The Public Health Burden of Influenza: Clustering, Modeling, and Predicting Incidence for Diseases Associated with the Influenza

• Activity Number: 513
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Fan Tang*+ and Joseph E. Cavanaugh
• Companies: University of Iowa and University of Iowa
• Keywords: model based clustering ; structural model ; state space model ; forecasting

Abstract

Investigating the associations among common diseases that follow a similar incidence pattern is of great interest in public health. Influenza has been linked with many seasonal diseases. However, diseases that follow a similar seasonal pattern to that of influenza may not necessarily be causally related. In this work, the first problem we consider is to identify empirically homogeneous diseases with influenza based on similarities in the temporal dynamics. We employ a structural time series model that characterizes each candidate disease incidence series as an additive combination of three latent processes: a long-term trend, seasonal variation, and a local anomaly. Correlation measures based on local anomaly and seasonal dynamics are applied to cluster the disease series. The second problem is to augment the state space models to evaluate the extent to which influenza can predict the incidence of those diseases to which it might be causally related. Specifically, we examine the utility of influenza activity in predicting incidence during an out-of-sample test year.