Methods for Direct Modeling of Restricted Mean Survival Time for General Censoring Mechanisms and Causal Inference

survival analysis;semi-parametric modeling;dependent censoring;instrumental variable;Public Health;Health Sciences;Biostatistics

Wang, XinMurray, Susan ;

University of Michigan

关键词: survival analysis;    semi-parametric modeling;    dependent censoring;    instrumental variable;    Public Health;    Health Sciences;    Biostatistics;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/140944/vivianwx_1.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Restricted mean survival time (RMST) is often of great clinical interest in practice. Several existing methods involve explicitly projecting out patient-specific survival curves using parameters estimated through Cox regression. It is often be preferable to directly model the restricted mean, for convenience and to yield more directly interpretable covariate effects.In the first chapter, we propose generalized estimating equation methods to model RMST as a function of baseline covariates. The proposed methods avoid potentially problematic distributional assumptions pertaining to restricted survival time. Unlike existing methods, we allow censoring to depend on both baseline and time-dependent factors. The methods are motivated by the end-stage liver disease (ESLD) setting and, in particular, consider survival in the absence of the preferred therapy, liver transplantation.In the second chapter, we propose generalized estimating equation methods to fit RMST models with multiplicative covariate effects.The proposed methods are applicable to several frequently occurring set-ups not considered in Chapter 1, including clustered data and data with a high-dimensional categorical covariate (e.g., center). Our proposed methods are motivated by modeling RMST among End-stage Renal Disease (ESRD) patients, in the presence of a high-dimensional covariate (1 million patients from over 5,000 dialysis facility).Estimation proceeds through a computationally efficient two-stage algorithm. In addition to evaluating large- and finite-sample properties, we demonstrate the considerable computational advantages of the proposed techniques.The third chapter is motivated by estimating the causal treatment in the presence of unmeasured confounding. We propose two-stage Instrumental Variable techniques for censored data. In particular, we develop closed-form, two-stage estimators for the causal treatment effect using an additive RMST model. Large sample properties are derived, with simulation studies conducted to assess finite sample properties. We apply the proposed methods to estimate the causal effect of peritoneal dialysis (PD) versus hemodialysis (HD) among End-Stage Renal Disease (ESRD) patients.