Urgent and emergency care generates large volumes of clinical data, yet much of the information most useful for research, surveillance, and service planning is recorded only in free-text clinical notes and never reaches the structured datasets used for analysis. As a result, the reasons patients present, the upstream pressures driving demand, and the care that was needed but unavailable remain largely invisible, with consequences that fall most heavily on people who move repeatedly across services. Existing initiatives connect previously separate records or predict future events, but neither recovers the clinical reasoning that explains why people reach crisis. In this Perspective we propose a three-layer knowledge graph architecture for urgent and emergency care: a first layer linking individual contacts into connected patient trajectories across services and time; a second recovering clinical detail lost to reductive coding; and a third capturing the reasoning behind each presentation, including barriers to care and intended but unavailable services. We describe why recent advances in natural language processing and artificial intelligence make this architecture achievable, assess the current evidence - strong for each component, though the integrated architecture awaits validation - and set out the validation, governance, and infrastructure that implementation would require. The architecture offers a route from data volume to usable knowledge, and a means of making patterns of unmet need legible to those who plan and govern care. The constituent technologies now exist; whether this information remains invisible is increasingly a question of design rather than capability.
