Our foundational research consisted of in-depth interviews with four medical professionals from Mount Sinai, including two nurses, one resident, and one physician. The goal of these conversations was to map the current ER triage workflow, identify specific pain points, and understand the real-world context in which triage decisions are made. These firsthand accounts provided invaluable qualitative insights into the daily pressures and systemic challenges of the ER, directly shaping our understanding of the problem.

To ground our project in established evidence, we conducted a comprehensive literature review of academic work on triage misses, the application of AI in emergency medicine, and the human factors that define the patient and staff experience. This review helped us quantify the scale of the problem and understand the clinical impact of issues like triage accuracy and wait times. It also validated that a core driver of negative patient experience is the uncertainty and duration of their wait.

We performed a focused competitive analysis of existing clinical decision support tools, including major EHR systems like Epic and specialized AI platforms like Aidoc and Viz.ai. The analysis revealed a critical market gap: most tools only activate after a patient has had a scan, leaving the pre-imaging waiting period as a clinical blind spot. This insight was pivotal, allowing us to position Aria.AI as a solution that continuously monitors and re-scores patient risk from the moment they arrive.

To broaden our perspective beyond clinical staff and overcome limited access to real patients, we utilized custom GPT agents to conduct synthetic interviews with AI-generated personas for patients and caregivers. This innovative method allowed us to explore the experiences of vulnerable populations, such as non-English speakers or caregivers for elderly patients. These interviews surfaced powerful emotional insights, revealing how "quiet patients" are often at the greatest risk of being overlooked in a chaotic ER and how crucial caregiver advocacy is.

Lynn Carter is an experienced ER triage nurse at Mount Sinai who is responsible for making rapid patient assessments in a high-pressure, chaotic environment. Her primary goal is to accurately prioritize the sickest patients first, but she is often hindered by incomplete patient information, constant interruptions, and the sheer volume of cases. She represents the expert user who needs a tool to reduce cognitive load and surface critical signals that are easy to miss.

Henry Mason is a 54-year-old man who comes to the ER with mild chest pain but has a history of downplaying his symptoms. He represents a high-risk patient archetype: one whose seemingly minor complaint could be masking a life-threatening condition. Because he is alone and tends to understate his condition, he is in danger of being under-triaged and left to deteriorate in the waiting room without being noticed.