🩸 Type & Screen + Transfusion Readiness Agent

FHIR-Based Clinical Decision Support Prototype

---

🧭 Overview

This project demonstrates how FHIR resources, synthetic lab data, and Python-based analytics can enable a pre-operative Type & Screen readiness alert and transfusion readiness dashboard.

It models how EHR/LIS integration (e.g., Cerner + WellSky) could automatically surface blood-readiness alerts to clinicians before surgery—supporting safer, faster, and more efficient perioperative workflows.

---

🎯 Goals

• Integrate Type & Screen status, specimen validity, and blood product readiness in one FHIR-driven view.
• Enable Clinical Decision Support (CDS) by analyzing Observation, Specimen, and ServiceRequest resources.
• Improve Patient Blood Management (PBM) by preventing same-day transfusion delays.

---

⚙️ Architecture

Workflow Summary

1. Data generation
   • make_synthetic_type_and_screen.py → Creates synthetic FHIR Observations (ABO, Rh, Antibody Screen).
   • make_synthetic_surgery_requests.py → Creates FHIR ServiceRequest resources for upcoming surgeries.
2. Data upload
   • upload_synthetic_type_and_screen.py and upload_synthetic_surgery_requests.py use Azure CLI tokens to POST resources to the FHIR server.
3. Alert evaluation
   • evaluate_tns_alerts.py queries the FHIR endpoint to detect:
     • ❌ No T&S before surgery
     • ❌ Latest T&S older than 72 hours pre-op
4. Visualization
   • Displays results in a Streamlit dashboard for OR or Blood Bank teams.

---

🧰 Tech Stack

Layer	Technology
Language	Python 3.11
Frameworks	Streamlit • Pandas • FHIR-Client
FHIR Server	Azure Health Data Services FHIR / HAPI-FHIR (local testing)
Data Source	Synthetic MIMIC-IV FHIR exports (Observation, Specimen, ServiceRequest)
Terminologies	LOINC (883-9, 10331-7, 890-4) • SNOMED CT (Blood Products)

---

🧪 Example FHIR Resources

Observation (ABO Group)

{
  "resourceType": "Observation",
  "code": { "coding": [{ "system": "http://loinc.org", "code": "883-9", "display": "ABO group" }] },
  "valueString": "O",
  "effectiveDateTime": "2025-11-19T08:00:00Z"
}

🏁 Quick Start

Create and activate a virtual environment (optional)

python -m venv .venv .venv\Scripts\activate # (Windows) source .venv/bin/activate # (macOS/Linux)

Install dependencies

pip install -r requirements.txt

Run key scripts

python scripts/make_synthetic_type_and_screen.py python scripts/make_synthetic_surgery_requests.py python scripts/upload_synthetic_type_and_screen.py python scripts/upload_synthetic_surgery_requests.py python evaluate_tns_alerts.py

🌟 About This Project

This repository is part of Bonnie K. Shackleford’s applied informatics work, connecting Laboratory Information Systems (LIS) and Electronic Health Records (EHR) using FHIR-based interoperability and AI-ready modeling. It demonstrates perioperative transfusion readiness logic and clinical decision support concepts aligned with Patient Blood Management (PBM) workflows.

👩‍💻 Author

Bonnie K. Shackleford Medical Laboratory Scientist • Health Informatics Graduate Student

Bridging LIS & EHR systems with FHIR interoperability and AI-driven clinical decision support.

🔗 LinkedIn: www.linkedin.com/in/bonnie-shackleford-mshim-candidate-mbahc-mls-ascp-89098815a

💻 GitHub: https://github.com/bkshackleford

📘 License

MIT License © 2025 Bonnie K. Shackleford Please use this project to improve healthcare interoperability and patient safety — adapt freely with attribution.