Factors affecting epidemics of African horse sickness in Spain were st
udied using a mathematical model. The model examined the likelihood of
an epidemic after the introduction of the virus, and the effectivenes
s of vaccination strategies. Two host species (horses and donkeys) and
one vector species (the biting midge Culicoides imicola) were include
d. A stratified random sampling method (Latin hypercube sampling) was
used for sensitivity analysis of the likelihood of an epidemic. System
atic variation of vaccination parameters was used to consider alternat
ive control strategies. In general, when an epidemic occurred most pot
ential hosts became infected. The peak prevalence in C. imicola was lo
w, and never exceeded 3%. The most significant factors in the likeliho
od of an epidemic were vector population size, the recovery rate in ho
rses and the time of year when the virus was introduced. The lag betwe
en virus introduction and protection, the proportion of hosts vaccinat
ed, and including donkeys in vaccination programmes where the factors
that most strongly affected the success of different vaccination strat
egies. These factors should be priorities for empirical research, and
should be considered in the design of control strategies in areas at r
isk of virus introduction.