Leonhard Held, University of Zurich 

Vendredi 7 décembre 2007 à 11h15, salle 5220

A major challenge in infectious disease epidemiology remains the statistical analysis of counts of notifiable diseases collected by national surveillance systems. Such data often show a regular pattern over time such as long-term trends or seasonality, but may also contain occasional outbreaks due to the infectiousness of the disease. In a surveillance setting, the available data are often spatially and temporally stratified while information about susceptibles is virtually never available.

Interdependencies between cases caused by different pathogens might particularly be of interest to further understand the dynamics of the diseases. Different pathogens infiltrating the same organ might interact or pathogen transmission between different population

subgroups (age groups, geographical areas etc.) are possible.

In this talk I will discuss some statistical models for the analysis of multivariate time series of infectious disease counts. The framework is applied to weekly surveillance counts on influenza and meningococcal disease in Germany, obtained from the German infectious

disease surveillance system administered by the Robert Koch Institute in Berlin. As a spatio-temporal example I consider the weekly number of deaths from influenza and pneumonia in the USA. The methods are particularly well suited for model validation based on one-step-ahead predictions, as will be illustrated.

This is joint work with Michaela Paul.