Time-aggregated mobile phone mobility data are sufficient for modelling influenza spread: the case of Bangladesh

Human mobility plays a major role in the spatial dissemination of infectious
diseases. We develop a spatio-temporal stochastic model for influenza-like
disease spread based on estimates of human mobility. The model is
informed by mobile phone mobility data collected in Bangladesh. We compare predictions of models informed by daily mobility data (reference) with
that of models informed by time-averaged mobility data, and mobility
model approximations. We find that the gravity model overestimates the
spatial synchrony, while the radiation model underestimates the spatial synchrony. Using time-averaged mobility resulted in spatial spreading patterns
comparable to the daily mobility model. We fit the model to 2014–2017 influenza data from sentinel hospitals in Bangladesh, using a sequential version
of approximate Bayesian computation. We find a good agreement between
our estimated model and the case data. We estimate transmissibility and
regional spread of influenza in Bangladesh, which are useful for policy planning. Time-averaged mobility appears to be a good proxy for human
mobility when modelling infectious diseases. This motivates a more general
use of the time-averaged mobility, with important implications for future
studies and outbreak control. Moreover, time-averaged mobility is subject
to less privacy concerns than daily mobility, containing less temporal
information on individual movements