Estimating the Reproductive Number of an Influenza Pandemic in Real-Time from Multiple Data Sources

Carlo G. Camarda, Max Planck Institute for Demographic Research
Jim Oeppen, University of Cambridge

An influenza epidemic is the outcome of an interaction between two populations – the virus and the human immune system – with very different demographic characteristics. Accurate real-time estimation of the parameters of this interaction, together with their confidence intervals, would be of enormous help to health planners. Conventional weekly reporting of influenza mortality and morbidity is being overtaken by real-time computerised medical record-keeping, rapid laboratory diagnosis, and indirect sources such as Google Flu trends. Our model aims to estimate the daily Reproductive Number of an epidemic in its growth phase (NRR in demographic terms) , while explicitly accounting for the mis-recording created by week-ends and public holidays. To evaluate the model we use data from two influenza pandemics: 1889-90 in Munich and 1918 in New York State. Treating these data as an ex ante estimation problem shows how well this key parameter might be estimated in a future pandemic.