AI-Assisted Triage Protocol for Medium-Complexity Clinics

Develop a complete AI-assisted triage protocol for [CLINIC NAME], specialty [SPECIALTY], with [NUMBER] patient visits/day across [NUMBER] locations.

**Regulatory Context (required understanding before implementation):**
- CFM Resolution 2.299/2021: AI use in medicine — AI assists, the physician decides
- LGPD: health data is sensitive data — requires explicit consent
- State CRM boards: verify local resolutions on digital triage systems
- Accountability: the physician remains responsible — AI does not replace clinical judgment

**Phase 1 — Triage Flow Definition:**

**Patient Entry Channel:**
- [ ] Proprietary app or partner platform (Doctoralia, iClinic, Metramed)
- [ ] WhatsApp Business with triage bot
- [ ] Digital kiosk at reception
- [ ] Web pre-consultation form
- [ ] All of the above with centralized data

**Triage Questions (customize by specialty):**

Block 1 — Identification (no sensitive data yet):
- Full name, date of birth, insurance/member ID
- Primary reason for visit (dropdown + open text field)
- Date of last visit to the clinic

Block 2 — Warning Signs (urgency flags):
- 'Are you currently experiencing any of these symptoms?' (list of critical symptoms for [SPECIALTY])
- 'How long have you had this?' (< 24h = possible urgency; > 7 days = elective)
- 'Have you been to an emergency room for this?' (yes = priority)
- 'Are you taking any continuous-use medications?' (list)
- 'Do you have any of these conditions?' (relevant comorbidities for [SPECIALTY])

Block 3 — Visit Context:
- Follow-up, first visit, or test results review?
- Do you have tests to show the doctor? (optional upload)
- Any mobility restrictions or special needs?

**Phase 2 — Priority Classification Algorithm:**

**Green Level (elective — wait in order of arrival):**
- Routine visit, check-up follow-up, prescription renewal without urgency
- No warning signs identified
- Service: standard appointment sequence

**Yellow Level (semi-urgent — priority on schedule):**
- Symptom with 24-72h progression
- Relevant comorbidity + new symptom
- Vulnerable patient (>70 years, pregnant, immunocompromised)
- Service: among the first 3 in queue when they arrive

**Orange Level (urgent — immediate slot):**
- 2+ simultaneous warning signs
- Symptom less than 24h with progression
- Service: immediate slot, physician notification via app

**Red Level (emergency — ER referral):**
- Immediate danger signs (chest pain + sweating, decreased consciousness, severe dyspnea, active bleeding)
- System does not treat — directs patient to nearest emergency room
- Automated notification to receptionist + WhatsApp message to patient

**Phase 3 — Technical Integration:**

**Implementation Options by Technical Maturity Level:**

Basic (cost < R$ 500/month):
- Google Forms + spreadsheet with conditional formatting + email notifications
- Typeform with conditional logic
- WhatsApp bot via ManyChat or Twilio

Intermediate (R$ 500-2,000/month):
- Digital anamnesis platforms: Anamnese.io, ProntoMed, Tria Saude
- Integration with existing EHR (Electronic Health Record)

Advanced (R$ 2,000+/month or custom development):
- LLM to interpret free-text complaints and classify
- HL7/FHIR integration for interoperability
- Real-time dashboard for nursing coordinator

**Phase 4 — Team Training:**

- Receptionists: how to handle patients who refuse to fill out the form
- Nursing technicians: how to use the triage level in the physical flow
- Physicians: how triage information arrives and what to do with it
- Managers: monitoring dashboard of triages by level throughout the day

**Phase 5 — Governance and Continuous Improvement:**

- Monthly audits: compare assigned triage level vs. actual level confirmed by physician
- Algorithm review criteria: if >10% of yellow cases are reclassified by the physician
- Technical responsible for the system: formally designated physician or nurse
- LGPD consent: collected before form begins, with clear purpose stated
- Success metrics: average wait time reduction, % of urgent cases reaching the physician in time, patient satisfaction (post-visit NPS)