Speaker
Description
Household transmission models are widely used in infectious disease
epidemiology, yet transmission ordering is typically ignored because it
does not affect final outbreak size. Consequently, many distinct
transmission histories produce identical epidemiological outcomes,
making it difficult to distinguish internal from external transmission
and limiting parameter identifiability.
Rather than introducing time-dependent transmission rates, we retain
the standard time-homogeneous continuous-time Markov chain formulation
and instead refine the household state representation. The model state
space is expanded to include transmission graphs describing infection
direction and order, while pathogen genomic data are used to restrict
the transmission histories compatible with observations.
Simulation studies show that incorporating genetic information
substantially sharpens the likelihood surface compared with models based
on epidemiological data alone. In particular, genomic data reduce the
dependence between internal and external transmission parameters,
removing the characteristic ridge associated with non-identifiability
and yielding more concentrated posterior distributions.
These results demonstrate that graph-resolved household models enable
improved transmission inference without sacrificing analytical or
computational tractability.
