ClarePath Health — Digital Patient Intake

Illustrative engagement — details anonymized.

Industry: Healthcare SaaS Service: Digital intake portal with AI pre-screening Timeline: 10 weeks Stack: Next.js, Supabase Auth, OpenAI

---

The problem

ClarePath Health operated a network of outpatient clinics where patient intake was still paper-based. New patients filled out 6–8 pages of forms in the waiting room, averaging 45 minutes. Front-desk staff then manually entered the data into the EHR system — another 15 minutes per patient, with frequent transcription errors.

The bottleneck created a ripple effect: appointments started late, waiting rooms were overcrowded, and patients arrived frustrated before they even saw a provider.

What we built

Digital intake portal: A mobile-first web application where patients complete intake forms before their appointment — at home, on their phone. Forms are adaptive: answers to early questions determine which follow-up sections appear, so patients only answer what's relevant.

AI pre-screening: An OpenAI-powered module that reviews completed intake forms and flags potential concerns for the clinical team. It identifies medication interactions, symptom patterns that may warrant priority, and missing information that should be collected before the visit.

Supabase Auth integration: Patients authenticate via magic link (no passwords to remember). Session management ensures partially completed forms are saved and resumed seamlessly.

EHR sync: A webhook-based integration that pushes structured intake data directly into the clinic's EHR system, eliminating manual data entry entirely.

Admin dashboard: Clinic staff see a real-time queue of incoming patients, completion status, and any AI-flagged items — all before the patient walks through the door.

The technical decisions

Adaptive forms over static PDFs: Rather than digitizing the existing paper forms 1:1, we built a conditional form engine. This reduced the average number of questions from 120 to 45 for a typical patient, which was the single biggest factor in the time reduction.

OpenAI with structured outputs: The AI pre-screening module uses structured JSON outputs to ensure consistent, parseable results. Flagged items include severity level, relevant form fields, and suggested follow-up questions. This avoided the "black box" problem — clinicians could see exactly why something was flagged.

Edge rendering for speed: The intake portal is edge-rendered via Vercel for sub-200ms load times. For patients on slow mobile connections in rural clinic areas, this made a measurable difference in completion rates.

Results

• Intake time: 45 minutes → 8 minutes (average)
• Data entry errors: Reduced by 94% (no manual transcription)
• NPS score: +34 points in the first quarter
• No-show rate: Decreased 12% (patients who completed intake digitally were more likely to show up)
• Staff time saved: ~20 hours/week across the clinic network

Handoff

Full codebase in client's GitHub. Supabase project transferred. OpenAI API keys rotated to client's account. Documentation includes form configuration guide, AI prompt templates, and EHR integration specs. 30-day bug warranty included.