Biostatistics Journal Club: An Introduction to Multistate Models

This talk will describe multistate models as a generalization of survival, competing risk, and semi-competing risk models. Multistate models are distinguished from survival and competing risk models in that they allow intermediate states and two-way transitions between states. The transition functions can depend upon subject time-varying covariates, time-varying treatments, and limited subject state history. Multistate models have been applied to a wide variety of conditions. Functionals of the models, for example a weighted sum of expected time in desirable and undesirable states, can be used to define outcomes, and the weighting can be tailored to patient preferences. The talk will introduce these models, discuss estimation of the transition functions, informative censoring, available software packages, and some applications for ongoing projects.

Required Reading:

Application of Multistate Modeling to Clinical Data Analysis

Speaker:

David Stein, PhD, Brigham and Women’s Hospital