A Case Study of Time Scale for Use in Cox's Proportional Hazard Model

• Activity Number: 209
• Type: Contributed
• Date/Time: Monday, July 30, 2007 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Qimei He*+ and Bradley Willcox and David J. Curb
• Companies: Pacific Health Research Institute and Pacific Health Research Institute and Pacific Health Research Institute
• Keywords: PH regression model ; time scale ; follow-up study ; Biases

Abstract

Cox's proportional hazard (PH) regression model is a widely used tool in the analysis of survival data. Most of the time, follow-up time is used as the time scale in the PH model. However, if the covariate of interest is correlated with age, the PH model using follow-up time as the time scale will produce biased results, even after adjusting for age at baseline. The Honolulu Heart Program has followed 8,006 Japanese-American in Honolulu, Hawaii for over four decades. We used the PH model to explore the relation between height and mortality in the HHP and to illustrate the time scale issue in the Cox regression model. We found that height is positively associated with mortality when using age as the time scale; conversely, a negative association is seen with mortality when follow-up time is used as the time scale, using 20-years of follow-up.