Algorithmical and Computational Procedures for a Markov Model in Survival Analysis

Abstract

A methodology for studying the homogeneous time-continuous Markov processes that allow the usual quantities of interest in survival studies to be expressed in a well-structured form is considered. This is performed by introducing the phase-type distributions, which enable algorithmic expressions for the performance measures used in survival analysis. A Markov model with several absorbent states is studied, incorporating the phase-type distributions. Following this procedure, we show that Markov models can be applied in a direct and tractable way. In this general model covariates are introduced and applied to analysing the behaviour of breast cancer. The computational implementation can be developed from the formulae proposed in a more convenient way than previous ones and this has been performed using the Matlab program.