Discovering congestion dynamics models in clinical pathways using background knowledge

Clinical Pathways (CPs) consist of structured multidisciplinary guidelines and protocols used to model steps of clinical treatments. The main objective of applying CPs is that of optimizing both outcomes and efficiency — however, the actual implementation of CPs can be complex and result in important deviations and unexpected inefficiencies. In this paper, we develop an approach to identifying and understanding such problems by leveraging process mining techniques and background knowledge. We design specific data structures aimed at properly capturing the data produced during the implementation of CPs, including the treatment of more than one disease for a single patient. We then provide a methodology to discover and characterize congestion dynamics in CPs. Since the resulting process discovery problem is theoretically intractable, we develop heuristic algorithms that, based on an extensive experimental assessment, prove capable of discovering meaningful knowledge with a reasonable computational effort.