LMbox
Vos données restent chez vous. Book a demo
← All agents
Healthcare · HDS pilot v0.1.0

Triage Patient HDS

Pre-triage of inbound patient requests (mail, web form) before the medical secretariat. 4-tier urgency classification, routes to the right specialty, recommends the care pathway. Stays on the HDS-compliant LMbox — 0 PII leaves the building.

Measured gain
28 min
par demande patient

On 200 patient mails / day / facility: 28 minutes saved per request, ~93 hours freed per week. At a typical medical-secretariat rate: ~€250k/year of non-clinical workload reclaimed for care coordination and physical reception.

The problem

The medical secretariat spends its day saying 'I'll call your doctor'.

Today

Pré-tri par secrétariat

30 min par mail patient, urgences vitales noyées dans la routine

30 min / demande patient
100-300 demandes / jour
Risque vital manqué douleur thoracique classée routine, traitée le lendemain
  • · 70 % du temps consommé par du routage trivial
  • · Urgence vitale glissée dans un mail à 17 h, traitée le lendemain
  • · Aucune traçabilité opposable au tribunal administratif
  • · Outils cloud → impossible (donnée patient interdite)
Le risque finit en procédure
With Triage Patient HDS

Pré-tri assisté HDS

Sur l'appliance HDS, 0 donnée patient vers le cloud

2 min / demande patient
SAMU 15 auto sur signal d'alerte
~93 h récupérées par semaine et par établissement
  • Données personnelles supprimées avant tout traitement aval
  • Détection immédiate des signaux d'alerte (AVC, idéation suicidaire…)
  • Protocoles internes validés par le médecin chef de pôle
  • Tout hashé dans la chaîne d'audit, opposable au patient et au juge
Pas de diagnostic — réserve médicale
How it works

Four steps, from incoming document to human decision.

  1. 1

    Inbound patient request

    Mail, web form or GP electronic referral. The agent reformulates in neutral terms and strips PII (national ID, name, address, phone) before any downstream processing.

  2. 2

    Red-flag scan first

    Before any tool call, the agent scans for vital-emergency symptoms (constrictive chest pain, sudden paralysis, respiratory distress, active suicidal ideation). On match → SAMU 15 + mandatory human handoff, specialty classification skipped.

  3. 3

    Internal protocol + available specialty

    Agent calls `lookup_protocole_triage` (protocols validated by the medical chief of staff) then `check_specialites_disponibles`. Every `protocol_id` and `specialty_id` in the output is cryptographically tied to a tool call from this turn — no invented protocol.

  4. 4

    Structured output + secretariat handoff

    Validated JSON: urgency level, recommended specialty, care pathway (in-person / teleconsultation / emergency / SAMU 15), patient next steps, warnings. Hashed into the audit chain — a doctor can prove to a patient what decision was made and why.

Architecture

Tools, connectors, deployment.

Tools (function-calling)

4
  • lookup_protocole_triage
  • check_specialites_disponibles
  • check_patient_existant
  • escalate_humain

Optional connectors

3
  • hds-protocoles
  • hds-dpi
  • hds-secretariat

Each connector activates based on the customer's subscription.

Deployment via the SDK
$ lmbox agent deploy ./triage-patient \
    --box BOX-XXX \
    --token "$LMBOX_BOX_API_KEY" \
    --api https://api.lmbox.eu

LMbox guarantees across the catalogue

Data stays with you

Model and data stay on the customer's LMbox appliance. No patient, contract or invoice data is ever sent to an external cloud.

Audit chain

Every tool call, every agent output is timestamped, hashed and admissible before the regulator (ACPR, ANSM, CNIL, EBA).

Human decision

The agent recommends, the human decides. No auto-signature, no auto-payment: final responsibility stays with the business.

Try Triage Patient HDS on the public demo.

One-click sign-in. You see the agent installed on a real LMbox, with its system prompt loaded and audit chain live.