Doubly Truncated Negative Binomial Regression for Evaluating the Age Distribution of Pertussis Infant Deaths

• Activity Number: 257
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Andrew Lewis Baughman*+ and Tracy Pondo and Margaret Cortese and Barry Sirotkin and Charles E. Rose, Jr. and Tejpratap Tiwari
• Companies: CDC and CDC and CDC and CDC and CDC and CDC
• Keywords: Doubly truncated negative binomial ; mixture model ; age distribution ; pertussis ; infant mortality

Abstract

As part of a study to examine long-term trends in pertussis infant mortality, we evaluated the age distribution of pertussis deaths by month (0-11) using the U.S. vital statistics for 1933-2006. A two-part mixture model was developed that included a point distribution for deaths that occurred in the first month of life, and a doubly truncated negative binomial (DTNB) distribution for deaths that occurred during months 1 to 11. The conditional mean and variance of the DTNB distribution were derived and used to calculate the predicted average age at death as well as estimate the risk ratio for death, by time period, race, ethnicity, and gender. Results from fitting the mixture model indicated that the age distribution of pertussis deaths shifted towards younger infants. The average age of pertussis infant deaths decreased throughout the study period.