Case Study: How We Built MediConnect for 50,000 Patients

In early 2025, a hospital chain approached us with a problem: their appointment wait times were 3-4 weeks, patients were frustrated, and doctors were overwhelmed. They needed telemedicine — fast. Here's how we built a platform that served 50,000 patients in 6 months.

The Challenge

HIPAA compliance was non-negotiable. The client had an existing EHR (custom-built, 10 years old) that we needed to integrate with. And they needed it live in 16 weeks — not a day more.

Our Architecture Decision

We chose Next.js for the web app, React Native for the patient mobile app, Node.js/FastAPI for the backend, WebRTC (LiveKit) for video, PostgreSQL for structured data, and AWS for infrastructure. All hosted in a dedicated VPC with end-to-end encryption.

HIPAA Compliance Implementation

• All PHI encrypted at rest (AES-256) and in transit (TLS 1.3)
• Business Associate Agreements (BAA) signed with AWS, Twilio, and all vendors
• Audit logs for every data access event, retained for 6 years
• Role-based access control — doctors only see their patients' data
• Automatic session timeout after 15 minutes of inactivity
• Signed BAA with video infrastructure provider (LiveKit)

The EHR Integration Challenge

The existing EHR had no public API. We built a custom HL7 FHIR adapter that polled the EHR database (read-only access) and synced patient data through our API gateway. This took 3 of our 16 weeks but was critical.

Scaling to 50,000 Patients

At launch, we had 2,000 patients. By month 3, 15,000. By month 6, 50,000 with 500+ active doctors. The auto-scaling setup on AWS ECS handled the growth seamlessly. Peak concurrent video calls: 800.

Key Metrics After 6 Months

• 50,000+ patients registered
• 500+ doctors onboarded
• Average wait time: 2 hours (from 3-4 weeks)
• Patient satisfaction score: 4.7/5
• Appointment booking increased 300%
• System uptime: 99.97%

What We'd Do Differently

We'd invest more time in the patient onboarding flow. Our initial registration had an 8-field form — we later A/B tested a 3-field version and improved completion rates by 40%. Always validate UX assumptions with real users before full build.