Pre-consult interview

The patient scans a QR at the consulting-room door. The interview opens on their own phone, and the doctor reads the answers before the patient walks in.

Status: Live at the Pinamar installation.

The interview asks in plain language and follows up on what the patient actually said.

How it works

  • The door carries the entry point. A printed poster with a QR code per consulting room — no app, no account, no password.
  • The narrative comes first. The patient describes the problem in their own words, typed or dictated. Structured questions come after, not before.
  • The AI asks the follow-ups a clinician would ask about that specific complaint, then stops. It does not run a fixed questionnaire past someone who already answered.
  • The result is a triage colour and a summary, waiting in the doctor’s list for that appointment.

Where the model is not trusted

Red-flag rules are deterministic code, not a prompt. Certain answers raise the triage level no matter what the model concluded. The model may raise a patient’s urgency; it may never lower it below what the rules set.

The doctor sees the raw answers next to the summary, so a wrong summary is visibly wrong rather than quietly wrong.

Editing what it asks

The three prompts behind the interview — the questioner, the triage judgment and the summary — are editable from the staff panel. The canonical text is English; staff edit the Spanish.

Next: Referral orders read by AI