Semiparametric Multi State Model for Time-To-Event Data

Survival Analysis is the study about time-to-event data. It stands apart from classical estimation, in the sense that it has censoring objects with incomplete information to be dealt with. Classical Survival models usually contain two events, of which one is treated as terminal event. Multi State Models (MSM) involve more than two states, of which some may be transient and others absorbing. The multi-state Markov model is a useful way of describing a process in which an individual moves through a series of states in continuous time. Multi-state models can be used to model the movement of patients between different states, such as, hospitalization, recovery, relapse and death. These models may offer a better understanding of the process due to transition specific nature of the events. Also, the estimated transition probabilities from one state to another throws more light on the nature of movements and the possible reasons behind the transitions. In this paper, a multi state model with three states is considered under Semiparametric multi state approach. This method enables us to identify transition specific covariates that throw more light on the entire transition process and the factors influencing the same. A Cox Proportional hazard multi state model is used to derive the necessary estimates and testing procedures are carried out using mstate package of R, a open source software.