monash_25660.pdf (2.28 MB)
An investigation of emergency department overcrowding using data mining and simulation : a patient treatment type perspective
thesisposted on 2017-01-05, 03:53 authored by Ceglowski, Andrzej Stefan
In ongoing efforts to limit the proportion of budgets allocated to healthcare, governments have closed hospitals, reduced bed numbers, and minimised staffing levels. These efficiency drives in Australia and elsewhere have led to the healthcare system being more sensitive to shocks and disruptions because there is little excess capacity to absorb extra demand. Among other impacts, this has resulted in hospital emergency departments (EDs) having to balance patient throughput with patient queues. This balance is often difficult to achieve and EDs become overwhelmed, resulting in long patient waits, overcrowded treatment areas and excessive stress for ED staff. Much research has been done in an effort to limit the frequency and severity of these “overcrowding” incidents. Projects have been driven by preconceived ideas about activities in EDs and often employ parametric methods to identify correlated factors. As such they are limited by the extent of analysts’ knowledge or observations. While these approaches have added to knowledge about ED operations, the problem of patient overcrowding persists. This research questioned whether patient treatment could be implicated in ED overcrowding. Process-based thinking was used in order to derive a simplified model of emergency department operations. This “process-focussed” model of ED operations directed thinking towards the identification of homogenous clusters of treatment with similar activities, so each treatment cluster could be considered to have matching inputs, outputs and resource consumption. Scientific Method was selected as an appropriate methodology for the research. Techniques from the dissociated methods of Data Mining and Management Science were combined within the hypothesis / experimentation framework of Scientific Method. Undirected clustering techniques from Data Mining were used to identify definitive treatment clusters. Discrete event simulation techniques from Management Science were used to drive the study towards the overcrowding problem. The treatment clusters were verified and validated through a number of studies. Process perspectives were employed together with the treatment clusters to simulate patient flows through the ED at an aggregated level. The clusters were combined with patient urgency and disposition to create “patient treatment types” that were tracked through the ED. Analysis of the simulated ED indicated that simultaneous occupation of the ED by certain patient types made the ED unable to accept any new patients for treatment. This thesis contributes to the understanding of ED overcrowding by confirming that exit block is the most likely direct cause of ED overcrowding, and by suggesting that the mix of patients types in, and arriving at, the ED are the most likely precursors of ED overcrowding. It concludes that there will always be a finite chance that a mix of patient types will occur who require admittance to hospital or have long ED treatment times, and consequently, are likely to block the ED. This suggests that it will never be possible to completely eliminate ED overcrowding. Rather, acceptable levels of risk of overcrowding need to be determined. The capacity of EDs and of hospitals to admit ED patients may then be determined based on how risk adverse the hospital is to ED overcrowding.