networks, patterns of disease spread have changed dramatically. In fact, a simple snapshot of the global spread of SARS will seem entirely sporadic and arbitrary without accompanying information. This taught epidemiologists that geographic distance is no longer a good indicator of the length of time it takes to travel between places, and thus, alone, may not be effective in predicting the spread of pandemics.

A New Predictive Model

In 2005, Dirk Brockmann PhD, an associate professor of engineering sciences and applied mathematics at Northwestern University, began research on a new method to predict epidemic and pandemic arrival and spread. This led to the development of a unique computational model that measures distance from an entirely new angle – by the amount of traffic that occurs between two locations, rather than by miles.

For instance, although New York City and London are geographically far apart, when compared to say, New York City and Milwaukee, the number of people traveling between New York and London is considerably higher, and thus, effectively, London is closer. “The amount of traffic is an indicator or measure of how close different places are,” says Brockmann. In today’s intricately connected world, geographic distance plays a much smaller role in the spread of infectious diseases.

“When we did this, my first impulse was ‘Okay, this must have been done before’ because it is so simple and so intuitive,” says Brockmann. “This is not only an academic pattern… If we look at it from a more intuitive notion of distance, a traffic-based notion, the effective distance… can predict the arrival time of an epidemic.”

Brockmann’s model depends on data from the worldwide air transportation network, which is comprised of some 4,000 airports and 40,000 connections. However, given its simplicity, it can be applied to any type of mobility network, including busses, trains, and even subways.

Similar to Brockmann’s approach, Dr. Kamran Khan, founder of Bio.Diaspora, is also dedicated towards leveraging the extensive global transportation network to better understand and prepare for public health threats. In collaboration with HealthMap, Khan uses GIS mapping to create an online accessible application that shows where infectious diseases are emerging and where they are likely to spread.  

In addition to revealing the onset of an epidemic and predicting its spread, transportation data can also be used to determine the location source of an outbreak. In fact, Germany’s 2011 E. coli epidemic inspired Brockmann to further develop his model.

“[In Germany] the situation was kind of reversed. There were all these cases and we wanted to know, where is this coming from?” The complexity of transportation networks and consequently the movement of food within Germany and across Europe made it difficult to ascertain the true vegetable culprit. Brockmann’s method can be used to determine the origin or starting point of the outbreak according to the pattern of spread. “With this technique, everything looks circular if you use the right reference point,” explains Brockmann, “So you can try various points, and if you find one where the pattern looks nice and simple, you found the right outbreak location.”

Computational epidemiology seems inherently complex. Traditionally, predictive models required scientists to run complicated and detailed computer simulations to make forecasts. But Brockmann hopes that this new model will open the door to additional basic and effective predictive models for global disease spread. “This is much simpler in its dynamic features than we originally thought it would be,” says Brockmann.

 

 

Brockmann presented his research at the American Association for the Advancement of Science (AAAS) annual meeting in Boston on Saturday, Feb. 16.

Related Posts