Background: The early growth dynamics of the West African Ebola virus epidemic has been
qualitatively different for Guinea, Sierra Leone and Liberia. However, it is important to
understand these disparate dynamics as trends of a single epidemic spread over regions with
similar geographic and cultural aspects, with likely common parameters for transmission rates
and the reproduction number R0.
Methods: We combine a discrete, stochastic SEIR model with a two-scale community network
model to demonstrate that the different regional trends may be explained by different community
mixing rates. Heuristically, the effect of different community mixing rates may be understood as
the observation that two individuals infected by the same chain of transmission are more likely to
know one another in a less-mixed community. Saturation effects occur as the contacts of an
infected individual are more likely to already be exposed by the same chain of transmission.
Results: The effects of community mixing, together with the effects of stochasticity, can explain
the qualitative difference in the growth of Ebola virus cases in each country, and why the
probability of large outbreaks may have recently increased. An increase in the rate of Ebola cases
in Guinea in late August, and a local fitting of the transient dynamics of the Ebola cases in
Liberia, suggests that the epidemic in Liberia has been more severe, and the epidemic in Guinea
is worsening, due to discrete seeding events as the epidemic spreads into new communities.
Conclusions: A relatively simple network model provides insight on the role of local effects such
as saturation that would be difficult to otherwise quantify. Our results predict that exponential
growth of an epidemic is driven by the exposure of new communities, underscoring the
importance of control measures that limit this spread.