The task of healthcare research is to analyse healthcare provision and to develop and evaluate new healthcare concepts. The aim is to improve the quality of care and at the same time reduce costs by increasing the efficiency of the healthcare system. Among other things, the focus here is on healthcare processes. However, these are often unknown, so they must first be identified and described.

Process mining is an alternative to the manual recording of care processes. This refers to techniques for automatically extracting, analysing and manipulating processes. These are able to automatically extract corresponding process models that describe the observed behaviour from the event data (event logs) recorded during process execution.

However, as the event logs are flat tables that contain all the events recorded for a process, process mining only provides an overall view of the process. For healthcare research, however, a differentiated analysis of different patient groups (e.g. with different ages or pre-existing conditions) is of particular interest. In addition, it is not necessarily known in advance what the possible differences between the individual patient groups might be. An exploratory analysis of the processes is therefore desirable, allowing the user to analyse and compare different patient groups step by step.

The basic idea behind the multidimensional process mining approach is that the properties of the patient groups, which appear as attributes in the event log, can be interpreted as dimensions that together form a multidimensional data cube. OLAP operations can then be used to define any patient groups using corresponding data extracts, which serve as the data basis for process mining (see illustration). This allows separate models to be created for the different patient groups. These can then be compared and contrasted by the user. However, if the expected differences are not found, the user can adapt the OLAP queries and define other patient groups accordingly, such as a differentiation according to previous illnesses instead of age. This enables the user to look at the care processes step by step from different perspectives and ultimately identify anomalies in them. These can then serve as a starting point for further analyses, for example by creating new research hypotheses.

The aim of the project group is to implement the described approach in a tool for multidimensional process mining in order to demonstrate the technical feasibility of the approach. The system to be developed should represent a vertical prototype that covers all the key concepts of the approach and enables exploratory analyses of supply processes to be carried out. This results in a series of specific tasks to be solved by the PG participants:

• Development of a DWH for the storage of multidimensional event data
• Integration of event logs in multidimensional data structure
• Implementation of exploratory navigation through the event data using OLAP operators
• Provision of process mining algorithms for extracting process models from the multidimensional event data
• Presentation of the analysis results in the form of process models

The project group is suitable for all students (Computing Science, Business Informatics or ESMR) - especially those specialising in ISSE / KISS or the specialisation "IT in Healthcare".

Contact: MSc Thomas Vogelgesang