Product Requirements Document (PRD): GoldenCare Check-in

1. Executive Summary

GoldenCare Check-in is a mission-critical safety platform designed to support elderly individuals living independently and provide peace of mind to their families. The service automates daily wellness checks via Interactive Voice Response (IVR) calls. If a senior fails to respond after a defined retry sequence, the system executes a multi-channel escalation protocol to alert family members or professional caregivers. By combining high-concurrency telephony with a modern, senior-accessible interface, GoldenCare bridges the gap between total independence and 24/7 assisted living.

2. Problem Statement

Millions of seniors live alone, facing risks ranging from medical emergencies (falls, strokes) to the debilitating effects of social isolation. Conversely, adult children and caregivers suffer from "caregiver burnout" and constant anxiety, often needing to interrupt their workdays for manual check-in calls. Current solutions are either too invasive (cameras), too expensive (full-time care), or rely on the senior proactively pressing a wearable button, which is impossible if they are unconscious or incapacitated.

3. Goals & Success Metrics

Safety Goal: Ensure 100% of missed check-ins result in a caregiver notification within 15 minutes of the final failed attempt.
Adoption Goal: Achieve a 90% "successful check-in" rate for active users through intuitive UI/IVR.
Operational Goal: Scale to handle 10,000+ concurrent morning calls with sub-100ms latency.
Metric 1: Mean Time to Resolution (MTTR) for alerts.
Metric 2: Caregiver Peace-of-Mind Score (via quarterly NPS).
Metric 3: False Alarm Rate (aiming for <5%).

4. User Personas

A. Independent Senior (e.g., "Sarah", 82)

Profile: Lives alone, tech-averse, values her privacy.
Need: A non-intrusive way to let her family know she’s okay without needing to use a complex smartphone app daily.

B. Adult Child/Caregiver (e.g., "Chris", 50)

Profile: Busy professional, lives in a different city, manages his mother’s well-being.
Need: Automated confirmation that his mother is safe and an immediate, reliable alert if she isn't.

C. Agency Administrator (e.g., "Alice", 40)

Profile: Operates a home health agency with 50+ clients.
Need: A multi-tenant dashboard to monitor all clients at once and audit logs for compliance.

5. User Stories

As a Caregiver, I want to schedule automated calls for 9:00 AM daily so that I don't have to remember to call every single morning.
As a Senior, I want to simply press "1" on my phone or speak "I'm okay" to verify my safety so that I don't have to navigate a complex app.
As a Caregiver, I want to receive a push notification and then an SMS if my parent misses their call, so that I am alerted even if my phone is on "Do Not Disturb" for work.
As a Senior, I want a "Vacation Mode" I can toggle so that the system doesn't call me when I'm staying with family.

6. Functional Requirements

6.1 Call Engine (IVR)

Automated Scheduling: Daily triggers based on preferredCallTime and timezone.
Verification: Support for PIN-based entry and Voice-to-Digit validation (Twilio <Gather>).
Retry Logic: Automatically retry 3 times at 10-minute intervals if the line is busy or no answer.

6.2 Caregiver Dashboard

Real-time Monitoring: A "Status Board" showing today’s check-in results (Pending, Success, Alert).
Escalation Management: Define a hierarchy of contacts (Primary, Secondary, Tertiary).
Activity Logs: Historical view of all calls, durations, and response types.

6.3 Notification System

Escalation Chain: Push Notification (Immediate) -> SMS (5 mins later) -> Automated Voice Call to Caregiver (10 mins later).
Resolution: Ability for a caregiver to "Mark as Safe" to stop the escalation chain.

6.4 AI & Voice (Nice-to-Have)

Mood Tracking: Use Hume AI or OpenAI Realtime to detect vocal strain or sadness during the check-in.
Natural Synthesis: Use ElevenLabs for warm, grandfatherly/grandmotherly tones.

7. Technical Requirements

7.1 Core Tech Stack (2026 Standards)

Backend: NestJS v11.1.12 (Express v5 engine).
Frontend (Admin/Web): React v19.2.3 with Tailwind CSS v4.1.18 (Oxide engine).
Mobile App: React Native v0.83.1 (New Architecture enabled).
Auth: Clerk (B2B Organizations for agency multi-tenancy).
Database: PostgreSQL 17 (with pgaudit and Row-Level Security).
Queue Management: BullMQ with Redis for call throttling.

7.2 Integrations

Telephony: Twilio (Programmable Voice with <Gather> for speech/DTMF).
Notifications: Firebase Cloud Messaging (Push), Twilio (SMS).
Email: SendGrid (Weekly PDF summaries).

8. Data Model

Entity: `SeniorProfile`

id: UUID
caregiverId: UUID (FK)
phoneNumber: E.164 string
checkInTime: Time (UTC)
status: Enum (Active, Paused, Alert)
verificationPin: Hashed String

Entity: `CheckInLog`

id: UUID
seniorId: UUID (FK)
timestamp: DateTime
outcome: Enum (Success, NoAnswer, FailedVerification, Busy)
retryCount: Integer

Entity: `EscalationPolicy`

id: UUID
seniorId: UUID (FK)
steps: JSONB (Order of contacts and methods)

9. API Specification (Sample)

POST `/v1/calls/trigger-manual`

Desc: Manually triggers a check-in call for a specific senior.
Request: { "seniorId": "uuid" }
Response: 202 Accepted

POST `/v1/webhooks/twilio/status`

Desc: Receives call status updates (busy, completed, failed).
Security: Twilio Signature Validation middleware.

GET `/v1/seniors/:id/status`

Desc: Returns current status and last 5 check-in logs.

10. UI/UX Requirements

10.1 Caregiver Interface (Web/Mobile)

Dashboard: High-contrast "Traffic Light" system (Green = Safe, Yellow = Calling, Red = Alert).
Configuration: Simple time-picker with timezone auto-detection.

10.2 Senior Experience (Audio)

Greeting: "Hi [Name], this is your GoldenCare check-in. Please say your 4-digit PIN or press it on your keypad now."
Error Handling: "I didn't quite get that. Let's try once more."

10.3 Accessibility (Seniors/Caregivers)

Touch Targets: Minimum 60x60dp for mobile buttons.
Contrast: 4.5:1 ratio for all text elements.
Screen Readers: Full accessibilityLabel support for VoiceOver/TalkBack.

11. Non-Functional Requirements

Security: HIPAA 2026 Compliance (Strict handling of PII, 72-hour breach reporting automation).
Privacy: Row-Level Security (RLS) ensuring caregivers only see their own seniors.
Availability: 99.9% uptime (Critical safety service).
Scalability: Horizontal scaling of BullMQ workers to handle peak 8:00 AM–10:00 AM call volumes.

12. Out of Scope

Direct 911 dispatch (We notify family; family calls emergency services).
In-home hardware (No proprietary "base stations").
Medical diagnosis or health advice.

13. Risks & Mitigations

14. Implementation Tasks

Phase 1: Project Setup & Infrastructure

[ ] Initialize NestJS v11.1.12 backend with PostgreSQL
[ ] Initialize React v19.2.3 Web Dashboard with Tailwind CSS v4.1.18
[ ] Initialize React Native v0.83.1 Mobile App (New Architecture)
[ ] Configure Clerk for Multi-Tenant Auth
[ ] Setup Redis and BullMQ for call queuing

Phase 2: Core Call Engine (IVR)

[ ] Implement Twilio Voice Service with <Gather> logic
[ ] Build PIN verification service (Hashing and Validation)
[ ] Implement BullMQ "Call Producer" (Cron-based scheduling)
[ ] Implement BullMQ "Call Consumer" (Throttled execution)
[ ] Build Webhook handler for Twilio status callbacks

Phase 3: Escalation & Notifications

[ ] Create Escalation State Machine (Push -> SMS -> Call)
[ ] Integrate Firebase Cloud Messaging for Mobile Push
[ ] Build "Alert Resolution" logic (Caregiver kill-switch)
[ ] Implement retry-logic (3 attempts, 10-min delay)

Phase 4: Caregiver Dashboard & Mobile

[ ] Build Senior Onboarding Flow (Phone, Timezone, PIN)
[ ] Create Real-time Status Dashboard (Web)
[ ] Implement High-Contrast UI for Senior-facing settings
[ ] Build "History" and "Log" views

Phase 5: Security & HIPAA Compliance

[ ] Implement pgaudit for database access tracking
[ ] Configure Row-Level Security (RLS) in PostgreSQL
[ ] Set up AES-256 Encryption at Rest for PII
[ ] Build Reproductive Health Data Protection logic (per 2026 HIPAA Update)
[ ] Implement automated 72-hour breach notification trigger

Phase 6: Testing & Launch

[ ] Conduct load testing for 1,000 concurrent calls
[ ] Accessibility audit (WCAG 2.2) on Mobile/Web
[ ] Beta testing with 10 caregiver/senior pairs
[ ] Production deployment to AWS (RDS, Lambda, S3)