Modifications of and Alternatives to the SMR in Evaluating Center-Specific Mortality

• Activity Number: 481
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Kevin He*+ and Douglas E. Schaubel
• Companies: University of Michigan and University of Michigan
• Keywords: Non-proportional hazards ; Standardized Mortality Ratio ; Stratification ; Center Effect ; Kernel smoothing ; Sup test

Abstract

Estimating center effects through fixed or random effects typically involves the unrealistic assumption of proportionality among center-specific hazards. As a solution, the standardized mortality ratio (SMR) has been used. However, existing asymptotic properties of SMR are based on person-year methods or parametric models with simple intensity functions. The behavior of the SMR under a Cox model is not well-studied. In this study, we first provide a rigorous examination of the SMR under the null and alternative hypotheses. We then develop alternative methods, which are based on the stratified Cox model and hence remedies some strong limitations of the usual measure. A sup test is developed to evaluate whether the center effects are constant over time. The finite-sample properties are examined in simulation studies. The proposed methods are applied to national kidney transplant data.