Joint Modelling of Longitudinal and Survival Data: A Comparison of Joint and Independent Models
Lisa McCrink, Adele Marshall, Karen Cairns
Centre for Statistical Science and Operational Research (CenSSOR), Queen's University Belfast, Belfast, United Kingdom

In the collection of longitudinal data, it is common to gather both repeated measures and survival data simultaneously, where the longitudinal variable is typically correlated with a patient's survival. As a result, independent models can cause biased estimates so it is necessary to use an alternative joint model approach [1].

A joint model provides longitudinal and survival estimates simultaneously and more accurately, with a clear reduction in the standard error of estimates compared to independent model estimates. The focus of this research is to demonstrate the benefits of using joint models to handle the typical association between repeated measures and survival data.

The medical applications of such models will be illustrated using UK Renal Registry data collected over a three year period commencing at the start of 2005. Joint models have been used to determine how the fluctuations in haemoglobin of 5860 haemodialysis patients, with a total of just over 59,000 observations, affects their survival. Previous research has found evidence associating fluctuations in haemoglobin over time with reduced survival rates for dialysis patients [2].

A multilevel mixed effects model has been joined with a Cox Proportional Hazards model to analyse the effects of clinical, biochemical and haematological variables on patient's fluctuations in haemoglobin over time and thus their survival.


[1] Henderson, R., Diggle, P. & Dobson, A. 2000, “Joint modelling of longitudinal measurements and event time data”, Biostatistics, vol. 1, no. 4, pp. 465-480.

[2] Gilbertson, D.T., Ebben, J.P., Foley, R.N., Weinhandl, E.D., Bradbury, B.D. & Collins, A.J. 2008, “Joint modelling of longitudinal measurements and event time data”, Clinical Journal of the American Society of Nephrology, vol. 3, no. 1, pp. 133-138.

Keywords: Longitudinal; Joint Model; Multilevel Mixed Model; Cox Proportional Hazards Model

Biography: Lisa McCrink is a PhD student at Queen's University Belfast, Northern Ireland. Her research focuses on the methodology of models used to handle unstructured longitudinal data, concentrating on medical applications. Lisa holds an MSci in Mathematics and Statistics & Operational Research from Queen's University Belfast and is a member of the Centre for Statistical Science and Operational Research (CenSSOR), Queen's University Belfast.