Assessing Bias Due to Competing Risk of Death in Traditional Survival Analysis of the Longitudinal Health and Retirement Study of the Elderly U.S. Population

• Activity Number: 554
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Jia Li* and Sharon Silver and Toni Alterman and Marie Sweeney and Walter Alarcon
• Companies: NIOSH and NIOSH and NIOSH and NIOSH and NIOSH
• Keywords: competing risk ; incidence ; longitudinal survey

Abstract

The Health and Retirement Study (HRS) is an ongoing national panel study conducted biannually since 1992. It surveys a representative sample of more than 26,000 Americans over the age of 50. In this study, we examined the association between smoking status and incidence of cancer. To assess the impact of competing risk of death in traditional survival analyses when the mortality rate changes, we stratified the sample by three age groups at baseline (50-59, 60-69, and 70+), and conducted both Cox proportional hazard regression analyses as well as Fine and Gray's competing risk regression analyses on the three subsets of data separately. Results showed that the Cox models can overestimate the hazard ratios and produce falsely significant statistical inferences. When the mortality rates increase from 13% in the youngest age group to 60% in the oldest age group, the magnitude of the bias also increased proportionally. Our example suggests that in an elderly population, the competing risk of death can be significant and needs to be taken into account in the analysis. Use of competing risk approaches is important in making valid statistical inferences for these studies.