Recurrent Event Data Analysis with Time-Dependent Covariates: The Impact of Streptococcus on Pharyngitis in Indian School Children

• Activity Number: 59
• Type: Topic Contributed
• Date/Time: Sunday, July 29, 2012 : 4:00 PM to 5:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Chiung-Yu Huang*+ and Dean Follmann and Shanshan Li and Richard M Krause and John Melbin
• Companies: National Institute of Allergy and Infectious Diseases and NIH/NIAID and The Johns Hopkins University and Fogarty International Center/National Institutes of Health and Christian Medical College
• Keywords: Kernel Smoothing ; Nonparametric imputation ; Survival analysis

Abstract

Although recurrent event data analysis has received much attention in the past few decades, rigorous studies on modeling and estimation of the effects of time varying covariates on the risk of experiencing recurrent events have been lacking. Estimation of recurrent event models usually assumes that the entire covariate process is observed. In practice, however, covariates are often observed periodically and the last observed value of the covariate is carried forward until it is observed again. This paper investigates the bias induced by the last covariate carried forward method in the setting of recurrent event analysis, and proposes alternative methods: a parametric likelihood approach based on an underlying Poisson assumption and a new semiparametric estimator for the proportional rate. The latter approach is based on an estimated score function where we kernel smooth the covariate process. The different methods are applied to an observational study designed to assess the effect of Group A streptococcus (GAS) on pharyngitis among school children in India.