Using self organising feature maps to unravel process complexity in a hospital emergency department: A decision support perspective

Abstract

In systems that are complex and have ill-defined inputs and outputs, and in situations where insufficient data is gathered to permit exhaustive analysis of activity pathways, it is difficult to get at process descriptions. The complexity conceals patterns of activity, even to experts, and the system is resistant to statistical modelling because of its high dimensionality. Such is the situation in hospital emergency departments, as borne out by the paucity of process models for them despite the continued and vociferous efforts of experts over many years. In such complex and ill-defined situations, it may be possible to access fairly complete records of activities that have taken place. This is the case in many hospital emergency departments, where records are routinely kept of procedures that patients undergo. Extracting process definitions from these records by self organized clustering is neither a pure technical analysis, nor a completely social one, but rather somewhere between these extremes. This chapter describe use of Self Organised Feature Maps to reveal general treatment processes - actual work practices - that may be monitored, measured and managed. © 2008 Springer-Verlag Berlin Heidelberg.