Application of Negative Binomial Regression Models for Estimating Influenza-Associated Deaths Using the CDC 122 Cities Mortality Reporting System Data and Final Complete Mortality

• Activity Number: 452
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Po-Yung Cheng*+ and Lynnette Brammer
• Companies: CDC and CDC
• Keywords: negative binomial regression ; influenza ; pneumonia ; death

Abstract

The 122 Cities Mortality Reporting System (122 CMRS) provides the Centers for Disease Control and Prevention (CDC) with timely monitoring of pneumonia and influenza (P&I) deaths. Complete mortality data from the National Centers for Health Statistics (NCHS) are not available until 2-3 years after a death occurred. In this study, we estimated influenza-associated deaths using 122 CMRS P&I data and NCHS P&I data from 1976-1977 through the 2009-2010 respiratory seasons. We estimated P&I mortality for persons < 65 and 65+ years using negative binomial regression models incorporating weekly influenza surveillance data as covariates. The models were designed to estimate the number of influenza-related P&I deaths. After population adjustment, average annual total number of influenza-associated deaths was 6613 (95% CI: 4240-18359) using 122 CMRS data and 6232 (95% CI: 5022-11127) using NCHS data. The correlation coefficient between these two sets of estimates was 0.99. Based on the high correlation between these two estimates, 122 CMRS data can be used to provide timely estimates of annual P&I deaths associated with influenza in the United States.