Optimal Event-Enhanced Heterogeneous Dynamic Knowledge Graph construction for patient clinical pathway analysis

Abstract

Purpose:: Clinical pathways (CPs) leave traces of patients and play a crucial role in the decision-making process of healthcare professionals. The purpose of this study is to uncover patient CPs from heterogeneous data and leverage the formalism method based on Event Graphs to depict them. Moreover, the structure gained can be memory space-consuming. Alternatively, this article aims to find a structural optimization algorithm to reduce memory space in polynomial time. Materials and methods:: The presented approach uncovers CPs from the events log of patients admitted to the Emergency Department (ED) and depicts them using Event Graphs methodology. The presented approach defines a structural optimization problem centered on information theory to reduce memory space. Results: The presented approach is evaluated on synthetic events data of patients extracted from MIMIC-3 and uses Neo4j framework and Docker to ensure respectively CPs analysis and method reproducibility. Conclusion:: The experiment demonstrated a high capability of the CPs uncover algorithm to derive patient traces from the events log. The optimal structure minimizes the number of events while maintaining the amount of information provided.