Guidelines for Generating Right-Censored Outcomes from a Cox Model Extended to Accommodate Time-Varying Covariates

• Activity Number: 39
• Type: Contributed
• Date/Time: Sunday, August 9, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Computing
• Author(s): Maria Montez-Rath* and Kristopher Kapphahn and Maya Mathur and Aya Mitani and David Hendry and Manisha Desai
• Companies: Stanford University School of Medicine and Stanford University and Stanford University School of Medicine and Boston University and Aarhus Universitet Institut for Statskundskab and Stanford University
• Keywords: right-censored outcomes ; extended Cox model ; time-varying covariates ; time-dependent covariates ; simulation studies ; censoring distribution

Abstract

Simulating studies with characteristics that closely resemble a motivating data set is crucial in some research projects. We were particularly interested in simulating studies with right-censored outcomes as functions of time-varying covariates. Hendry implemented Zhou's method, which generates survival times that follow a Cox model with time-varying covariates. We provide guidelines on the use of the algorithm and through simulation studies we characterize the sensitivity of the method to user inputs. We recommend the use of a Weibull distribution, which allows for a better specification of the baseline hazard in a wide range of scenarios and whose parameters can be informed from the data one wants to mimic. We highlight how some specifications do not affect the estimated effects but have an impact in the overall run times. This is particularly true for the choice of censoring distribution being implemented.