Handoff

A web app that turns patient records from EHRs and whiteboard notes into clear, reliable summaries, tasks, and routes. Nurses save time, reduce errors, and get the info they need in one place.

Comment

🩺 Handoff — Project Story

🌟 Inspiration

This project started with noticing how stressful and fragmented nurse handoffs can be. Critical details live across EMRs, handwritten notes, sticky pads, and whiteboards. Nurses are expected to synthesize all of this under intense time pressure, with any missed detail impacting care quality and patient safety. Most problems fall into one of three categories: loss of data, duplicate data, and time sinks.

We wanted to build something that reduces this cognitive overload nurses have, supporting faster decision-making and ensuring that nurses walk into each shift confident and informed. That goal inspired the idea of an AI-powered assistant that could unify all handoff inputs into one trustworthy, structured summary.

🧭 What It Does

Handoff streamlines shift-to-shift communication by turning scattered clinical information into a single, clean hands-off (get it?) summary.

Key capabilities:

  • AI-generated and editable handoff notes with auto-save from structured patient data
  • Image analysis of whiteboards or documents, merged into the patient record and turned into logs
  • Detailed Patient View with vitals, meds, tasks, allergies, and safety alerts
  • Claude support for flexible AI inference
  • Real-time updates via Supabase whenever patient/room/task data changes
  • AI-generated routes to visit patients most efficiently.
  • AI-generated scheduler for efficient handoff.

The result: a fast, accurate handoff workflow that fits naturally into a nurse’s routine.

🛠️ How We Built It

Frontend: React + Vite v4

  • A modern dashboard for viewing rooms, tasks, schedules, and logs
  • Patient Detail view for clinical context and AI handoff generation
  • Image upload pipeline for whiteboard/document analysis
  • Smooth animations and responsive UI built for clarity and speed

Backend: Express + Node.js

  • Endpoints for multimodal AI summarization
  • File upload handling via Multer
  • Handoff logging, task creation, and patient record management
  • Secure environment-variable-based configuration

Database & Auth: Supabase (PostgreSQL)

  • Tables for patients, rooms, tasks, nurses, logs, and hand_offs
  • Real-time channels (postgres_changes) to sync UI instantly
  • Supabase Auth for secure nurse account access
  • Automatic nurse account creation using service role keys

AI Integration

AI merges multiple inputs:

[ \text{Patient Record} + \text{Whiteboard Image} \xrightarrow{\text{AI}} \text{Unified, Structured Handoff} ]

Claude Sonnet 4 is supported!

🚧 Challenges We Ran Into

1. Keeping AI clinically accurate

Large models sometimes hallucinated clinical details. We had to redesign prompts to enforce:

  • strict factual grounding
  • uncertainty notation
  • structured outputs
  • no invented symptoms or meds

2. Merging structured & unstructured data

Combining EHR-style fields with handwritten whiteboard content required careful logic to avoid contradictions and duplication.

3. Real-time syncing

Building a responsive dashboard that updates instantly with every patient/room/task change pushed us to deeply understand Supabase’s subscription model.

4. Designing for nurses under pressure

We iterated heavily to build a UI with:

  • minimal clicks
  • scannable sections
  • consistent layouts
  • zero clutter

Clinical software must be fast, reliable, and intuitive.

🏆 Accomplishments That We're Proud Of

  • A fully working multimodal AI pipeline that merges image + structured data
  • A realtime clinical dashboard built from the ground up
  • A smooth patient detail interface designed with nurse workflows in mind
  • Robust database modeling for tasks, vitals, handoffs, and logs
  • Security-minded implementation with Supabase Auth and server-side AI calls
  • Built an experience that genuinely helps clinicians instead of adding more complexity

📚 What We Learned

  • How to design prompts for high-stakes domains like healthcare
  • That multimodal AI (image + text) can dramatically simplify documentation
  • How to orchestrate a full-stack app using React + Express + Supabase
  • How nurses think about handoffs, what they worry about, and where tools often fall short
  • How to design UI for speed, clarity, and cognitive ergonomics

🚀 What's Next for Handoff

We're excited to expand this tool into something nurses can rely on every shift.

Upcoming ideas:

  • Handoff timeline/history so teams can track changes over time
  • Voice-to-handoff: letting nurses dictate updates hands-free
  • Mobile-first version for bedside use
  • Multilingual support for diverse care teams
  • Hospital system integrations (Epic, Cerner, etc.)
  • Smart alerts when AI detects missing info or inconsistencies

Ultimately, we envision Handoff as a real-world clinical safety tool — a bridge between human expertise and AI assistance during one of the most critical transitions in patient care.

Built With

Share this project:

Updates