The ER nurse's
second opinion,
in real time

TriaVox listens to the patient-nurse intake conversation, analyzes vital signs and medical history, and recommends an evidence-based triage level. Three AI architectures. One clinical decision.

Triage Scale FRENCH / GEMSA
Market Europe (FR, BE, LU, CH)
Regulation EU AI Act Compliant
Triage Concordance with Senior Physicians
90%
TriaVox AI
vs
30%
Standard nursing
How It Works

Three inputs. One recommendation.

TriaVox captures multimodal data from the intake process and synthesizes it through three specialized AI models running in parallel.

01
🎤

Voice Analysis

Real-time NLP processing of the patient-nurse interview. Captures symptoms, onset, severity, and clinical context from natural conversation.

02

Vital Signs

Integrates blood pressure, heart rate, SpO2, temperature, and respiratory rate as entered by the nurse during assessment.

03
📋

Medical History

Structured data from the patient's electronic medical record: allergies, chronic conditions, current medications, prior ED visits.

Why TriaVox

What no one else does

The market has triage tools. None of them listen.

Real-time audio capture

The only triage AI that processes the live patient-nurse conversation. No manual data entry of chief complaints. No information lost in translation.

FRENCH/GEMSA native

Built for the FRENCH triage scale (priority) and GEMSA flow prediction used across French-speaking European emergency departments. Not an American ESI adaptation.

Three AI models in parallel

NLP, LLM, and JEPA architectures cross-validate each other. Each model was published and peer-reviewed independently.

Assists, never replaces

The nurse makes the final call. TriaVox provides a recommendation with confidence levels and reasoning. Full EU AI Act compliance by design.

Scientific Foundation

Peer-reviewed. Award-winning.

Built on years of clinical research at CHU de Lille, published in top medical informatics journals.

JMIR Preprint — AcceptedNew
Artificial Intelligence Models for Predicting Triage in Emergency Departments
7-month retrospective comparative study of NLP, LLM, and JEPA architectures for emergency triage prediction. Evaluates three distinct AI models in a real clinical setting at CHU de Lille.
IEEE/ACM BDCAT 2025
Comparative evaluation of NLP, LLM, and JEPA models for ED triage prediction
Best Paper Award. 7-month retrospective proof-of-concept at Roger Salengro Hospital, Lille.
Am J Emergency Medicine
Navigating the landscape of medical triage: LLMs and beyond
Comprehensive analysis of large language model potential and challenges in clinical triage.
Multicenter (2026)
TIAEU-2: 5-center validation study
Ongoing multicenter study across 5 French emergency departments. TRIADE randomized trial (PHRC-I) in planning.
Active Research

EIMLIA

Health Economics Evaluation of AI-Assisted Emergency Triage. A medico-economic study evaluating the organizational and financial impact of deploying AI triage systems in emergency departments.

Economic

Cost-Effectiveness

Cost-effectiveness and cost-utility analysis of AI-assisted triage vs. standard nursing triage, measured in cost per QALY gained and cost per patient managed.

Organizational

ED Workflow Impact

Discrete Event Simulation combined with Process Mining to model the impact of AI triage on patient flow, waiting times, and resource utilization.

Clinical

Patient Outcomes

Analysis of clinical outcomes including triage accuracy, undertriage rates, length of stay, and downstream effects on hospital admissions and mortality.

Cost-per-Patient
Length of Stay
Waiting Times
Resource Utilization
Undertriage Rate
QALY
Staff Workload
Bed Occupancy

Emergency triage hasn't changed in decades. That ends now.

TriaVox is building the standard for AI-assisted emergency triage in Europe. One conversation at a time.