Loading...
Loading...
AI-powered emergency medicine feedback using ACGME EM Milestones, ABEM, and Ottawa frameworks
lattice template use lattice/emergency-evalThis will create a new workspace using this template configuration.
A non-provisioned evaluation template that configures the AI feedback pipeline for emergency medicine clinical transcripts. Uses a 6-pass analysis architecture aligned with ACGME Emergency Medicine Milestones 2.0, ABEM oral exam format, AAMC EPA framework, Ottawa EM Assessment criteria, and the CORD Teaching Framework.
| Pass | Weight | What it evaluates |
|---|---|---|
| Initial Assessment | 25% | ABCDE approach, primary survey, triage acuity, life threat recognition, resuscitation initiation |
| Diagnostic Workup | 20% | Differential breadth, worst-first thinking, time-sensitive diagnoses (STEMI, stroke, sepsis, PE), imaging/lab interpretation |
| Management & Disposition | 20% | Evidence-based ED management, risk stratification tools (HEART, PERC, Wells, CURB-65), disposition decisions, discharge safety |
| Procedural & Communication | 15% | Team communication, SBAR/I-PASS handoff, consultant communication, de-escalation, breaking bad news (SPIKES) |
| Systems & Safety | 15% | Throughput, multi-patient management, medication safety, bounce-back prevention, documentation timeliness |
| Synthesis | 5% | Overall EM competency, Milestone level mapping, entrustability for independent ED shifts, learning goals |
| Score | Milestone Level | Description |
|---|---|---|
| 9.0+ | Level 5 | Aspirational — ready for unsupervised practice in complex settings |
| 7.5+ | Level 4 | Ready for unsupervised practice |
| 6.0+ | Level 3 | Advancing — manages most patients with indirect supervision |
| 4.0+ | Level 2 | Developing — requires direct supervision |
| 0-4.0 | Level 1 | Entry level |
export LATTICE_URL=https://access.latticeruntime.com
export LATTICE_TOKEN=$(lattice tokens create)
./scripts/setup-evaluation-template.sh
All parameters are mutable and can be overridden per evaluation run:
anthropic or openaiclaude-sonnet-4-5-20250929, gpt-4oThe server must have an AI provider API key set:
ANTHROPIC_API_KEY for Claude modelsOPENAI_API_KEY for GPT models