Suen, Daniel and Worthington, David and Sperrin, Matthew (2015) The development and application of an analytical healthcare model for understanding and improving hospital performance. PhD thesis, Lancaster University.
DSuenThesis.pdf - Published Version
Available under License Creative Commons Attribution-NoDerivs.
Download (1MB)
Abstract
Healthcare systems are tasked with balancing a variety of conflicting priorities such as increasing patient demand, minimising waiting times and a limited budget. With an ageing population the pressure on hospitals to maintain the quality of patient care is only going to rise as demand increases. Improving the management of these systems is important for avoiding potential dangers such as patient overcrowding and bed blockages which can result in or exacerbate problems such as increased risk to patient care, excessive patient lengths of stay, and staff burnout. This thesis develops and applies an analytical model, which uses queueing theory to approximate healthcare systems in order to provide a means of analysing these systems mathematically. In particular we consolidate, simplify and extend the theory underpinning both single node and networks of infinite server queues, using ideas and concepts developed in Gallivan and Utley (2005) and Massey and Whitt (1993) in order to derive explicit and easy to use formulae for the mean and variance of bed demand. We demonstrate the use of the analytical model by using the model outputs to produce model-based performance indicators to measure hospital performance and hence identify hospitals deserving further investigation. A difficulty in evaluating hospitals is determining how to measure their performance and produce fair and meaningful results while accounting for factors beyond their control, for example hospital size impacts on the relative variability of patient demand and needs to be incorporated into any analysis. We analyse the elective and emergency work of 30 hospitals using model-based performance indicators as a point of comparison for the observed results, allowing for hospital size. For the emergency work we focused on a single length of stay distribution but a key difference arose in the elective case, where we incorporated day-of-week dependent patient lengths of stay. In cases where day-of-week dependent length of stay data is not available, we also devise and evaluate a statistical approach for model calibration.