CareThread

CareThread connects OPD visits, pharmacy counters, and post-discharge follow-ups into a continuous patient thread, with AI running silently at each handoff.

Comment

💡 Inspiration

The current healthcare journey is broken into information black holes across three critical touchpoints: the OPD consult, the pharmacy, and the post-discharge period.

  • Doctors lack patient context upon arrival
  • Pharmacists dispense drugs without full medication histories
  • Patients deteriorate silently after leaving the hospital

We wanted to weave these isolated failure points into a unified system that catches errors and readmission risks early.

🩺 What It Does

CareThread is a hospital-deployed web platform that operates across three core modules:

Module 1 — Smart OPD Intake

Patients scan a QR code in the waiting room and complete an NLP-powered intake form. This converts free-text symptoms into a structured clinical brief for the doctor before the consultation begins.

Module 2 — Pharmacy Interaction Guard

As pharmacists enter drugs for billing, the system cross-checks prescriptions against the patient's full medication history. It provides simple 🔴 / 🟡 / 🟢 flags to warn against:

  • Dangerous drug interactions
  • Polypharmacy risks (especially for elderly patients)

Module 3 — Post-Discharge Pulse

The system sends automated WhatsApp messages for quick daily check-ins. An ML model monitors response patterns using time-series anomaly detection, alerting the care team only when intervention is statistically warranted.

🔧 How We Built It

The platform runs entirely server-side — no new hardware or app installations required.

Module Technology
OPD Intake NLP entity extraction → structured clinical brief
Pharmacy Guard OpenFDA API · real-time drug interaction cross-check
Post-Discharge WhatsApp Business API · time-series anomaly detection

The pharmacy module integrates with drug interaction databases (such as OpenFDA) to run real-time cross-checks during the existing billing workflow.

🚧 Challenges We Ran Into

1. Zero behavioral change constraint Designing a system that monitors complex medical data — polypharmacy risks, dynamic readmission scoring — while requiring no additional steps from patients or staff.

2. Ambiguous free-text parsing The AI needed to accurately parse responses like "the wound smells a bit" and flag genuine deterioration signals without triggering false alerts for every minor complaint.

🏆 Accomplishments We're Proud Of

"The whole is greater than the sum of its parts."

By linking the three touchpoints, CareThread passively builds a longitudinal patient health record purely as a byproduct of normal hospital visits — no extra effort from anyone.

We successfully created an early warning system that delivers massive value to doctors and pharmacists at near-zero marginal cost.

📚 What We Learned

  • Meet patients where they are. Using WhatsApp messages and QR codes — instead of new apps — drastically reduces friction and drives adoption.
  • Pharmacists are the last line of defense. Equipping them with simple decision-support tools at the billing counter is one of the highest-leverage interventions in the entire patient journey.

🚀 What's Next for CareThread

  1. Scale deployment to hospital administrators to prove impact on:

    • Reducing readmissions
    • Improving OPD throughput
    • Building longitudinal patient data assets

  2. Refine the readmission risk scoring model so it dynamically updates based on long-term post-discharge response data

Built With

Share this project:

Updates

Submission history