A visual analytics framework for analysis of patient trajectories
The problem of analyzing patient trajectories is fundamental to our ability to understand and characterize diseases and how we treat them in our hospitals, and to devise and explore effective alternative strategies for healthcare. In this paper, we present a new approach to analyze hospital patient trajectories. Based on visual analytics, our approach is aimed at aiding the domain scientist (in this case, a hospital bioinformatician or a data analyst) to visually navigate and analyze patient health trajectories in a scalable manner. More specifically, we view the problem as one of structure discovery and tracking how such structure evolves with time over the course of patients’ stay at the hospital(s). An ability to scalably track and view the temporal progression of context variables associated with patients in conjunction with health indicator variables could provide vital clues on how practices affect outcomes. Furthermore, by enabling compact and consolidated views of complex patient trajectories, our approach can help to delineate subpopulations (i.e., subgroups of patients) that show divergent behavior. As a concrete case study in application and evaluation, we present results and initial findings on a large patient data set obtained from the Duke Antimicrobial Stewardship Outreach Network (DASON) database, with an aim of extracting factors relevant to antibiotic usage and stewardship in hospitals.
