Surveillance data on various notifiable diseases usually consist of multiple time series of daily, weekly, or monthly counts of new infections. Data are typically reported in several strata defined through administrative geographical areas, gender and/or age groups. Statistical modeling of the resulting multivariate time series is an important task in infectious disease epidemiology. We will discuss time series models - specifically developed for multivariate surveillance count data - that can be used for two distinct roles, understanding and prediction of disease spread.

The surveillance task when faced with small area health data is more complex than in the time domain alone. Both changes in time and space must be considered. Such questions as ‘where will the infection spread to next?’ and, ‘when will the infection arrive here’, or ‘when do we see the end of the epidemic?’ are all spatially specific questions that are commonly of concern for both the public and public health agencies.  Hence both spatial and temporal dimensions of the surveillance task must be considered.

Overview of the Institute for Health Metrics and Evaluation

Presenter

William Heisel, Director of Global Engagement, IHME