Stable heuristic miner: Applying statistical stability to discover the common patient pathways from location event logs

Abstract

Purpose: The classic heuristic miner algorithm has received lots of attention in the healthcare sector for discovering patient pathways. The extraction of these pathways provides more transparency about patient activities. The previous versions of this algorithm receive an event log and discover several process models by using manually adjustable thresholds. Then, the expert is left with the difficult task of deciding which discovered model can serve as the descriptive reference process model. Such a decision is completely arbitrary and it has been seen as a major structural issue in the literature of process mining. This paper tackles this problem by proposing a new process discovery algorithm to facilitate patient pathways diagnosis.Approach: To address this scientific challenge, this paper proposes to consider the statistical stability phenomenon in an event log, and it introduces the stable heuristic miner algorithm as its contribution. To evaluate the applicability of the proposed algorithm, a case study has been presented to monitor patient pathways in a medical consultation platform.Originality: Thanks to this algorithm, the value of thresholds will be automatically calculated at the statistically stable limits. Hence, instead of several models, only one process model will be discovered. To the best of our knowledge, applying the statistical stability phenomenon in the context of process mining to discover a reference process model from location event logs has not been addressed before.Findings/Practical implications: The results enabled to remove the uncertainty to determine the threshold that represents the common patient pathways and consequently, leaving some room for potential diagnosis of the pathways.

Keywords

• Process mining
• Statistical stability
• Healthcare processes
• Patient pathways
• Indoor localization systems