{"data":{"kind":"file","path":"README.md","version_id":"a0iewx4uhldo4a0srb65hscw","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":3284,"modified_at":"2026-08-28T03:23:09.455000","content_hash":"7a3550db7df7d52577113b9b6524c6ee1275805bcf9d4b6a29d697d4ebe1f211"},"entries":[],"content":"# Hospital Patient Scheduling & Triage Agent\n\nA multi-turn environment that evaluates an AI agent's ability to triage patients, schedule hospital resources, handle emergencies, and communicate effectively in realistic clinical scenarios.\n\n## Overview\n\nThis environment simulates the complex decision-making required of hospital staff who must:\n- **Triage** incoming patients using the Emergency Severity Index (ESI) 5-level system\n- **Schedule** appointments considering provider availability, equipment, labs, and patient preferences\n- **Handle emergencies** by reorganizing existing schedules when urgent cases arrive\n- **Communicate** clearly and empathetically with patients, families, and medical staff\n\n## Scenarios\n\nThe environment includes 10 diverse clinical scenarios:\n\n| Scenario | Type | Key Challenge |\n|----------|------|---------------|\n| Chest Pain | Emergency Triage | ACS recognition, time-critical ESI-2 |\n| Pediatric Fever | Triage | Age-appropriate assessment, ESI-3 |\n| Ankle Fracture | Triage + Scheduling | Orthopedic pathway, pain management |\n| Stroke | Emergency Triage | Time-critical ESI-1, tPA window |\n| Routine Follow-up | Scheduling | Patient preferences, resource matching |\n| Mental Health Crisis | Triage + Safety | Psychiatric emergency, safety protocols |\n| Multi-Patient Scheduling | Complex Scheduling | Constraint satisfaction, priority balancing |\n| Emergency Rescheduling | Surge Management | Mass casualty, resource reallocation |\n| Medication Scheduling | Inpatient Coordination | Critical path analysis, NPO constraints |\n| Discharge Planning | Care Coordination | Multi-disciplinary timeline, family education |\n\n## Evaluation Metrics\n\n| Metric | Weight | What It Measures |\n|--------|--------|-----------------|\n| Triage Accuracy | 25% | Correct ESI level assignment with clinical reasoning |\n| Action Quality | 20% | Completeness and prioritization of immediate actions |\n| Scheduling Quality | 20% | Time-specific planning with constraint awareness |\n| Communication | 10% | Empathy, clarity, and structured patient interaction |\n| Reasoning Depth | 15% | Clinical justification, differential thinking, contingency planning |\n| Constraint Satisfaction | 10% | Respect for stated constraints (time, resources, availability) |\n\n## Usage\n\n```python\nimport verifiers as vf\n\nenv = vf.load_environment(\"hospital-scheduling-agent\")\n# or\nfrom hospital_scheduling_agent import load_environment\nenv = load_environment()\n```\n\n## Design Principles\n\n1. **Realistic Scenarios**: Based on actual clinical presentations and hospital workflows\n2. **Multi-dimensional Evaluation**: No single metric can be gamed for a high score\n3. **Constraint-Rich**: Every scenario has specific resource, time, and personnel constraints\n4. **Emergent Complexity**: Scenarios range from simple triage to complex multi-patient coordination\n5. **Safety-Critical**: Emphasizes the consequences of incorrect triage and scheduling decisions\n\n## Clinical Accuracy\n\n- ESI levels follow the official 5-level Emergency Severity Index framework\n- Time windows reflect real clinical guidelines (e.g., tPA within 4.5 hours, door-to-CT for stroke)\n- Resource constraints mirror typical hospital operations\n- Medication and procedure scheduling follows standard clinical workflows\n","encoding":"utf-8","truncated":false,"total_bytes":3284},"status":null}