COMPASS

Abstract

Clinical decision-making related to hospital stays requires clinicians to synthesize medical, functional, and social evidence scattered across lengthy, unstructured clinical narratives produced by multiple clinical teams, including physicians, nurses, therapists, and social workers. Although automated text summarization methods can condense clinical records, they often provide limited support for tracing evidence over time, relating contributions across disciplines, or verifying summarized claims against source documentation. We present a visual analysis system, COMPASS — Clinical Observation Multidisciplinary Progression Analysis Support System, for exploring clinical event sequences from narrative clinical data related to patients hospitalized for falls. A central contribution is a phase-based representation that models clinical care as a multi-stage, multidisciplinary process rather than a single endpoint or status. The system integrates a multi-team timeline representation based on impressionist art renderings, event-based summary views, and daily evidence views to support temporal reconstruction, coordination analysis, and evidence-grounded review. COMPASS was developed through an activity-centered design process with domain collaborators and is illustrated through a representative case study and expert qualitative feedback. These results suggest that COMPASS can help medical teams reconstruct patient trajectories, identify multidisciplinary documentation patterns, and examine medical, functional, and social evidence within its temporal context. COMPASS supports the evidence-grounded visual analysis of how clinical events and multidisciplinary documentation evolve over time.