Multi-state models and cost-effectiveness analysis

Abstract

Multi-state models provide a relevant tool for studying the observations of a continuous-time process at arbitrary times. the underlying process is often supposed to be Markov for sum-up the history. these models have been successfully implemented in various applications. SemiMarkov processes provide an interesting alternative to describe a succession of observations. indeed, the process evolution depends on the time spent in the current state. the distribution of sojourn times is a parameter of model whereas an exponential distribution is implicitly supposed in Markov models. a review of the main estimation methods for various multi-state models is proposed. First, the homogeneous Markov model (constant intensities) and non homogeneous Markov models (time-dependent intensities) are considered. a parametric estimation method in semi-Markov models is then proposed. Distributions of the sojourn times can be chosen and maximum likelihood estimation in a parametric framework is considered. Cost-effectiveness analysis based on multi-state Markov models is discussed. indeed, a cost model can easily be considered to evaluate costs and effectiveness of an intervention. application to severe asthma follow-up will illustrate the potential of multi-state models.