Assessing the Impact of a Movement Network on the Spatiotemporal Spread of Infectious Diseases

Authors

  • Birgit Schrödle,

    1. Division of Biostatistics, Institute for Social and Preventive Medicine, University of Zurich, Hirschengraben 84, 8001 Zurich, Switzerland
    Search for more papers by this author
  • Leonhard Held,

    Corresponding author
    1. Division of Biostatistics, Institute for Social and Preventive Medicine, University of Zurich, Hirschengraben 84, 8001 Zurich, Switzerland
      email:  leonhard.held@ifspm.uzh.ch
    Search for more papers by this author
  • Håvard Rue

    1. Department of Mathematical Sciences, Norwegian University of Science and Technology, 7491 Trondheim, Norway
    Search for more papers by this author

email:  leonhard.held@ifspm.uzh.ch

Abstract

Summary Linking information on a movement network with space–time data on disease incidence is one of the key challenges in infectious disease epidemiology. In this article, we propose and compare two statistical frameworks for this purpose, namely, parameter-driven (PD) and observation-driven (OD) models. Bayesian inference in PD models is done using integrated nested Laplace approximations, while OD models can be easily fitted with existing software using maximum likelihood. The predictive performance of both formulations is assessed using proper scoring rules. As a case study, the impact of cattle trade on the spatiotemporal spread of Coxiellosis in Swiss cows, 2004–2009, is finally investigated.

Ancillary