A Longitudinal Comparison of Supervised and Unsupervised Learning Approaches to Iso-Resource Grouping for Acute Healthcare in Australia

Abstract

Estimating resource consumption of hospital patients is important for various tasks such as hospital funding, and management and allocation of resources. The common approach is to group patients based on their diagnostic characteristic and infer their resource consumption based on their group membership. This research looks at two alternative forms of grouping of patients based on supervised (classification trees) and unsupervised (self organising map) learning methods. This research is a longitudinal comparison of the effect of supervised and unsupervised learning methods on the groupings of patients. The results for the four-year study indicate that the learning paradigms appear to group patients similarly according to their resource consumption.