Analysis of episodic data with application to recurrent pulmonary exacerbations in cystic fibrosis patients

We consider a special type of recurrent event data, termed "recurrent episode" data, arising in episodic illness studies. When an event occurs, it lasts for a random length of time. A naive recurrent-event analysis disregards the length of the episodes, which may contain important information about the severity of the disease, the associated medical costs, and quality of life. Bivariate gap time models have been suggested in which length of episodes and time between episodes are modeled jointly. These models are useful but may obscure the overall effects of treatment and other prognostic factors. The analysis can be further complicated if covariate effects change over time, as may occur when the effects vary across episodes. This article reviews the existing methods applied to recurrent episode data and approaches the problem using the recently developed temporal process regression. Novel endpoints are constructed that summarize both episode frequency and the length of episodes and time between episodes. Time-varying coefficient models, with inferences based on functional estimating equations, are proposed. Both existing and new methods are applied to a clinical trial to assess the efficacy of a treatment for patients with cystic fibrosis, many of whom experienced multiple episodes of pulmonary exacerbations.