Dr Kari Auranen, Senior Scientist, Department of Vaccination and Immune Protection
National Institute for Health and Welfare, Finland
How uncertain should I be? – experiences from epidemiologic modelling
Models are increasingly used to predict outcomes of intervention programmes, e.g., to assess the likely effectiveness of large-scale vaccination. Although statistical modelling seems appropriate when addressing questions about such future outcomes, predictive models in epidemiology have defied formal requirements on coherent statistical reasoning. This is at least partly due to the need of causal inferences and more generally to uncertainties in choosing appropriate model structures. Luckily, when nature has spoken, predictive models become useless. Models may then serve as basis to address more scientific questions. In this talk I discuss different aspects of uncertainty in modelling infectious disease data for answering scientific questions on one hand, and for making predictions about intervention programmes on the other. The view will be a practical one, with examples from my recent work on modelling vaccination against pneumococcus and varicella.
