Proportional hazards modelling for current lifetime data

Abstract

Current lifetime data represent a unique subset of length-biased data where only the times from the initiating event to the cross-section date are observed without follow-up. We propose an efficient estimation procedure to fit the semi-parametric proportional hazards model using current lifetime data. This estimation procedure is based on the EM algorithm which has two versions based on the support points of the non-parametric baseline hazard function. We apply the method to simulated data and Parkinson’s disease current lifetime data drawn from the Canadian-Open Parkinson Network (C-OPN). We estimate the effects of clinical and epidemiological covariates using the proportional hazards model on the survival of subjects with Parkinson’s disease.