Comparison of Methods for Estimating Influenza-Associated Deaths from the 1972/1973 Through 2003/2004 Seasons

• Activity Number: 315
• Type: Contributed
• Date/Time: Tuesday, August 5, 2008 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Hong Zhou*+ and William W. Thompson and Po-Yung Cheng and Eric Weintraub and Cecile Viboud and Corinne Ringholz and Joe Bresee and David Shay
• Companies: Centers for Disease Control and Prevention and Centers for Disease Control and Prevention and Centers for Disease Control and Prevention and Centers for Disease Control and Prevention and National Institutes of Health and National Institutes of Health and Centers for Disease Control and Prevention and Centers for Disease Control and Prevention
• Keywords: influenza-associated deaths ; Serfling regression model ; robust regression model

Abstract

Several methods for estimating influenza-associated deaths were compared, including Serfling regression models, robust regression models, and a Serfling spline regression model. Weekly underlying pneumonia and influenza (P&I) deaths were modeled from the 1972/73 through 2003/04 respiratory seasons. Influenza epidemic periods were defined by using either P&I deaths or influenza deaths. Ninety-five percent confidence limits were used to define epidemic weeks. The average annual numbers of epidemic weeks estimated by these methods were similar (range11-14) and significantly correlated (r>=.44). Epidemic weeks defined by one Serfling model and the Serfling spline regression model were significantly correlated with epidemic weeks defined by influenza viral data (r>=.52). Annual influenza-associated deaths estimated by all methods were similar (range 5643-8403) and highly correlated (r>=.85).