Bayesian inference was performed using a Markov chain Monte Carlo approach using Win-
BUGS version 1.4, which was also used to calculate outcome measures of interest, accounting
for parameter uncertainty [37]. The dynamic transmission model was implemented in Component
Pascal as a hard-wired plug-in to WinBUGS using the BlackBox Component Builder
(Oberon Microsystems). Full implementation details are presented in S1 Text, and model code
is available at http://goo.gl/htrvrk.