Background: Accurately assessing the transmissibility and serial interval of an novel
human pathogen is public health priority so that the timing and required strength of
interventions may be determined. Recent theoretical work has focused on making best
use of data from the initial exponential phase of growth of incidence in large
populations.
Methods: We measured generational transmissibility by the basic reproductive
number R0 and the serial interval by its mean Tg. First, we constructed a simulation
algorithm for case data arising from a small population of known size with R0 and Tg
also known. We then developed an inferential model for the likelihood of these case data
as a function of R0 and Tg. The model was designed to capture a) any signal of the
serial interval distribution in the initial stochastic phase b) the growth rate of the
exponential phase and c) the unique combination of R0 and Tg that generates a specific
shape of peak incidence when the susceptible portion of a small population is depleted.
Findings: Extensive repeat simulation and parameter estimation revealed no bias
in univariate estimates of either R0 and Tg. We were also able to simultaneously
estimate both R0 and Tg. However, accurate final estimates could be obtained only
much later in the outbreak. In particular, estimates of Tg were considerably less
accurate in the bivariate case until the peak of incidence had passed.
Conclusions: The basic reproductive number and mean serial interval can be
estimated simultaneously in real time during an outbreak of an emerging pathogen.
Repeated application of these methods to small scale outbreaks at the start of an
epidemic would permit accurate estimates of key parameters.