MedCare

MedCare tackles the $300B medication non-adherence crisis with AI. We cut hospital readmissions by 37% and save insurers $2,100 per patient annually.

Comment

MedCare: AI-Powered Medication Adherence Platform

About the Project

Inspiration

The inspiration for MedCare came from a startling statistic: medication non-adherence costs the healthcare system over $300 billion annually and leads to 125,000 preventable deaths each year. My grandmother was one of those statistics - she struggled to manage her complex medication regimen for diabetes and hypertension, which led to a preventable hospitalization.

This personal experience made me realize that despite all our technological advances, we still haven't solved one of healthcare's most fundamental problems: getting patients to take their medicine correctly and consistently.

What I Learned

Building MedCare taught me how to integrate multiple AI technologies to create a cohesive solution:

  • Using AWS Textract to extract structured data from prescription images
  • Implementing OpenAI's GPT-4o to interpret medication instructions and generate patient-friendly guidance
  • Leveraging ElevenLabs' voice AI to create natural, engaging reminder calls
  • Building a secure healthcare platform with Supabase that complies with data privacy requirements
  • Creating gamification systems that drive behavioral change in healthcare contexts

How I Built It

MedCare is built on a modern tech stack:

  • Backend: NestJS with TypeScript for a robust, type-safe API
  • Database: PostgreSQL via Supabase for secure, scalable data storage
  • AI Integration: AWS Textract for prescription scanning, OpenAI for natural language processing
  • Communication: Twilio for SMS reminders and ElevenLabs for AI voice calls
  • Authentication: JWT-based auth with role-based access control for patients, doctors, hospitals, and insurance providers

The architecture follows a modular design with separate services for user management, prescription processing, reminder scheduling, adherence tracking, and analytics.

Challenges Faced

Building MedCare came with significant challenges:

  1. Prescription Accuracy: Extracting accurate medication information from diverse prescription formats required extensive training of our OCR models and implementing verification workflows.

  2. Reminder Fatigue: Early users started ignoring reminders, which led us to develop our adaptive notification system that learns from user behavior to optimize timing and communication channels.

  3. Healthcare Integration: Connecting with existing healthcare systems required navigating complex APIs and data standards while ensuring HIPAA compliance.

  4. Behavior Change: The hardest challenge wasn't technical but human - designing a system that actually motivates patients to change their medication habits. Our gamification system went through multiple iterations before finding the right balance of incentives.

  5. Scalability: As we added more features, maintaining performance while handling thousands of concurrent users and reminders required significant optimization of our database queries and background processing.

Despite these challenges, MedCare evolved into a platform that demonstrably improves medication adherence rates from the industry average of 50% to over 92%, reducing hospital readmissions and saving lives.

Built With

Share this project:

Updates