# Blood Test Explainer — Changelog

**Project:** Blood Test Explainer Tools
**Live URL:** https://tools.aryash.health
**Maintained by:** Dr Yadavakrishnan Pasupathi (MBBS MBA MRCGP)
**GMC:** 6050795

---

## Version 2.2.0 — 4 January 2026

### Regulatory & Compliance Documentation
- **DPIA Screening Record** — Formal Data Protection Impact Assessment screening for all Aryash Health sites. Confirms no personal data processing; full DPIA not required.
- **Clinical Safety Case (DCB0160)** — Comprehensive hazard identification, risk assessment, and mitigations following NHS DCB0160 standard. 6 hazards documented with acceptable residual risk.
- **Intended Use Statement** — Formal statement defining purpose, intended users, clinical scope, limitations, and regulatory classification for the platform.

### Platform Scope Extended
- Governance documents now cover all Aryash Health sites:
  - aryash.health (landing page)
  - tools.aryash.health (blood tests, audio guides, infographics)
  - heart.aryash.health (cardiovascular health education)
  - mens.aryash.health (men's health education)

### Governance Portal Updated
- **New "Regulatory & Compliance" section** — Added to governance/index.html as first section
- **Multiple formats available** — All documents in .md, .html, and .docx formats
- **Public transparency** — Documents made publicly accessible to build trust

### Content Updates
- tools.aryash.health now lists 35+ blood tests, 40+ audio guides, 15 infographics
- heart.aryash.health covers 5 conditions, medications, investigations, recovery
- mens.aryash.health covers 6 health topics plus PSA/testosterone test guides

---

## Version 2.1.1 — 2 January 2026

### AI Chat Feature Removed
- **Removed Blood Test Q&A chatbot** — After extensive testing, LLM reasoning errors were identified (e.g., stating correct reference range then incorrectly classifying result). Decision made to prioritise clinical accuracy over AI features.
- **Rationale:** Static tools (learn.html, results.html) are fully controlled and clinically accurate. Live AI chat introduces uncontrollable risk where errors reach patients without clinician review.
- **Privacy policy updated** — Removed AI Chat section
- **Homepage updated** — Removed Online Features section

### Lesson Learned
AI is excellent as a **development tool** (helping build static content, reviewed before deployment) but requires extreme caution as a **patient-facing tool** (responses not reviewed before reaching users).

---

## Version 2.1.0 — 2 January 2026

### Blood Test Q&A (Added then Removed — see v2.1.1)
- **Patient Education Chatbot** — Added RAG-based Q&A feature using Lyzr AI agent with curated NHS-aligned knowledge base
- **Online Features Section** — New homepage section distinguishing online (Q&A) from offline tools
- **Renamed to "Blood Test Q&A"** — Changed from "Chat" to reduce "AI doctor" perception
- **Transparency Line** — Added: "Answers are based on reviewed NHS-aligned content and general medical education" (refined to accurately describe RAG+LLM behaviour)
- **n8n Integration** — Webhook proxy at `/webhook/ask-health` for secure API handling
- **Feature removed same day** — See v2.1.1 for details

### Safety & Governance
- **75+ Safety Tests** — Comprehensive testing of chatbot covering emergencies, boundaries, accuracy
- **100% Emergency Detection** — All emergency scenarios (chest pain, self-harm, overdose) trigger 999/Samaritans
- **100% Boundary Compliance** — Chatbot never diagnoses, never prescribes, always refers to GP
- **10 Stress Tests** — Additional adversarial testing (jailbreak attempts, unit confusion)
- **Hazard Log (DCB0129)** — Clinical risk assessment following NHS standards, 7 hazards documented with mitigations
- **Governance Index Page** — Created `governance/index.html` (previously 404)
- **Chat Security Audit** — Full security, privacy, accessibility assessment
- **Comprehensive Test Report** — Documented all 75+ test scenarios and results

### Privacy & Compliance
- **Privacy Policy Updated** — Added sections for "Request a Test" email field and AI chat data handling
- **Summary Revised** — Changed from "no data collection" to "minimal data collection" for accuracy
- **Request Page Safety** — Added emergency banner (999/111) to request.html

### Performance
- **Image Optimization** — Compressed infographics from 7-12 MB to ~130 KB (98% reduction)
- **Web Thumbnails** — Created `/images/resources/web/` with optimized JPEGs for faster page loads
- **Retained Print Quality** — Original PNGs kept for downloads

### Credibility & Trust
- **GMC Number** — Added to homepage footer (6050795)
- **Last Updated Date** — Added "January 2026" to footer
- **Clinically Reviewed Badge** — Added to footer
- **Feedback Link** — Added email link for user feedback

### External Reviews
- Received and addressed feedback from ChatGPT, Gemini, and Perplexity
- All major concerns resolved (privacy contradiction, governance 404, emergency handling)
- Risk assessment revised from "High Risk 7/10" to "Moderate Risk 3-4/10" after RAG clarification

---

## Version 2.0.0 — 28 December 2025

### Clinical Corrections
- **Fasting Glucose Boundary** — Changed from 6.0=normal to 5.9=normal, 6.0-6.9=pre-diabetic (aligns with NHS/NICE guidance)
- **Potassium Range** — Extended upper limit from 5.0 to 5.3 mmol/L (reduces false anxiety, within UK lab variation)
- **PSA Warning** — Added prominent warning box about PSA limitations when test selected in Learning Tool
- **Lab Variation Disclaimers** — Added to both tools: "Always compare your result to the reference range printed on your actual test report"

### Legal/Compliance
- **Privacy Policy** — Created dedicated privacy.html page documenting zero data collection
- **Terms of Use** — Created dedicated terms.html page with educational disclaimers
- **Interpretation Boundary** — Changed marketing example from "what does ferritin 15 mean?" to "My ferritin is marked 'low' — what should I ask my GP?"
- **Comparison Table** — Changed "Unlimited" to "Any test" for clarity

### Patient Experience
- **Category Filters** — Added quick-filter buttons to Results Guide (Diabetes, Kidney, Liver, Thyroid, etc.)
- **Tests Preview** — Added 35-test tag cloud to landing page
- **Clean URLs** — Created /learn.html and /results.html for cleaner, more trustworthy URLs
- **Mobile Statement** — Added explicit statement about device compatibility
- **GP Questions Teaser** — Added prominent section on hub highlighting key differentiator
- **Offline Capability** — Made offline feature more prominent in descriptions

### Patient Safety
- **PWA Support** — Added manifest.json for "Add to Home Screen" functionality
- **Reference Range Education** — Added expandable "Why do reference ranges vary?" section to every test
- **Print GP Questions** — Added feature to print questions with checkboxes for appointments

### Governance
- Created governance/ folder with clinical documentation
- Added .vercelignore to exclude governance from deployment
- Created outreach materials (poster, leaflet, email templates, QR code)
- Conducted and documented 4 external AI reviews

---

## Version 1.0.0 — December 2025

### Initial Release
- **Learning Tool (Claudius)** — Interactive sliders for 35 blood tests with educational explanations
- **Results Guide (Aryash)** — Look up any test result with GP questions for each test
- **Landing Page** — Hub page comparing both tools with feature matrix

### Core Features
- 35 common blood tests covered
- UK NHS reference ranges used throughout
- Plain English explanations at reading age 12
- Emergency signposting (999/111) on all pages
- Educational disclaimers throughout
- Zero data collection — all processing local to browser
- Works offline after initial load
- Mobile-responsive design
- Accessibility features (skip links, keyboard navigation)

### Tests Included
Full Blood Count (5), Kidney Function (6), Liver Function (7), Diabetes (3), Lipids (5), Thyroid (3), Iron Studies (4), Bone/Calcium (2)

---

## Planned Future Features

| Feature | Priority | Status |
|---------|----------|--------|
| Email signup for Ask tool waitlist | Low | Not started |
| User testimonials section | Low | Not started |
| Embed documentation for IT teams | Low | Not started |
| Search-as-you-type | Low | Not started |
| Privacy-preserving analytics | Low | Not started |
| Symptom-to-test mapper | Future v2 | Not started |
| ARIA audit on sliders | Low | Not started |

---

## Version Numbering

This project uses semantic versioning:
- **Major** (X.0.0): Significant new functionality or breaking changes
- **Minor** (0.X.0): New features, backwards compatible
- **Patch** (0.0.X): Bug fixes and minor improvements

---

## External Reviews Conducted

| Reviewer | Date | Verdict |
|----------|------|---------|
| ChatGPT | 28 Dec 2025 | "Bulletproof for patients" |
| Perplexity | 28 Dec 2025 | "Well-executed MVP" |
| Claude.ai | 28 Dec 2025 | "Production-quality" |
| Gemini | 28 Dec 2025 | "Polished, professional" |
| ChatGPT | 2 Jan 2026 | "Strong concept, needs governance docs" → Fixed |
| Perplexity | 2 Jan 2026 | "Privacy contradiction, broken links" → Fixed |
| Gemini | 2 Jan 2026 | "RAG approach is fundamentally sound" |

Full review documentation available in governance folder.

---

*Last updated: 4 January 2026*