64th ISI World Statistics Congress - Ottawa, Canada

64th ISI World Statistics Congress - Ottawa, Canada

Multi-state models: an appraisal with an application to real data


Claudia Adriana Castro Kuriss


  • V
    Victor Leiva


64th ISI World Statistics Congress - Ottawa, Canada

Format: CPS Abstract

Keywords: "survival, data-analysis, multi-state

Session: CPS 05 - Statistical modelling VI

Monday 17 July 8:30 a.m. - 9:40 a.m. (Canada/Eastern)


Survival and reliability analysis are extremely popular and helpful tools to handle problems with censored data that arise from health sciences and engineering. The most common model is the Cox regression, and some extensions of this model are available. However, another problem frequently appears when the death is due to another cause rather than that followed, or more than one event is registered for an object under study. Here, events of the same type, like recurrences, can be considered. There are different approaches to coping with it as Counting Process, which uses the Cox proportional hazard model, and frailty models. For the last problem in particular, but also for others models(like for example, competing risks, mortality and disability models), there are more recent developments, such as the multistate models, where the object can be allocated each time in one state and move to other states, which are called transitions, or reaches an absorbing state (like death or destruction), which can no longer be abandoned. This can be visualized as an oriented graph with nodes as states and connections within them as edges. This is an appealing way to assess the problem and can be undertaken with the R software. One of this proposal's main challenges is establishing the possible states and interactions, presenting the data appropriately to formalize the problem, and statistically investigating it. The main objective of this work is to revise some of the available methods considered in multistate survival models, especially with R, and apply these results to study a real data set from the area of health sciences.