Analysis of recurrent event processes with dynamic models

Abstract

The analysis of past developments of processes through dynamic covariates is useful to understand the present and future of processes generating recurrent events. In this study, we consider two essential features of recurrent event processes through dynamic models. These features are related to monotonic trends and clustering of recurrent events, and frequently seen in medical studies. We discuss the estimation of these features through dynamic models for event counts. We also focus on the settings in which unexplained excess heterogeneity is present in the data. Furthermore, we show that the violation of the strong assumption of independent gap times may introduce substantial bias in the estimation of these features with models for event counts. To address these issues, we apply a copula-based estimation method for the gap time models. Our approach does not rely on the strong independent gap time assumption, and provides a valid estimation of model parameters. We illustrate the methods developed in this study with data on repeated asthma attacks in children. Finally, we propose some goodness-of-fit procedures as future research.