Survival analysis for recurrent event data: an application to childhood infectious diseases

摘要: Many extensions of survival models based on the Cox proportional hazards approach have been proposed to handle clustered or multiple event data. Of particular note are five Cox-based for recurrent data: Andersen and Gill (AG); Wei, Lin Weissfeld (WLW); Prentice, Williams Peterson, total time (PWP-CP) gap (PWP-GT); Lee, Wei Amato (LWA). Some authors compared these by observing differences that arise from fitting real simulated However, no attempt has made systematically identify components appropriate We propose a systematic way characterizing such using four key components: risk intervals; baseline hazard; set, correlation adjustment. From definitions interval set there conceptually seven permissible, which those previously identified. The two new variant termed 'total - restricted' (TT-R) 'gap unrestricted' (GT-UR) models. aim paper is determine data components. fitted sets childhood infectious diseases. LWA model not because it allows subject be at several times same event. WLW overestimates treatment effect recommended. conclude PWP-GT TT-R useful analysing data, providing answers slightly different research questions. Further, applying robust variance any does adequately account within-subject correlation.