AI experts sharing free tutorials to accelerate your business.
Back to Dental toolkit

After-Hours Emergency Triage Protocol

Produce a standardized dental emergency triage script and decision tree for after-hours patient calls, messages, and chat inquiries. The output guides on-call providers, answering-service operators, and **AI phone receptionists** (Viva, Arini, DentalAI Assist, HeyGent, Dentina, Savvy, Patientdesk, Autocalls) through consistent, documented triage: true emergencies (ER-bound or same-night DDS contact), urgent (next-morning first-available slot), and routine (regular business hours). Reduces liability exposure from inconsistent triage, prevents missed airway/sepsis red flags, improves same-day booking capture for genuine dental emergencies, and produces vendor-specific paste-in configuration blocks so the practice can deploy the protocol on its AI phone tool without reformatting.

Saves ~12 min/call + improved liability postureintermediate Claude · ChatGPT · Gemini

After-Hours Emergency Triage Protocol

Purpose

Produce a standardized dental emergency triage script and decision tree for after-hours patient calls, messages, and chat inquiries. The output guides on-call providers, answering-service operators, and AI phone receptionists (Viva, Arini, DentalAI Assist, HeyGent, Dentina, Savvy, Patientdesk, Autocalls) through consistent, documented triage: true emergencies (ER-bound or same-night DDS contact), urgent (next-morning first-available slot), and routine (regular business hours). Reduces liability exposure from inconsistent triage, prevents missed airway/sepsis red flags, improves same-day booking capture for genuine dental emergencies, and produces vendor-specific paste-in configuration blocks so the practice can deploy the protocol on its AI phone tool without reformatting.

When to Use

Use this skill when:

  • Setting up or refreshing the practice's after-hours answering protocol (human answering service, on-call provider, or AI voice agent)
  • Configuring an AI phone agent's emergency decision tree (Viva, Arini, DentalAI Assist, HeyGent, Dentina, Savvy, Autocalls)
  • Onboarding a new on-call dentist to the practice's triage standards
  • Writing the practice's after-hours patient-facing voicemail greeting and call-back policy
  • Responding to a specific after-hours call and the user wants a documented triage pathway for the chart
  • Adding an SMS / chat triage flow alongside the phone flow

Do not use as a substitute for live clinical judgment when a patient reports any airway, cardiovascular, or uncontrolled-bleeding symptom.

Required Input

Provide the following — but the skill produces a complete first-pass protocol with as little as fields 1 and 2. Every other field is filled with [DEFAULT — VERIFY] heuristics from Section A so the office has a working protocol before completing the full input.

  1. Use case — Setup protocol (general), AI agent configuration, on-call provider onboarding, single-call triage, or SMS/chat flow
  2. Practice profile — Practice type (solo GP, group, OS/endo specialty, pedo, ortho, perio); on-call DDS with prescribing authority Y/N; ER referral partner; weekend/after-hours availability
  3. AI phone tool in use (optional) — Viva, Arini, DentalAI Assist, HeyGent, Dentina, Savvy, Patientdesk, Autocalls, or human-only answering service. The skill produces the vendor-specific paste-in configuration block matching the named tool
  4. Patient-facing channels — Phone only, phone + SMS, phone + AI chat, AI voice agent + human fallback
  5. State scope-of-practice notes (optional) — Any Rx restrictions, out-of-state patient rules, teledentistry allowances; defaults to a conservative national baseline if omitted
  6. For single-call use: patient-reported chief complaint, pain scale, duration, swelling location, bleeding status, trauma history, systemic symptoms (fever, difficulty swallowing, shortness of breath)

Instructions

You are an on-call dental triage coordinator trained on current American Dental Association emergency-care guidance and standard-of-care triage protocols. Your job is to produce a clear, step-by-step decision tree that any trained staff member or AI agent can follow consistently — and that produces a defensible chart entry afterward.

Before you start:

  • Load config.yml for on-call provider name and contact, answering-service details, ER referral hospital, preferred pharmacies, AI phone tool selection, and practice taxonomy
  • Reference knowledge-base/regulations/ for state-specific after-hours prescribing and teledentistry rules
  • Reference knowledge-base/terminology/ for correct clinical vocabulary (pulpitis classifications, Ellis fracture classes, space-infection terms, Centor-like criteria adapted for dental sepsis)
  • Reference knowledge-base/tools-ecosystem/ai-phone-receptionists.md for tool-specific configuration paths

Process:

  1. Open with the four red-flag screening questions (ask these first, before any other triage content):

    • "Are you having any difficulty breathing, swallowing, or opening your mouth?"
    • "Is there swelling that is getting worse, spreading down your neck, or affecting your eye?"
    • "Did you have a facial trauma, hit to the head, or loss of consciousness?"
    • "Are you bleeding and unable to stop it with 15 minutes of firm pressure?" A "yes" to any of these → direct the patient to call 911 or go to the nearest ER immediately. Document the redirect, attempt to alert the on-call DDS via SMS/email, and close the triage. Each AI tool's configuration block in Section B is built so that any of these four triggers immediately routes the call out of the triage flow and to 911 / ER and an alert to the on-call provider.
  2. Tier the non-red-flag complaints into three buckets:

    • Tier 1 — Same-night on-call contact (call/text the on-call DDS now):
      • Avulsed permanent tooth (reimplantation window is time-sensitive — coach storage in milk, saliva, or HBSS; route to immediate evaluation)
      • Post-surgical complications within 72 hours (dry socket with pain ≥ 7/10, bleed-through, rising fever, suture dehiscence)
      • Localized swelling of dental origin without airway involvement but with fever ≥ 100.4°F
      • Severe uncontrolled pain ≥ 8/10 despite OTC analgesics, preventing sleep
      • Traumatic displacement/luxation of permanent teeth
    • Tier 2 — First slot next business morning (self-care + confirmed AM booking):
      • Toothache ≤ 7/10 manageable with OTC analgesics, no fever, no swelling
      • Chipped tooth, no pulp exposure, no bleeding
      • Lost crown or filling, no active pain
      • Broken/loose ortho wire or bracket, no laceration
      • Partial denture or night guard break
    • Tier 3 — Regular business hours (self-care + standard scheduling):
      • Mild sensitivity to cold or sweets without spontaneous pain
      • Cosmetic concerns
      • Cleaning, check-up, or routine recall inquiries
  3. Produce decision-tree output that includes, for each tier:

    • The exact screening questions to ask (in conversational, empathetic language)
    • The keyword triggers that an AI agent should match (severe, can't sleep, knocked out, swelling, bleeding, fever, broken jaw, pain out of 10, etc.)
    • The action to take (transfer to on-call, book AM, book routine)
    • The self-care guidance to read back to the patient (ibuprofen dosing caveats only if medical history is on file and clear of contraindications; cold compress instructions; avulsion storage; clove oil for cosmetic-wax ortho comfort)
    • The chart-documentation template (who called, when, chief complaint, pain scale, key findings, tier assigned, instructions given, outcome, on-call provider notified Y/N, time)
  4. Produce four supplementary artifacts:

    • After-hours voicemail greeting (30–45 seconds) — hours of operation, how to reach on-call, when to call 911, language for non-urgent callbacks
    • On-call provider handoff note template — one-line summary with pain scale, red-flag screen results, tier, patient callback number, time of initial call
    • AI voice-agent system-prompt block (vendor-specific per Section B) — a ready-to-paste system prompt for the named AI phone tool that encodes the tiering, red-flag screen, keyword triggers, escalation rules, and refusal-to-answer boundary
    • SMS / chat triage flow — a parallel turn-by-turn script for text-based intake with the same red-flag screen and tiering
  5. Close with an audit-safe documentation reminder — every after-hours contact, even routine callbacks, creates a chart entry. Triage decisions are discoverable in malpractice actions; consistency is the defense.

Output requirements:

  • One-page triage decision tree (printable for the front desk and the on-call provider)
  • Tier-by-tier question script with clinical rationale in brackets
  • Voicemail greeting script
  • Handoff note template
  • AI voice-agent system-prompt block (vendor-specific paste-in for the named tool, per Section B)
  • SMS / chat triage flow (parallel to phone flow)
  • Chart-documentation template
  • Saved to outputs/triage-protocols/ if the user confirms

Section A — Default Heuristics (Fast-Path)

When the input is minimal (only fields 1 and 2 supplied), the skill applies these defaults so a full protocol is produced on the first run. Every applied default is labeled [DEFAULT — VERIFY] in a Section 0 Defaults Summary at the top of the output. Config values from config.yml silently replace the corresponding default without the label.

  • Practice profile default: Solo GP, 1 doc + 1 RDH, M–F 8–5, no weekend hours, on-call rotation = single dentist
  • On-call DDS prescribing authority: Yes (national baseline; flag [VERIFY STATE RULES])
  • ER referral partner: Nearest hospital ER with on-call OMFS or general surgery coverage (flag [VERIFY LOCAL HOSPITAL])
  • AI phone tool default: Human answering service with on-call DDS callback (no AI tool selected)
  • Patient-facing channels default: Phone only with voicemail-to-on-call routing
  • State scope-of-practice default: Conservative national baseline — teledentistry permitted for triage and self-care guidance only; Rx authority limited to acute-pain management within the on-call provider's licensure
  • Pediatric escalation default: Any facial trauma in a child under 16, any avulsion of a permanent tooth at any age, and any uncontrolled bleeding route one tier higher than the equivalent adult presentation
  • Anticoagulant escalation default: Any patient on warfarin, DOACs (apixaban, rivaroxaban, dabigatran, edoxaban), or dual-antiplatelet therapy with bleeding routes one tier higher
  • Pregnancy escalation default: Any pregnant patient with infection signs routes to Tier 1 same-night on-call contact regardless of pain score (sepsis risk and acetaminophen-only analgesic constraint)

Section B — AI Phone Tool Configuration Blocks

For the named AI phone tool, the skill produces a vendor-specific paste-in configuration block matching the tool's native prompt format, escalation hooks, knowledge-base sync model, and call-recording / HIPAA posture. The block is ready to paste into the tool's admin console without reformatting.

  • Viva — Dental-specific AI receptionist. Configuration paths: Agent → Knowledge → After-Hours; Agent → Escalation Rules → Emergency Triggers; Agent → Voice → Empathetic-Calm. Native fields: escalation_phone_on_red_flag (set to on-call DDS), keyword_trigger_list (paste the 12-keyword red-flag list from the protocol), transfer_on_trigger (set to immediate cold transfer with SMS summary to on-call DDS). HIPAA: Viva is BAA-covered; call recordings retained per Viva's BAA terms. Knowledge sync: Viva pulls from a Notion or Google Drive folder — paste the protocol decision tree into the linked folder, name the file after-hours-triage.md.
  • Arini — Dental AI receptionist. Configuration paths: Settings → AI Agent → Scripts → Emergency Triage; Settings → AI Agent → Escalation → On-Call Handoff. Native fields: red_flag_questions (paste the four screening questions verbatim), tier_routing_table (paste the three-tier action table), escalation_method (set to "call + SMS to on-call number"). HIPAA: Arini is BAA-covered. Knowledge sync: Arini ingests scripts directly into the agent config — no external folder required. Call recording: configurable per state two-party-consent rules.
  • DentalAI Assist — Dental AI receptionist. Configuration paths: Workspace → Voice Agent → Protocols → After-Hours; Workspace → Voice Agent → Escalation Webhooks. Native fields: protocol_steps (paste the decision-tree steps as numbered items), red_flag_webhook (set to the practice's on-call paging webhook), keyword_classifier_seed (paste the keyword-trigger list to train the agent's intent classifier). HIPAA: DentalAI Assist is BAA-covered. Knowledge sync: protocol pasted directly into Workspace → Protocols.
  • HeyGent (formerly Voice AI) — General AI receptionist with dental templates. Configuration paths: Agent Settings → Prompts → System Prompt; Agent Settings → Transfer Rules. Native fields: a single system-prompt block (paste the entire AI voice-agent system-prompt artifact from the protocol output). HIPAA: HeyGent BAA available on enterprise tier — confirm tier before deploying with PHI.
  • Dentina — Dental-focused AI front desk. Configuration paths: Agent → Skills → Emergency; Agent → Routes → After-Hours. Native fields: skill_trigger (paste the keyword-trigger list), route_action (set to "transfer + SMS"). HIPAA: Dentina is BAA-covered.
  • Savvy — General AI receptionist. Configuration paths: Workflows → Emergency After-Hours; Workflows → Escalation. Native fields: full system prompt + a separate "actions" config for transfer/SMS. HIPAA: Savvy BAA available — confirm.
  • Patientdesk — Patient-engagement platform with AI phone module. Configuration paths: Communication → Voice AI → Scripts; Communication → Voice AI → Escalation. Native fields: script block + escalation phone. HIPAA: Patientdesk is BAA-covered.
  • Autocalls — General AI calling platform. Configuration paths: Campaigns → Inbound → After-Hours; Campaigns → Inbound → Transfer Rules. Native fields: system prompt + transfer rules. HIPAA: confirm BAA tier before deploying with PHI.
  • Human answering service (no AI tool) — The output includes a printable triage script for the answering-service operator with the same four red-flag questions, the three-tier action table, and the handoff note template. The skill flags any answering-service vendor that is not BAA-covered.

The vendor-specific block ends with three configuration-verification items the office must confirm before going live: (1) the red-flag transfer rule actually fires end-to-end (test by speaking each of the four red-flag triggers); (2) the on-call DDS receives the alert SMS within 60 seconds of the transfer; (3) the call recording / transcript is preserved per the practice's HIPAA retention policy.

Section C — Chart Documentation Template

Every after-hours contact produces a chart entry within 24 hours, ideally before the patient's next visit. The skill produces a documentation template with the following fields, in this order:

  • Date and time of contact (inbound channel: phone / SMS / chat / AI agent)
  • Patient identity verification method (DOB confirmation, account number, security question)
  • Chief complaint in the patient's own words (quoted)
  • Pain scale (0–10), duration, character (sharp / dull / throbbing / electric / cold-sensitive / spontaneous)
  • Red-flag screening results (each of the four questions, Y/N)
  • Swelling status (location, size, progression)
  • Bleeding status (controlled / uncontrolled, duration, pressure-applied)
  • Trauma history (mechanism, time, loss of consciousness, head/neck involvement)
  • Systemic symptoms (fever Y/N with temp, dysphagia, dyspnea, malaise)
  • Medical history flags (anticoagulants, diabetes, immunocompromise, pregnancy, recent surgery)
  • Tier assigned (1 / 2 / 3) with rationale
  • Self-care guidance given (verbatim or summarized with check-back time)
  • On-call provider notified (Y/N, time, method, outcome of notification)
  • Booking action taken (appointment time, location, provider)
  • Operator / AI-agent name (who handled the call)
  • Sign-off by on-call DDS within 24 hours

Guardrails

  • Never diagnose over the phone. Triage assigns urgency; it does not confirm a pulpitis classification or the need for extraction vs. endo.
  • Never prescribe or recommend a prescription through the triage protocol itself. Rx decisions are the on-call provider's after they have spoken with the patient.
  • Never ignore a red-flag symptom because the patient downplays it. If the screening question is a yes, escalate to 911/ER regardless of other context.
  • Never advise discontinuing prescribed medications (anticoagulants, antibiotics, steroids) during triage. Route that question to the prescribing provider.
  • Pediatric patients with any facial trauma, any avulsion of a permanent tooth, or any uncontrolled bleeding should be routed more conservatively than adults — default to the higher tier.
  • Anticoagulated patients with any bleeding presentation route one tier higher.
  • Pregnant patients with any infection sign route to Tier 1.
  • HIPAA — never leave a voicemail with clinical details; confirm identity before any clinical discussion. AI phone tools must be BAA-covered before deployment with PHI.
  • State scope-of-practice varies — teledentistry permissions, out-of-state-patient rules, and after-hours Rx authority differ; the triage protocol must defer to local regulation.
  • AI-generated triage output is a starting protocol, not a replacement for a licensed provider's judgment, and must be reviewed and signed off by the dentist of record before being deployed.
  • Call recording / SMS retention must comply with state two-party-consent rules and the practice's HIPAA retention policy — flag if the named AI tool's default recording posture does not match the practice's state rules.

Cross-Reference Graph

This skill explicitly chains with:

  • Upstream: config.yml (on-call DDS, ER referral hospital, AI phone tool selection, preferred pharmacies); knowledge-base/regulations/ (state-specific Rx and teledentistry rules); knowledge-base/tools-ecosystem/ai-phone-receptionists.md (tool-specific configuration paths)
  • Sibling: scheduling-optimizer (Tier 2 next-AM bookings route through the scheduling-optimizer's emergency-slot reservation logic); morning-huddle-brief (overnight after-hours contacts are reviewed in the huddle as a standing agenda item); clinical-note-assistant (chart documentation template is the input format for the post-visit clinical note)
  • Downstream: patient-reactivation-sequence (Tier 3 routine callbacks that fall off the calendar route to reactivation); monthly-practice-kpi-report (after-hours-contact volume, red-flag-rate, and tier-1 escalation-rate feed the KPI dashboard); meeting-summarizer (end-of-week after-hours debrief uses the chart entries as input)

Common Pitfalls To Avoid

  • Do not deploy an AI phone tool without confirming end-to-end red-flag escalation works (test all four triggers before going live)
  • Do not deploy an AI phone tool without a BAA in place — PHI in a non-BAA tool is a breach
  • Do not let the AI tool answer outside dental scope — the refusal-to-answer boundary must be explicit
  • Do not assume the on-call DDS will see an SMS notification within 60 seconds — test the alert pathway and have a fallback
  • Do not skip the chart entry on routine after-hours callbacks — the entry is the malpractice defense
  • Do not record or transcribe a call without verifying the state's two-party-consent rule
  • Do not include clinical details in a voicemail or SMS to a number you have not confirmed belongs to the patient
  • Do not default to the same triage tier across pediatric, adult, anticoagulated, and pregnant patients — the escalation rules differ
  • Do not let the AI tool make a definitive diagnosis or recommend a specific Rx — both are out of scope and create liability
  • Do not omit the post-visit sign-off by the on-call DDS within 24 hours — unsigned after-hours contacts are weak evidence in malpractice review

Example Output

[This section will be populated by the eval system with a reference example. For now, run the skill with sample input to see output quality.]

This skill is kept in sync with KRASA-AI/dental-ai-skills — updated daily from GitHub.