← Stochastic Processes for Biology

Extinction and outbreak probabilities

Stochastic epidemic models can assign probabilities to qualitatively different outcomes, such as early extinction or a substantial outbreak.

Early extinction

Even when a corresponding deterministic model predicts initial epidemic growth, a small number of infectious individuals can recover before producing enough new infections. The stochastic process can therefore reach the absorbing disease-free state.

Branching approximation

During the very early phase, when susceptible depletion is negligible, transmission can sometimes be approximated by a branching process. If each infectious individual independently produces offspring according to a distribution with probability-generating function \(G(s)\), the extinction probability \(q\) is the smallest solution in \([0,1]\) of

\[q=G(q).\]

Simulation estimate

For more complicated models, extinction or outbreak probability can be estimated from repeated simulations by the fraction satisfying a clearly defined outcome criterion.

Key idea. A reproduction threshold describes whether growth is possible on average; stochastic modelling additionally quantifies the chance that an outbreak succeeds or dies out.