Hypothesis Testing and Prediction of the Self-Triggering Cox Model for Recurrent Event Data

• Activity Number: 445
• Type: Contributed
• Date/Time: Tuesday, August 2, 2016 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Learning and Data Science
• Author(s): Jung In Kim* and Jason Fine and Feng-Chang Lin
• Companies: and The University of North Carolina at Chapel Hill and The University of North Carolina at Chapel Hill
• Keywords: recurrent event data ; Cox proportional hazard model ; cystic fibrosis

Abstract

Recurrent/Repeated event data frequently appear in longitudinal studies when study subjects experience more than one event during the observation period. In reality, one may observe the subsequent events influenced by previous events; hence, the triggering scheme of event occurrence shall be considered when modeling such data. In this paper, we extend the Cox proportional hazard model with time-varying information of previous events to enhance the model fitness and prediction. Parameter estimation and statistical inferences can be easily achieved via a partial likelihood function. A jointed statistical test is provided to assess the existence of the effects from previous events. We demonstrate our approach via comprehensive simulation studies and cystic fibrosis registry data in chronic pseudomonas infections. Significantly, our model provides a better prediction amongst currently existing ones.