B0202
Title: Continuous-time multi-state survival models for cross-sectional data
Authors: Ardo van den Hout - University College London (United Kingdom) [presenting]
Luz Sanchez-Romero - University College London (United Kingdom)
Abstract: A multi-state survival model describes change of status over time. The archetype example is the three-state illness-death model, where individuals are in state 1 when they are healthy, in state 2 when they are ill, and in state 3 when they are deceased. Typically, longitudinal data are available with repeated observations within individuals. When transition times between the states are interval censored, the data are called panel data. There is range of methods in the literature for fitting continuous-time multi-state survival models to panel data. With longitudinal data there is individual-specific information on the transition process during the follow-up. With cross-sectional data, this information is not available. In case the time scale of the process is age in years, cross-sectional data typically consist of a table with sample frequencies pertaining to different individuals for each age/state cross-classification. We consider the statistical modelling of a continuous-time multi-state survival process using cross-sectional data. Included topics are identifiability problems, hazard models, estimation, and goodness of fit. The modelling will be illustrated with a data analysis.
