Signposting Toolkit — Project Summary

A multi-surgery GP signposting and care-navigation web application used by primary care teams to route patients safely and consistently.
Built by Dr Daniel Webber-Rookes, Ide Lane Surgery, Exeter.

The toolkit provides a structured, standardised symptom library, local customisation, AI-assisted clarity tools, an appointment directory, and a full governance workflow for clinical review — all delivered through a clean, modern, NHS-aligned interface.

1. Purpose

The Signposting Toolkit supports reception and care-navigation teams to:

• send patients to the right service first time,
• follow clear and clinically-approved guidance,
• reduce variation and inappropriate bookings,
• identify high-risk symptoms quickly,
• and maintain a robust clinical audit trail.

Each GP surgery has a fully independent configuration, symptom library, governance workflow, and appointment directory.

2. Key Features

🔹 Symptom Library (Base + Local Overrides)

• 200+ standardised symptoms included by default.
• Each surgery can override any field (title, wording, instructions, ages, icons).
• Local-only symptoms can be added.
• Clear badge indicating base or locally modified.
• Alphabet filter, age-group filters, search, and high-risk quick buttons.

🔹 Symptom Detail Page

• Clean NHS-styled layout.
• Highlight engine automatically formats:
  • orange slot, red slot, pink, purple, green, etc.
• Embedded icons (per symptom) displayed consistently.
• AI tools available directly inside the symptom page:
  • “Customise instructions”
  • “Get Questions to Ask”
• Buttons for suggestion feedback, hiding symptoms, and navigation back.

🔹 AI-Assisted Features

• AI Instruction Editor: rewrites instructions for clarity; admin reviews before publishing.
• AI Suggested Questions: generates grouped triage-style questions:
  • red flags
  • urgency/timing
  • pathway suitability
• All AI text must pass through clinical review before going live.
• Smart symptom updates available when base content changes centrally.

🔹 Appointment Directory

• Fully integrated top-level navigation.
• Simple searchable catalogue of appointment types.
• Colour-coded by staff team.
• CSV import for rapid updates.
• Built for reception staff: plain language, simple filtering, clear rules.

🔹 Preferences System (per browser)

• Appearance: Modern white or Classic blue cards.
• Header layout: Classic toolbar vs Split layout.
• Quick-scan mode: minimal card display.
• High-risk button style: pill or tile.

🔹 Clinical Review Workflow

• Every surgery must clinically approve all symptoms.
• Pending symptoms trigger a banner warning to all users.
• Approval recorded with reviewer identity and date.
• Re-review can be requested at any time (e.g. annually).
• All AI-generated content enters PENDING state.

🔹 Engagement Tracking and Suggestions

• Logs which symptoms are viewed and by whom.
• Tracks high-risk vs low-risk usage.
• Suggestions tab shows staff-submitted improvement ideas.

🔹 User Management & RBAC

Three-level access hierarchy:

1. Superuser (global)
2. Surgery Admin
3. Standard User

Each user may belong to multiple surgeries with differing roles.

🔹 Multi-Surgery Tenancy

• Each surgery has its own overrides, symptoms, appointment directory, users, and setup status.
• Superusers can switch between surgeries.

3. Technical Architecture

🔧 Stack

• Next.js 15 (App Router)
• TypeScript
• Tailwind CSS
• Prisma ORM — Neon Postgres
• NextAuth (Credentials Provider + JWT)
• Azure OpenAI (secure server-side calls)

🔧 Server Actions

Used extensively for:

• symptom updates
• override creation
• appointment directory CRUD
• AI generation endpoints
• clinical review actions

🔧 Rendering

• Mix of Server Components (data fetching) and Client Components (interactive UI).
• Caching layer for common public queries.

4. Data Model Overview

Core Models

• User
• Surgery
• UserSurgery (junction table)
• BaseSymptom
• SurgerySymptomOverride
• SurgeryCustomSymptom
• AppointmentEntry
• StaffTeam
• Suggestion
• EngagementEvent
• FeatureFlag
• ImageIcon

5. Feature Flags

Feature availability controlled via hierarchical flags:

• Superuser → Surgery → User

Current active flags include:

• ai_instructions
• ai_questions
• smart_symptom_updates
• (future flags can be added via admin panel)

6. Clinical Governance Model

Principles

• Surgeries are responsible for their own final wording.
• Base library is centrally maintained but overrideable.
• AI output must be clinician-reviewed.

States

• PENDING — awaiting local review
• APPROVED — clinically signed-off
• CHANGES_REQUIRED — flagged for updates

The system maintains an audit trail for:

• reviewer identity
• date of approval
• which fields changed

7. Application Flows

🟦 Staff Workflow

1. Search for symptom
2. Check age-group filters
3. Open symptom
4. Follow instructions
5. Book appointment or signpost service
6. Use high-risk guidance if needed

🟧 Admin Workflow

1. Select symptom
2. Edit fields or override base
3. Review highlight logic
4. Approve via Clinical Review
5. Monitor usage via Engagement tab

🟩 AI Workflow

1. Admin clicks “Customise Instructions”
2. AI generates suggestion → stored as DRAFT
3. Draft enters Clinical Review
4. Admin approves, modifies, or rejects

🟨 Appointment Directory Workflow

1. Import CSV or add entries manually
2. Assign staff teams & colours
3. Receptionists browse by search or team
4. Used as reference when interpreting instructions

8. External Dependencies

• Azure OpenAI API
• Neon managed Postgres
• Vercel serverless platform
• TailwindCSS + Heroicons
• TipTap (for future rich text improvements)

9. Known Issues & Future Improvements (Short List)

• Need surgery profile module.
• Multi-line membership badges in Users table (UX).
• Analytics dashboard not yet surfaced for admins.
• Symptom linking pathways could be expanded.
• Setup checklist can be made more dynamic.