CareLink Project Overview
Inspiration
Globally, medication non-adherence is one of the most pressing issues in healthcare. According to the World Health Organization, only about 50% of patients with chronic illnesses in developed countries take their medications as prescribed. In the U.S. alone, the CDC estimates that non-adherence leads to 125,000 preventable deaths and 10% of hospitalizations each year. Beyond medication, patients frequently forget or misinterpret doctors’ instructions after leaving the clinic, especially when dealing with multiple prescriptions or complex care plans. This lack of follow-through directly impacts patient outcomes, increases healthcare costs, and places additional strain on providers. We were inspired to build CareLink to address this gap and ensure that what happens in the clinic translates into action at home.
What it Does
CareLink bridges the gap between doctors and patients by transforming medical visits into actionable, trackable care plans. The platform:
- Captures consultations
- Generates professional SOAP notes for clinicians
- Produces simplified summaries for patients in plain language
- Provides a compliance dashboard visualizing adherence trends (green: good, yellow: missed, red: critical)
Patients receive reminders and summaries that make care instructions easier to remember and follow, while providers gain proactive insight into adherence before small lapses become critical issues.
CareLink is fully HIPAA-compliant, ensuring that all patient data—including audio recordings, transcripts, and summaries—is securely stored, encrypted, and accessible only to authorized users.
How We Built It
CareLink is a full-stack web application with distinct roles for doctors and patients:
Frontend: React + Tailwind CSS
- Doctor Dashboard
- Capture Screen (live transcription & entity detection)
- Review Screen (SOAP notes & patient summaries)
- Patient Portal (visit recaps & reminders)
Backend: Supabase for authentication, storage, and database management
- Custom API layer handles encounters, transcripts, SOAP notes, and summaries
- HIPAA-compliant encryption at rest and in transit, strict access control
ML/NLP Pipeline:
- Audio streams transcribed in real-time using OpenAI’s Whisper for high medical transcription accuracy
- Speaker diarization separates doctor and patient voices
- Medical entity extraction via spaCy, medspaCy, scispaCy
- Transformer-based summarization generates:
- Professional SOAP notes for clinicians
- Plain-language summaries for patients (6th–8th grade reading level)
- Audio streams transcribed in real-time using OpenAI’s Whisper for high medical transcription accuracy
Integration: FHIR-compliant EHR exports, QR-based patient intake, secure PDF sharing
Challenges We Ran Into
- Medical transcription accuracy: Clinical conversations include jargon, drug names, acronyms, and overlapping speech. Whisper improved transcription reliability.
- Dual summarization: Producing both clinician-ready SOAP notes and patient-friendly summaries while balancing accuracy and latency.
- HIPAA compliance: Ensuring security and privacy across the ML/NLP pipeline required careful design and strict access enforcement.
Accomplishments We're Proud Of
- Created a working prototype in under 8 hours generating dual outputs from the same encounter
- Built a compliance dashboard giving providers visibility into adherence for proactive intervention
- Developed a HIPAA-compliant solution addressing a critical gap in healthcare delivery
What We Learned
- Balancing usability for two roles without compromising functionality
- Integrating HIPAA-compliant transcription, summarization, and data visualization into a cohesive pipeline
- Designing impactful solutions for systemic healthcare problems like adherence and comprehension
What's Next for CareLink
- Integrate with existing EHRs using FHIR standards
- Add multilingual support
- Develop mobile features: medication reminders, push notifications
- Long-term vision: Become a trusted, HIPAA-compliant bridge between clinical visits and everyday care, reducing preventable complications and improving health outcomes at scale
Built With
- ai
- data
- javascript
- openai
- python
Log in or sign up for Devpost to join the conversation.