Putter, Hein

ABSTRACT Multistate models offer a powerful framework for studying disease processes and can be used to formulate intensity‐based and more descriptive marginal regression models. They also represent a natural foundation for the construction of joint models for disease processes and dynamic marker processes, as well as joint models incorporating random censoring and intermittent observation times. This article reviews the ways multistate models can be formed and fitted to life history data. Recent work on pseudo‐values and the incorporation of random effects to model dependence on the process history and between‐process heterogeneity are also discussed. The software available to facilitate such analyses is listed.