Continuous Dynamic Bayesian networks for predicting survival of ischaemic heart disease patients

This paper introduces a Dynamic Bayesian network (DBN) model for representing survival of patients suffering from ischaemic heart disease (IHD). The main purpose of the model is to investigate the potential association between patient variables, the risk of developing cardiovascular disease (IHD) and survival. Of particular interest is whether, a combination of risk factors known as Metabolic syndrome are the key variables of interest in determining IHD risk or whether there are others considered just as significant, such as age, smoking and BMI that are not associated with the syndrome. The resulting Dynamic Bayesian network provides a straightforward illustration of the causal relationships between patient variables, disease occurrence and survival with the aim of understanding patient needs and the possibility of highlighting health interventions. It is hoped that such a model can help inform patient management decisions by illustrating where a change in certain patient characteristics could produce a health improvement and reduced risk. The DBN has the additional capacity to allow the representation of repeated measures data where patient variables may be available at more than one time point.