Restricted Mean Models for Transplant Benefit and Urgency

• Activity Number: 126
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Section on Health Policy Statistics
• Author(s): Fang Xiang*+ and Susan Murray
• Companies: University of Michigan and University of Michigan
• Keywords: Dependent censoring ; Pseudo observation ; Restricted mean life ; Survival ; Time-dependent covariates ; Transplant benefit

Abstract

National lung allocation policy relies on statistical estimation of each individual's urgency and transplant benefit in defining a lung allocation score (LAS),both of which require accurate estimation of waitlist days lived.Risk factors are available to estimate patient urgency at their listing time,with more urgent patients removed from the waitlist as they either die or get transplanted.LAS is highly linked to both a patient's survival time and censoring (transplant) time. Therefore,it is crucial to adjust for dependent censoring in modeling estimated days of life. For estimation relevant to modeling an individual patient's urgency, we develop a model for the restricted mean as a function of covariates, using pseudo observations that account for dependent censoring. Simulation results show that our method performs well in situations comparable to the lung waitlist setting. A restricted mean model is also used to estimate days lived post-transplant based on individual risk factors at the time of transplant. The difference in LAS score for an individual, when properly accounting for dependent censoring, has high impact on the priority and timing of an organ offer for these patients