A Robust Regression Model for Estimating Influenza-Associated Deaths in 122 Cities' Mortality Data

• Activity Number: 113
• Type: Contributed
• Date/Time: Monday, July 30, 2007 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Po-Yung Cheng*+ and William W. Thompson and David K. Shay
• Companies: Centers for Disease Control and Prevention and Centers for Disease Control and Prevention and Centers for Disease Control and Prevention
• Keywords: robust regression ; influenza ; death

Abstract

CDC uses a robust regression model (Andrews) to forecast the baseline for the percentage of all deaths attributed to pneumonia and influenza (P&I) deaths using the 122 Cities Mortality Reporting System. Influenza-associated death estimates were obtained using the Andrews weight function, a Serfling linear regression model and two other weight functions (Bi-square and Huber). For the Andrews model, an annual average of 1,383 (95% CI: 947 - 1,804) P&I deaths were associated with influenza. Using the percentage of the standard error over the parameter estimates, the results indicated that the Andrews model was the best fitting model. Using these results to project national estimates, the Andrews model estimated an annual average of 4,534 (95% CI: 3,101 - 5,922) P&I deaths associated with influenza. The robust regression model provides a timely estimate of deaths associated with influenza.