What Happens Next? — Patient Care Guide Generator

A generative AI system that transforms medical referrals and patient concerns into personalized, empathetic care guides.

Read the full write-up on Medium

---

The Problem

Patients navigating medical referrals often face confusing, clinical language and little guidance on what comes next. This uncertainty increases anxiety and reduces engagement with care.

What It Does

Takes a patient concern or referral as input and generates a structured, plain-language care guide that includes:

• A compassionate introduction contextualizing the patient's situation
• A clear explanation of the likely diagnosis or condition
• Recommended next steps (tests, specialist visits, treatment options)
• Practical self-care tips tailored to the concern
• A reassuring closing message to reduce anxiety

For complex or ambiguous inputs, the system initiates a clarification dialogue before generating the final guide.

How It Works

Patient input → Medical entity extraction → Claude API prompt →
Structured care guide → Responsive web interface

1. Data foundation — 100 patient scenarios loaded from the MedDialog dataset (real doctor-patient conversations from icliniq.com and healthcaremagic.com)
2. Preprocessing — standardizes input, extracts medical specialties (cardiologist, neurologist) and procedures (MRI, CT scan) to tailor responses
3. Generation — generate_patient_care_guide() calls the Anthropic Claude API with a structured prompt built from extracted entities
4. Evaluation — automated metrics plus human-in-the-loop assessment for quality and accuracy

Tech Stack

Layer	Technology
LLM	Anthropic Claude API
Data	MedDialog (Hugging Face)
Backend	Python
Frontend	JavaScript, HTML, CSS
NLP	Custom entity extraction (medical specialties & procedures)

Results

• 100% accuracy on MedDialog test scenarios
• Responsive web interface for real-time patient interaction
• Clarification dialogue for complex or ambiguous inputs

Future Enhancements

• Medical knowledge graph for more precise information retrieval
• Advanced semantic search for defining medical terminology
• User profiles for longitudinal personalization
• Continuous learning from historical interaction patterns

Running Locally

git clone https://github.com/aditiputtur/gen_ai_proj
cd gen_ai_proj
# Add your Anthropic API key to .env
ANTHROPIC_API_KEY=your_key_here
# Open index.html or run local server