Monday, February 1st, 12:00pm, 2411 Tupper Hall
Speaker: Yolanda Hagar (Postdoctoral Researcher, Applied Mathematics Department, University of Colorado Boulder)
Title: Modeling of Long-Term Survival Outcomes
Abstract: Modern data storage methods allow researchers to track patient health outcome information through different registries, clinical trials, and other studies. These methods yield large-scale databases containing long-term follow-up of patients, measures of high-dimensional patient characteristics, and multi-varying comorbidities. We examine a family of survival models designed to accommodate features of modern data using the multi-resolution hazard (MRH) methodology. The MRH model is a Bayesian, semi-parametric survival model for estimation of the hazard rate and the effects of covariates on survival time. The model has been extended to incorporate non-proportional hazards as well as periods of time with sparsely observed failures. A case study of long-term prostate cancer survivors is examined, using RTOG clinical trials data, and implementation performed via the “MRH” package in R. Future directions for incorporation of time-varying covariates and multiple outcomes are examined, presented with corresponding applications in electronic health records data and cleaner cook stove analyses.