The extended mixture distribution survival tree based analysis for clustering and patient pathway prognostication in a stroke care unit

Abstract

In our previous work we proposed a special class of survival distribution called Mixture distribution survival trees, which are constructed by approximating different nodes in the tree by distinct types of mixture distributions to achieve more improvement in the likelihood function and thus the improved within node homogeneity. We proposed its applications in modelling hospital length of stay and clustering patients into clinically meaningful patient groups, where partitioning was based on covariates representing patient characteristics such as gender, age at the time of admission, and primary diagnosis code. This paper proposes extended Mixture distribution survival trees and demonstrates its applications to patient pathway prognostication and to examine the relationship between hospital length of stay and/or treatment outcome. 5 year retrospective data of patients admitted to Belfast City Hospital with a diagnosis of stroke is used to illustrate the approach.