A0324
Title: Semiparametric analysis of multivariate recurrent events with informative censoring
Authors: Yang Li - Indiana University School of Medicine (United States) [presenting]
Bin Zhang - Cincinnati Children's Hospital Medical Center (United States)
Abstract: In healthcare and clinical studies, recurrent events are frequently encountered both during hospitalizations and after hospital discharge. By recurrent events, we mean that one subject can potentially experience the same type of event repeatedly. In practice, it is common that two or more related types of recurrent events are of interest during the follow-up and thus multivariate recurrent events (MREs) arise. A possible complicating factor in many recurrent event studies is informative censoring. Compared to rate or intensity models, MRE mean functions can be clinically more interpretable especially when the event recurrence is likely fatal. We consider a semiparametric regression analysis on the mean functions with informative censoring. For modeling capacity and flexibility, both additive and multiplicative covariate effects are included. Marginal models will be employed to avoid distributional assumptions or specified correlation structures between MREs and informative censoring. An estimating-equation based inference procedure is developed for both the parametric and nonparametric components. The simulation study shows that the proposed inference procedure performs well. The proposed approach is applied to analyze a motivating dataset from the Mother's Gift Study to evaluate the effectiveness of maternal influenza vaccine and infant pneumococcal conjugate vaccine (PCV7) in reducing infant illnesses in Bangladesh.