With a rise in the length of time that people are living, comes an increase in the pressure which hospital departments face in providing care for elderly patients. The research presented in this thesis introduces the conditional Coxian phase-type distribution, a novel methodology which may be used to describe the movement of elderly patients between various stages of hospital and community care. The rate parameters governing the movement of patients within each stage of care are conditioned on the length of stay experienced at the previous stage of care. Two additional methodologies are presented, extending the conditional Coxian phase-type distribution to account for the natural heterogeneity present in elderly patient length of stay in care. Firstly, a phase-type survival tree is used to partition elderly patients into subgroups (or cohorts), before the conditional Coxian phase-type distribution is used to model the pathway of each patient subgroup through the remaining stages of care. Secondly, patient covariates are included directly into the probability density function for the conditional Coxian phase-type distribution; through the use of log- linear equations, further reducing the assumption of homogeneity. To illustrate the results, a separate methodology is presented, alongside the aforementioned approaches, which allows the estimation of predictive intervals, to predict when a new elderly patient admission or discharge is likely to leave a given stage of care. Such an approach would enable patient-centred length of stay predictions to be made at the beginning of each stage of care. This would mean that community care may be organised ahead of time and according to these predictions so that instances of hospital bed blocking, together with the number of potential readmissions, may be reduced. Two sets of elderly patient hospital readmission data, from the Lombardy and Abruzzo regions of Italy, are used as applicative examples of each approach.
Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:726636 |
Date | January 2017 |
Creators | Gordon, Andrew Samuel |
Publisher | Queen's University Belfast |
Source Sets | Ethos UK |
Detected Language | English |
Type | Electronic Thesis or Dissertation |
Page generated in 0.2633 seconds