Online Program

282238
Comparing methods using auxiliary information in estimating the restricted mean lifetime in the presence of dependent censoring


Monday, November 4, 2013

Tzu-Ying Liu, Department of Biostatistics and Department of Mathematics, University of Michigan, Ann Arbor, MI
Introduction In survival analysis, when the mean lifetime is of primary interest but the longest observation is censored, the mean lifetime will not be defined. In addition, it is not infrequent that the assumption of independence between the censoring time and the event time fails. Therefore, we are interested in estimating the restricted mean lifetime, the average survival time out the next fixed period, under the presence of dependent censoring. In particular, since auxiliary information that either predicts a subject's probability of being dependently censored, or predicts future event time is often available, we focus on comparing three methods that exploit these information: the inverse probability of censoring weighting (IPCW), the joint latent class modeling (JLCM), and the non-parametric risk set imputation.

Method We will generate simulation data in which the event times and the censoring times are associated through an internal time-dependent covariate. Then, we will estimate the restricted mean lifetime based on the three methods: IPCW, JLCM, non-parametric risk set imputation, and a naïve method: the usual proportional hazard model that ignores dependent censoring, and compare the consistency and efficiency of these four estimators.

Conclusion To our knowledge, JCLM has not been applied to the estimation of the survival time in the presence of dependent censoring and no published comparison has been made on estimators for the restricted mean lifetime based on these three approaches. With this study, we will inform clinical researchers reliable and precise options when dependent censoring is expected.

Learning Areas:

Biostatistics, economics

Learning Objectives:
Compare methods using auxiliary information in estimating the restricted mean lifetime in the presence of dependent censoring.

Presenting author's disclosure statement:

Qualified on the content I am responsible for because: I have completed two years of training in the master program of biostatistics at the University of Michigan. Before studying biostatistics and mathematics, I have completed my medical eduction and two years of residency in Taiwan. I am interested in survival analysis and longitudinal analysis with application to clinical research. One of my major career goal is to make these statistical methods approachable to public health researchers.
Any relevant financial relationships? No

I agree to comply with the American Public Health Association Conflict of Interest and Commercial Support Guidelines, and to disclose to the participants any off-label or experimental uses of a commercial product or service discussed in my presentation.