Summarizing patient daily activities for clinical pathway mining

Abstract

Clinical Pathway is ubiquitous and plays an essential role in clinical workflow management. The combination of topic modeling and process mining is an efficient approach to get a non-static and topic-based process model. Topic modeling is used to group the activities of each clinical day into the latent topics, and process mining is used to generate a concise workflow model based on these topics. However, because of the specificity of clinical data, it usually suffers from the performance of topic modeling. In this paper, we take an important clinical practice, all the same activities in one clinical day tend to represent the same clinical goal, into account to enhance the effectiveness of topic modeling. The experiments on real data show significant performance gains of our approach.