Preview
CHAI Layout
Applied Model Card · CHAI schema v0.2 · Draft v1

Aster Triage

Aster Health Ltd. · Dev · v123

Uses and Directions

  • SummaryAI-assisted triage tool that helps emergency nurses prioritize incoming patients. Intended use: Prioritize incoming emergency department patients by acuity at triage. Website: https://example.comDeveloper Reported
  • Intended use and workflowUse: Prioritize incoming emergency department patients by acuity at triage. Settings: Emergency department, Inpatient Workflow: Runs when a triage assessment is saved in the EHR and shows a suggested acuity level. User action: Review the suggested acuity and confirm or change it before assigning a bed.Developer Reported
  • Primary intended usersNurses, PhysiciansDeveloper Reported
  • Target populationAge: 18 and older Group: Adult Conditions: Adults presenting to the emergency department with undifferentiated complaints. Geography: US only, English-language documentation only Not evaluated: Pregnant patients and patients with limited English proficiency.Developer Reported
  • Out-of-scope usePediatric emergency care and prehospital settings.Developer Reported
  • Known risks and limitationsLimitations: Less accurate when the triage note is very short or missing. Failure modes: Under-triages atypical chest pain presentations. Do not rely when: The patient is unstable or vitals are missing. Safeguards: Nurse confirmation is required before any acuity is assigned. Human review: AlwaysDeveloper Reported
  • Known biasesGroups: Age, Sex, Race / ethnicity Findings: Agreement stayed within 2 points of the overall rate in every subgroup.Developer Reported
  • Populations with uncertain evidencePatients over 90 had too few cases to report.Developer Reported

AI System Facts

  • Inputs / outputsInputs: Demographics, Vitals, Clinical notes Outputs: Risk score, RecommendationDeveloper Reported
  • EHR compatibilityEHR systems: Epic, Oracle Health (Cerner) Interfaces: FHIR R4 API, SMART on FHIR appDeveloper Reported
  • Security and privacyIn transit: Yes Certifications: SOC 2 Type 2, ISO 27001, HIPAA security risk assessmentDeveloper Reported
  • Regulatory approvalPathway: FDA 510(k) Clearance: Cleared / authorizedDeveloper Reported
  • Funding sourceSelf-fundedDeveloper Reported

Key Metrics

  • Usefulness / Efficacy

    • GoalShow the suggested acuity agrees with nurse-assigned acuity.
    • ResultAgreement: 91.2% (95% CI 89.5-92.8) Under-triage rate: 2.1%
    • InterpretationAgreement is comparable to nurse-to-nurse agreement.
    • Test typeRetrospective, External
    • Validation42,000 adult emergency visits held out from model development.

  • Fairness

    • GoalAge, Sex, Race / ethnicity
    • ResultAgreement by subgroup, difference from overall
    • InterpretationAgreement stayed within 2 points of the overall rate in every subgroup.
    • Test typeSubgroup analysis
    • ValidationThe same 42,000 held-out visits.

  • Safety

    • GoalConfirm that unsafe under-triage is rare.
    • ResultCritical under-triage: 0.4%
    • InterpretationRate is within the range accepted by the clinical safety board.
    • Test typeRetrospective chart review
    • Validation1,200 cases reviewed by two emergency physicians.