When humans travel, diseases travel with them. Events like the 2003 SARS outbreak and 2009 H1N1 flu pandemic have revealed the role that air travel plays in the spread of illness on a global scale. Now, researchers at MIT have developed a computer simulation that suggests which U.S. airports are most likely to contribute to the spread of a pathogen in the first few days of a global health crisis. It's a glimpse of where our next epidemic could start.

The new model was created by a team led by MIT computer engineer Ruben James, and is the first of its kind to simulate the propagation of an infectious disease during the first ten days of a contagion, throughout what's known as the "early-time behavior" of a spreading epidemic (past models, by comparison, have examined the final stages of the contagion process). Unlike previous models, this one also considers things like an airport's geographic location, the patterns that passengers typically follow while traveling, how flights connect between airports, and average waiting times between successive flights.

The team's observations indicate that the airports most likely to spread disease at the beginning of an outbreak are significantly different from those implicated in the later stages of a public health crisis. To address this discrepancy, the researchers propose a new metric — called the "geographic spreading centrality," or "GSC" — that they claim can be used to "identify and rank influential spreaders of infectious diseases in human transportation networks." In other words: it allows them to predict which airports pose the greatest risk to public health and safety, should an outbreak originate in their home city. When they ran their simulation (a visualization of which can be seen here) through its paces, it came up with a small list of "early-time super-spreaders," led by New York's J. F. Kennedy (JFK), Los Angeles International (LAX) and Honolulu (HNL).


JFK and LAX aren't really too surprising. But Honolulu? That, say the researchers, is strange. To help make sense of this finding, they compare it to Atlanta International (ATL) — an airport which, prima facie, one would expect to rank among the small group of super-spreaders, yet somehow barely manages to slip into the top ten:

Classical measures of centrality, such as total traffic or connectivity, would suggest that Atlanta International airport (ATL) would have the largest spreading ability. This is clearly not the case, as it ranks 8th in terms of spreading power. The reason is that much of that traffic is of regional nature, within North America, and that many of the connected airports are not, themselves, strong spreaders. The GSC metric allows for a rationalization of the surprising fact that an airport like Honolulu (HNL) ranks third in early-time spreading, very close to JFK and LAX. Despite having a relatively low connectivity and total traffic, HNL combines three important features that catalyze contagion spreading: (1) it is dominated by long-range travel; (2) it is well connected to other massive hubs, which are themselves powerful spreaders; and (3) it is geographically located such that East-West travel is balanced, thereby maximizing TSD growth [TSD stands for "total square displacement" and is used by epidemiologists to measure the physical spread of disease].


James explained in a statement that his team's findings "could form the basis for an initial evaluation of vaccine allocation strategies in the event of an outbreak, and could inform national security agencies of the most vulnerable pathways for biological attacks in a densely connected world."

James also said that "if ground zero for the zombie apocalypse is in New York, L.A. or Honolulu, we're all pretty much humped." Just kidding, he didn't say that. But he probably should have.

The researchers' findings are published in the latest issue of PLoS ONE (no subscription required).