Dynamic prediction with penalized joint frailty model of high-dimensional recurrent event data and a survival outcome
The Annals of Applied Statistics · 2025 · Volume 19 · Issue 4
Abstract
High-dimensional recurrent event data have become increasingly popular these days. For example, many of the data captured in electronic health records (EHR) may be viewed as recurrent event data. In addition, a survival outcome is the primary outcome of interest in many studies. It is of great interest to study the associations between recurrent events and the survival outcome as well as to build dynamic prediction models. Joint frailty models are commonly used to study the associations between recurrent events and the survival outcome as well as to build dynamic prediction models. However, existing methods only consider one or a few recurrent events and cannot deal with high-dimensional recurrent events. In this article we propose a novel penalized joint frailty model that can deal with high-dimensional recurrent events. We adopt proportional hazards models for both the recurrent event submodels and the survival submodel. In addition, we utilize a penalized likelihood approach with an adaptive lasso penalty to select the informative recurrent events related to the survival outcome. A computationally efficient algorithm is also developed for model estimation based on the Gaussian variational approximate inference method, which can be implemented using the PJFM package in R. We conduct extensive simulation studies to evaluate the performance of our proposed method. The performance of our proposed method is further demonstrated on an EHR dataset of patients with type 2 diabetes mellitus.
Keywords
Adaptive lasso penalty; Dynamic prediction; high-dimensional recurrent event data; joint frailty model; variational approximation;
Enable JavaScript for citation tools and additional paper interactions.