Use this URL to cite or link to this record in EThOS:
Title: Unified modelling for care of the elderly
Author: Garg, Lalit
ISNI:       0000 0004 2721 5212
Awarding Body: University of Ulster
Current Institution: Ulster University
Date of Award: 2009
Availability of Full Text:
Access from EThOS:
The overall objective of this thesis was to develop and validate mathematical and statistical methods to aid informed healthcare decision making to help hospital staff, managers and policy makers to ensure the quality of elderly care while at the same time reducing the cost. As the result of the work of this thesis, holistic methods are proposed to facilitate understanding of the healthcare process dynamics and management, monitoring and performance measurement of healthcare systems. These models are illustrated and validated using three different datasets: a historical dataset on geriatric patients from an administrative database of a London hospital, a nationwide dataset available from the English Hospital Episode Statistics database on stroke- related patients, and a 5 years' retrospective dataset of stroke-related patients admitted to the Belfast City Hospital. In this thesis we first present a non-homogeneous Markov model to compute key performance measures for the whole patient care system, including both hospital and community components. We then describe different ways of modelling patients' length of stay and clustering patients into meaningful groups and a novel mixture distribution is proposed to have a significantly improved fit to length of stay data. We propose two novel techniques based on survival trees; phase-type survival trees and mixed distribution survival trees, to cluster the patients with respect to their length of stay considering the importance and effect of various patient characteristics, such as gender, age at the time of admission and disease diagnosed and their interrelation with patients' length of stay. We then illustrate how these models can be used for better understanding the care system and extracting exceptional or interesting patient pathways based on a given criterion of interest in terms of probability of occurrence, cost or duration. Based on this work, a novel application of data mining technique called sequential pattern mining is proposed to identify anomalous sequential patterns which require attention for efficiently managing scarce healthcare resources. We present two novel models for optimally scheduling patient admissions to satisfy resource restrictions, resource requirement forecasting, budgetary estimations, and/or comparing different admission scheduling strategies for a care system. First, a model based on sequential pattern mining technique, which is simple, easy to implement, has better explainability. The second model is a more sophisticated non- homogeneous discrete time Markov model and can be used for more complex admission scheduling such as a variable number of admissions each day to allocate resources to satisfy the fluctuating demand for care services or resource constraints. In addition, we describe an extension of our phase type survival tree based analysis to examine the relationship between length of stay in hospital and destination on discharge among these patient groups and illustrate applications of this approach to patient pathway prognostication, capacity planning and modelling discharge delay and its detrimental effects on length-of-stay and cost of care of patients whose discharge have been delayed. In summary this thesis proposes models to provide useful information readily available to make strategic decisions for effective care management and improvement.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available