AI Emergency Department Triage Agent

How Cassidy automates this using AI

Step 1: Trigger on patient arrival

The Workflow activates when a new patient registers or completes digital pre-arrival intake, capturing chief complaint, vitals, pain score, and mental status assessment.

Step 2: Ingest and normalize clinical data

Cassidy pulls relevant EHR signals—demographics, allergies, medications, recent labs, prior ED visits, and comorbidities—via FHIR integration, while NLP structures free-text triage notes into actionable data.

Step 3: Run risk stratification and acuity prediction

The Agent analyzes all inputs to recommend an acuity level with probability scores, flags undertriage risk, and runs condition-specific screens for sepsis, stroke, STEMI, and other critical presentations.

Step 4: Generate routing and intervention recommendations

Based on risk scores and predicted resource needs, Cassidy suggests split-flow routing (resuscitation bay, main ED, or Fast Track), standing orders, and bundle activations—with transparent rationale for each recommendation.

Step 5: Human-in-the-loop review and action

The triage nurse reviews AI recommendations with full visibility into contributing factors, approves or overrides suggestions, and initiates appropriate pathways—all logged for audit and continuous model improvement.