Abstracts:

 
Theresa Smith, Bath
 
Title: A stratified age-period-cohort model for spatial heterogeneity in all-cause mortality
 
Summary: A common goal in modelling demographic rates is to compare two or more groups. For example comparing mortality rates between men and women or between geographic regions may reveal health inequalities. A popular class of models for all-cause mortality as well as incidence of specific diseases like cancer is the age-period-cohort (APC) model. Extending this model to the multivariate setting is not straightforward, because the univariate APC model suffers from well-known identifiability problems. Often APC models are fit separately for each stratum, and then comparisons are made post hoc. A stratified APC model is introduced to directly assess the sources of heterogeneity in mortality rates using a Bayesian hierarchical model with matrix-normal priors that share information on linear and nonlinear aspects of the APC effects across strata. Computing, model selection, and prior specification are addressed and the model is then applied to all-cause mortality data from the European Union.
 

 
Ruwanthi Kolamunnage-Dona
 
Title: Joint modelling of longitudinal data and event-times with applications in health research
 
Abstract: Joint modelling of longitudinal data and event-time processes has gained its popularity in last decade as they yield more accurate and precise estimates. However, adopting this framework in health research has been limited. For example, in many clinical trials with longitudinal outcome data, a common situation is where some patients withdraw or dropout from the trial before completing the measurement schedule but the dropout may be non-ignorable. In such cases, the longitudinal outcome data alone may not reflect a genuine change over time, it may be an artefact caused by selective dropout, which could result in a biased comparison between the treatment groups.  In other research, a relatively large number of quantities such as biomarkers are measured over patients’ follow-up over time to fully explore the damage caused by adverse clinical events, and harnessing all such information in a single model could lead to improved estimation and prediction. In this talk, the methodology of joint modelling and its advances for competing risks and multiple longitudinal outcomes will be discussed with real applications in health research.