People interact. They travel. And diseases might travel with them.

By Vittoria Colizza

Airneth Column
September 2009
By Dr. Vittoria Colizza

The current H1N1 influenza pandemic is just the last example of the crucial role that human mobility has on the spread of infectious diseases. Transportation systems keep expanding, becoming faster and more efficient, shrinking distances on the planet, and allowing increasing travel volumes. We are more mobile than ever before, and we can mix and interact at degrees and speeds that were unthinkable not so long ago. As a dangerous byproduct, we eased the way to pathogens to spread at unprecedented conditions, leaving our increasingly advanced world also increasingly vulnerable to biological invasions of pandemic proportions.

Human history has been mostly characterized by populations relatively isolated one from each other. Traveling means were limited and diseases would mainly spread through a slow diffusion phenomenon. Black Death in 14th century Europe spread at an approximate velocity of 200-400 miles/year1. The advent of modern transportation dramatically altered this picture. The 1957 influenza pandemic swiped the world in about 6 months, and it took approximately two months since the international alert to the H1N1 influenza virus to be declared a pandemic during spring 20092.

If the Boeing 747 is thought to be today’s most dangerous and efficient disease vector system3, how can we limit the rapid worldwide spread of diseases?

A key issue dominating the recent years discussions on mitigation plans against a pandemic is the effectiveness of restriction measures on the international air travel. In the very few real-world examples – as the reduced traffic to South East Asia during SARS outbreak4 and the decline in air travel after the 9/11 attack5 – the effect of reduced travel is hindered by the co-occurrence of several other factors that might affect the slowing down of an epidemic. However, computational models can help to shed light on the blurry future of an unfolding pandemic under several uncertainties. Recent findings based on data-driven computer simulations pointed out the limited benefit provided by feasible travel reductions in slowing down the global spread of an infectious disease6. Unless 99% effective, border controls and reductions on travel volumes would not delay the pandemic peak by more than 2-3 weeks.

The travel volume alone is unable to explain the extraordinarily rapid spread of an infectious disease and the mechanisms leading to the unexpected failure of travel restrictions in a pandemic situation. The rapid mixing between people due to mobility is phenomenal in numbers but most importantly it is unlimited by geographical distance or country boundaries. Flights compose an intricate web of connections, composed mainly by airports with poor/medium connectivity along with few hubs7, that dramatically favors the invasion of a disease hopping from city to city in the world8. A model network composed by exactly the same number of airports and connections, with connections distributed homogeneously among airports, would be less vulnerable to pandemics. Approximately a 10% reduction of the travel flows on the ‘homogeneous’ network would be enough to halt a pandemic. 99% or more is required in the complex and highly interconnected real world.

Although strategies based on travel flows are ineffective, analysis of traffic patterns can quantify and predict the risk of cases importation. Moreover, data on travelers infected by H1N1 allows to estimate epidemic incidence9, and to estimate unknown aspects of the new virus key to predictions for the future unfolding of the pandemic10. While the complex nature of human mobility make traditional strategies for disease control powerless, approaches that incorporate travel data can still prove to be crucially useful for understanding and facing epidemics in real time.


1. JD Murray, Mathematical Biology, second ed., Springer, New York, 1993.
2. M Chan, WHO Director-General, June 11, 2009. 
3. M Enserink, Science 313, 1555 (2006).
4. DA Skowronski et al, Ann Rev Med 56, 357 (2005).
5. JS Brownstein et al, PLoS Med 3, e401 (2006).
6. BS Cooper et al, PloS Med 3, e212 (2006), TC Germann et al, Proc Natl Acad Sci USA 103, 5935 (2006), NM Ferguson et al, Nature 442, 448 (2006), JM Epstein et al, PLoS One 2, e401 (2007), V Colizza et al, PLoS Med 4, e13 (2007)
7. A Barrat et al, Proc Natl Acad Sci USA 101, 3747 (2004) 
8. V Colizza & A Vespignani, J Theor Biol 251, 450 (2008)
9. C Fraser et al, Science 324, 1557 (2009), M Lipsitch et al, PloS One 4, e6895 (2009)
10. D Balcan et al, BMC Med 7, 45 (2009)

I thank D. Balcan, H Hu, B Goncalves, P Bajardi, C Poletto, JJ Ramasco, N Perra, M Tizzoni, W Van den Broeck for contributing the seeding data, and A. Vespignani for helpful comments. I am partially supported by the ERC Starting Grant EpiFor, and the EC Future Emerging Technologies projects Epiwork and Dynanets.

Category: news & columns, columns