← Back to projects
Ongoing - Prospective simulated study

HARMONIA

Humans (Call Dispatchers + Physicians) vs AI in Emergency Medical Dispatch, One-to-one Impact Analysis

Project Description

HARMONIA is a prospective cohort study with continuous enrollment, evaluating the performance of a GPT model compared to Medical Regulation Assistants (ARM) and dispatch physicians at the SAS-Centre 15 of Lille University Hospital. The study uses 60 simulated call scenarios, managed by ARM–physician pairs and simultaneously analyzed by AI. The goal is to determine whether AI can match or exceed human performance in telephone triage using the ARM triage scale.

Design

Prospective cohort with continuous enrollment, double-blind and intra-protocol randomization

Population

60 simulated scenarios
13 dispatch physicians
10 volunteer ARMs

Period

January - March 2026
(3-month enrollment)

Center

SAS-Centre 15
Lille University Hospital

Objectives

Primary: Compare GPT algorithm performance with actual ARM triage at the SAS-Centre 15 of CHU Lille against the gold-standard ARM triage scale.

Secondary:

  • Concordance with medical regulation decision (final regulation decision)
  • Participant experience analysis (ARMs + physicians) via NASA-TLX questionnaire
  • Reliability and speed of GPT algorithm transcription
  • Reliability of semantic analysis and predicted resource deployment by GPT

Methodology

Each scenario is managed by an ARM–physician pair, formed interchangeably from a volunteer pool. Matching follows a double-blind scheme with intra-protocol randomization. Each volunteer is exposed to all simulated cases, ensuring a complete crossover evaluation.

The gold standard is composite: strict application of the ARM triage scale to call data and expert opinion (panel of 3 blinded physicians).

Statistical Analyses

Primary Metrics

MAE, RMSE, weighted Kappa, exact agreement and ±1 class, AUC-ROC

Composite Score

F1 micro/macro, Spearman correlation, multiclass Brier score

Secondary Analyses

Bland-Altman plots, confusion matrices, reliability diagrams

Tools

R® (v4.2.2+)
α = 5% threshold, Bonferroni correction

Ethics and Data

The study does not modify patient care in any way. It relies exclusively on simulated data (anonymized by default, AES-256 encrypted). AI models do not store or archive audio recordings. Signed consent from participating ARMs and physicians. Analysis conducted on the secured CHU Lille platform (SNDS security framework).

Timeline

December 2025

Protocol writing, scientific council validation, DPO declaration

January 2026

Start of enrollment

March 2026

End of enrollment

April 2026

End of data extraction

June 2026

Analysis and results validation

October 2026

Final report

2026-2027

Peer-reviewed journal publication

Project Team

André Filipe Gomes Botelho

Principal Investigator
Emergency Medicine Resident
Lille University Hospital

Dr Flavie Vanbrugge

Thesis Director
Head of SAMU Unit
Lille University Hospital

Dr Edouard Lansiaux

Scientific Committee
Junior Doctor, Emergency Medicine
AI Engineer
Lille University Hospital

Dr Jonathan Hennache

Scientific Committee
SAMU Unit
Lille University Hospital

Dr Roch Joly

Scientific Committee
Head of Emergency Department
Lille University Hospital