A Novel Application of a Mutual Information Measure for Analysing Temporal Changes in Healthcare Network Graphs.

We have previously demonstrated that network graphs generated using patient administrative data can represent health services functional structures through which hospital inpatient care is delivered. However, hospitals have to respond to changes in the external environment within which they function. These changes affect the network graphs and can be captured with a sequence of graphs of the same system over time. We developed a novel method for measuring temporal changes in healthcare network graphs based on the concept of mutual information, modified to allow for modest variations in node identities. The method is straightforward to apply, and we demonstrate its efficacy in a series of monthly snapshots of the functional structures in two hospitals.