Forecast and Control of Epidemics in a Globalized
World
The application of mathematical modeling to the
spread of epidemics has a long history and was initiated by
Daniel Bernoulli’s work on the effect of cowpox inoculation
on the spread of smallpox in 1760. Most studies concentrate
on the local temporal development of diseases and
epidemics. Their geographical spread is less well
understood, although important progress has been achieved
in a number of case studies. The key question, as well as
difficulty, is how to include spatial effects and quantify
the dispersal of individuals. Today’s volume, speed, and
globalization of traffic, increasing international trade
and intensified human mobility promote a complexity of
human travel of unprecedented degree.
Figure 1: Nearly 95% of
the worldwide air transporation network. Lines
represent routes between connected airports, the
colorcode quantifies the number of passengers per day
along the routes.
The severe acute respiratory syndrome (SARS) which spread
around the globe in a matter of months in 2003 has not only
demonstrated that the geographic spread of modern epidemics
vastly differs from historic ones but also the potential
threat of emergent infectious diseases such as Hanta, West
Nile and Marburg fever which can be contained to their
endemic region only with increasing difficulty.
Particularly in the light of an imminent H5N1 influenza A
pandemic the knowledge of dynamical and statistical
properties of human travel is of fundamental importance and
acute.
In a this project we focused on the key mechanisms of the
worldwide spread of modern infectious diseases. In
particular we investigated to what extend the worldwide air
transportation network plays a role in disseminating
epidemics. To this end we compiled approximately 95% of the
entire traffic network, which amounts to nearly 2 million
flights per week between the 500 largest airports
worldwide. Some of the questions we aimed to answer were:
Is it possible to describe the worldwide spread of an
epidemic based on a conceptionally lean model which
incorporates the worldwide aviation network? If so, how
reliable are forecasts put forth by the model? How
important are fluctuations? What facilitates the spread?
And, last but not least, what containment strategies and
control measures can be devised as a result of computer
simulations based on such a model?
Figure 2:
(A) geographic spread of SARS in May
2003 as reported by the World Health Organization (WHO)
and the Center for Disease Control and Prevention
(CDC). The number of infecteds per country are encoded
by color. (B) The expected spread as
predicted by the model after a time of 90 days.
Depicted ist the expectation value of the number of
infecteds after 10000 simulations with an initial
outbreak in Hong Kong.
Our model consists of two parts: a local infection dynamics
and the global traveling dynamics of individuals. The local
dynamics is described by the SIR reaction kinetic scheme in
which individuals are initially susceptible (S), become
infected (I) and recover (R) from the disease, become
immune and cannot be infected a second time.
Furthermore we assumed that the transport between urban
areas is directly proportional to the traffic flux of
passengers between them. Both constituents of our model are
treated on a stochastic level, taking full account of
fluctuations of disease transmission, latency, and recovery
on the one hand and of the geographical dispersal of
individuals on the other.
Fig. 2 depicts a comparison of the spread of SARS in May
2003 as reported by the World Health Organization and
simulations of our model. Despite some differences the
overall prediction of our model is surprisingly close to
the observed global spread of SARS. The degree of agreement
seems particularly surprising if one recalls the conceptual
simplicity of our model and the fact that we incorporate
fluctuations in the infection dynamics as well as the
spatial dispersal.