Learning the context of a clinical process

Abstract

Clinical guidelines provide recommendations to assist clinicians in making decisions regarding appropriate medical care for specific patient situations. However, characterizing these situations is difficult as it requires taking into account all the variations that patients may present. We propose an approach which helps with identifying and categorizing the contexts that need to be taken into account within a clinical process. Our methodology is based on a formal process model and on a collection of process execution instances. We apply machine-learning algorithms to group process instances by similarity of their paths and outcomes and derive the contextual properties of each group. We illustrate the application of our methodology to a urinary tract infection management process. Our approach yields promising results with high accuracy for some of the context groups that were identified. © 2010 Springer-Verlag.