Discovery of patient pathways from a national hospital database using process mining and integer linear programming

Abstract

The analysis of patient pathways from event log is gaining importance in the field of medical information. It provides deep insights about the care process and the ways to improve it. This paper combines optimization and process mining. A new Integer Linear Programming model is proposed to discover the care process at a macroscopic scale from a large-size database. When dealing with health-care data, the main challenge to overcome is the considerable variability of patients' behaviors. An original size constraint and an aggregation method are used to create simple but significant process models. The results of a case study on heart failures confirm the ability of the approach to reveal the process information behind the data.