Accessibility-Adjusted Emergency Room Burden Platform

FairQ is a real-time accessibility equity layer for emergency departments. It models how ER wait times disproportionately burden patients with functional accessibility needs and provides frontline staff and administrators with actionable insights to reduce inequity — without replacing triage or making clinical decisions.

Emergency departments measure wait time. They do not measure waiting burden.

Two patients with identical wait times can experience radically different distress due to:

• Mobility impairment
• Chronic pain
• Language barriers
• Sensory sensitivity
• Being alone

This creates invisible inequity.

FairQ introduces an accessibility-adjusted burden index that weights wait time by vulnerability. Staff see who is at elevated risk of distress or leaving without being seen (LWBS), with suggested actions to support patients equitably.

• Accessibility-weighted burden modeling engine
• Real-time burden refresh with backend simulation
• Alert threshold system (Green/Amber/Red)
• LWBS risk integration from Alberta HQCA data
• Vulnerability multiplier derived from StatsCan disability data
• Modular NestJS architecture with RESTful endpoints
• Angular SPA with route guards and session-based auth
• Dockerized full-stack deployment

Queue view with burden index, LWBS risk, accessibility flags, and suggested actions per patient.

Model health and alert distribution.

Burden analysis by accessibility needs (mobility, pain, sensory, cognitive, language, alone).

Step 1: Hospital & Discomfort	Step 2: Accessibility	Waiting View

flowchart TD
    A[Patient Intake] --> B[Accessibility Profile]
    B --> C[Vulnerability Multiplier]
    A --> D[Wait Time Snapshot]
    D --> E[AHS + HQCA]
    C --> F[Burden Engine]
    E --> F
    F --> G[CIHI + McMaster]
    G --> H[Staff Dashboard Alerts]
    G --> I[Admin Equity Overview]

Patient Intake → Accessibility Profile → Vulnerability Multiplier → Wait Time Snapshot (AHS + HQCA) → Burden Engine (CIHI + McMaster) → Staff Dashboard Alerts

nomoney/
├── backend/          # NestJS API (Node.js + TypeScript)
├── frontend/         # Angular SPA
├── docker-compose.yml
├── package.json      # Monorepo scripts
└── README.md

• Base URL (dev): http://localhost:3002/api (or http://localhost:4200/api via frontend proxy)

Endpoints:

• Base: http://localhost:4200

Routes:

Landing — At /, user chooses Staff or Patient.

Patient flow:

1. Click "Patient" → setPatientSession() → /patient/intake/1
2. Step 1: Select hospital, discomfort level, "thinking about leaving?"
3. Step 2: Accessibility toggles (mobility, pain, sensory, cognitive, alone, language)
4. Step 3: Confirm → POST /patients → store patient ID → /patient/waiting
5. Waiting: View patient ID, estimated wait. Link to /patient/checkin
6. Check-in: Discomfort, needs (interpreter, quiet space, etc.), planning (staying / unsure / leaving) → POST /patients/:id/checkins → back to waiting

Staff flow:

1. Click "Staff" → enter hospital code (001–005) → setStaffSession(code) → /staff
2. Dashboard: Patients for selected hospital, stats (total, alerts, avg wait), time controls (+30 min, Reset)
3. Patient cards: Burden index, LWBS risk, accessibility flags, latest check-in, suggested actions
4. Actions: "Staff checked in with patient", "Sent to Doctor — Off Queue"
5. Link to Admin → /admin

Admin flow — Same auth as Staff. Read-only metrics: Average Burden, Total Patients, High LWBS Risk, alert distribution, equity by flag.

• Staff: Hospital codes 001–005 map to backend keys (uofa, royalAlexandra, greyNuns, misericordia, sturgeon). Stored in sessionStorage.
• Guards: staffGuard protects /staff and /admin; redirects to / if not staff.
• Patient: No auth; patient ID and hospital key stored in sessionStorage for check-in.

• Node.js 20+
• npm

# Install all dependencies
npm run install:all

# From root directory run:
npm run dev

• Frontend: http://localhost:4200 (proxies /api to backend)
• Backend (direct): http://localhost:3002

docker-compose up --build

• API: http://localhost:3000
• Frontend: http://localhost:4200 (proxied to backend)

• Backend: Uses PORT (default 3002 for dev; 3000 for Docker), FRONTEND_URL (default http://localhost:4200) for CORS
• Frontend: src/environments/environment.ts — apiUrl is /api (proxied in dev to backend)
• Dev proxy: proxy.conf.json forwards /api/** → http://localhost:3002

backend/src/patients/patients.service.ts — In-memory patient store. Patients are registered at intake and receive check-ins. Staff can record a check-in (POST /patients/:id/staff-checkin) and remove patients from the queue (DELETE /patients/:id) when sent to doctor. lastStaffCheckInAt tracks when staff last checked in with a patient. All patients are assumed checked-in after completing the check-in form (no missed-check-in logic).

Data model — StoredPatient: id, waitStart, vulnerabilityScore, burdenIndex, alertLevel, flags (AccessibilityFlags), checkIns (discomfort, needsHelp, planningToLeave, assistanceRequested, planningToLeaveChoice, timestamp), lastStaffCheckInAt, assignedHospitalKey, estimatedCtasLevel, etc.

backend/src/wait-times/alberta-waittimes.snapshot.ts contains snapshot data for five Alberta hospitals:

• Wait times: Alberta Health Services (AHS)
• LWBS rates: Health Quality Council of Alberta (HQCA), Apr–Jun 2025

Update snapshotTakenAt and waitMinutes / lwbsRate before demos as data changes.

backend/src/burden-modeling/vulnerabilityWeights.ts — Source: Statistics Canada – Disability in Canada (2024). Weights reflect relative burden impact (not raw prevalence): chronicPain 0.25, mobility 0.20, cognitive 0.15, sensory 0.15, language 0.10, alone 0.10. Pass profile: { chronicPain?: boolean; ... } in burden compute; when provided, vulnerabilityMultiplier is derived from profile (0 → ~0.95). Otherwise use vulnerabilityMultiplier directly.

backend/src/burden-modeling/modelConstants.ts:

• MEDIAN_TOTAL_STAY_MINUTES = 238 (CIHI NACRS Alberta 2024–25)
• MEDIAN_TO_PHYSICIAN_MINUTES = 90
• MCM_MASTER_MEDIAN_TIME_TO_PHYSICIAN_MINUTES = 87 — McMaster median time-to-physician; reference for gradual escalation. Source: published ED LWBS study.
• MEDIAN_LWBS_TRIGGER_MINUTES = 87 — Gate for disengagement warning (show only when waited this long or planning to leave or burden ≥ 70).

1. Base waiting impact (CIHI) — Time-based burden from computeBaseWaitingImpact(minutesWaited):

   • Normalized against median total stay (238 min → ~60 burden)
   • +8 acceleration after 90 min (median time to physician)
   • Capped at 75

2. Post-87 min gradual escalation (McMaster) — applyPostMedianPhysicianDelayAdjustment(burden, minutesWaited):

   • +0 to +10 bump when minutesWaited exceeds 87 min
   • Capped at 10; then final burden clamped 0–100

3. LWBS integration (HQCA) — Environment-level disengagement:

   • computeLeaveSignalWeight(hospitalKey) normalizes LWBS around 5% baseline
   • If intendsToStay === false: burden += 15 * leaveSignalWeight

4. Vulnerability scaling (StatsCan) — burden *= (1 + vulnerabilityMultiplier) when using profile-derived multiplier.

5. Alert status — RED if burden > 75 or planning to leave; AMBER if burden > 50; else GREEN.

6. Amber check-in suggestion — suggestAmberCheckIn: true when minutesWaited > 87 and burden >= 55.

7. Disengagement gating (backend) — disengagementWindowMinutes returned only when: planningToLeave OR burden >= 70 OR minutesWaited >= 87 OR minutesWaited / expectedWaitMinutes >= 0.30.

8. Burden API response — burden, alertStatus, suggestAmberCheckIn, disengagementWindowMinutes, burdenCurve, equityGapScore, baselineCurve

Read-only view computed from frontend PatientStoreService (same data as Staff). Polls GET /patients every 3s and runs burden refresh so it reflects the same simulations and burden data as Staff.

• Hero metrics — Average Burden Index, Total Patients, High LWBS Risk (green gradient card).
• Model Health — Alert distribution (% Green/Amber/Red), average burden (30–55 normal, 70+ strain).
• Equity Overview — Average burden by flag (mobility, chronicPain, sensory, cognitive, language, alone) and % Red by flag.
• Model Data Sources — CIHI, McMaster Health Forum, HQCA, Statistics Canada.

Safety — Admin cannot change thresholds, weights, patients, LWBS scaling, or triage.

Check-in assumption — Every patient is treated as checked-in after filling the check-in form. No "Missed check-in" label or badge is shown.

LWBS risk — "Risk of leaving without being seen" uses burden-curve-based LWBS probability. When a patient indicates "Thinking of leaving" in a check-in, the metric is fixed at 80%.

Check-in display — Recent check-ins (within 5 min) show "New check-in"; older ones show "Latest check-in" with time ago.

Staff actions:

• Staff checked in with patient — Button to record that staff has checked in with the patient (updates lastStaffCheckInAt).
• Sent to Doctor — Off Queue — Button to mark patient as sent to doctor and remove from queue. Calls DELETE /patients/:id so the patient is permanently removed and does not reappear on sync.

Disengagement warning — Shown when: minutesWaited >= 87 OR intendsToStay === false OR burden >= 70.

Suggested actions — Flag-based (e.g. Provide seating support, Support staff check-in, Confirm patient intends to stay when planning to leave).

FairQ does not diagnose, prioritize treatment, or provide medical advice. It operates as an accessibility and system-equity support layer alongside existing clinical workflows.

• Uses snapshot wait-time data (not real-time AHS API integration)
• Vulnerability weights are impact-informed, not clinically validated
• In-memory patient store (no persistent database in current version)

UNLICENSED (private project)