team_04

Blood Transfusion Modeling and Operational Delay Analysist

---

🏆 Presentation & Award

Our final project presentation (linked below) summarizes the key findings, visualizations, and model insights developed throughout this work.
The project was awarded 3rd place.

🔗 Final Presentation (Canva): View Presentation

---

📌 Project Overview

This project investigates a clinically critical question:

Does the timing of a blood transfusion influence patient outcomes, and can we predict the number of units a patient will require?

Delays in blood transfusion—caused by logistics, cross-matching, inventory shortages, or ICU workload—can significantly affect survival, especially for unstable patients. Our goal is to translate these operational challenges into a data-driven predictive model that supports clinicians and hospital systems in urgent settings.

---

🎯 Objectives

1. Predict how many blood units a patient will need based on clinical features at presentation.
2. Assess whether time-to-transfusion correlates with poorer outcomes.
3. Provide insights to support:
   • Faster clinical decision-making
   • Better blood bank resource allocation
   • Reduced complications associated with delayed transfusion
   • Operational efficiency across the transfusion pipeline

---

💡 Why This Matters

• ⏱ Time sensitivity: Ideally, transfusions should occur within <60 minutes, yet this is rarely achieved in practice.
• 🩸 Resource scarcity: Especially in under-resourced settings or countries with limited blood supply.
• 🧬 Population relevance: High prevalence of genetic disorders (e.g., in regions with frequent cousin marriages) increases demand for transfusions.
• 🏥 Operational burden: Every step—cross-matching, tagging, releasing units—adds significant lab workload and delays.

Understanding these factors through data is essential for improving logistics and clinical outcomes. Thus, this model aims to mitigate these challenges using predictive analytics.

---

📂 Repository Structure (Scripts Folder)

This folder contains all the modelling and preprocessing work used for the analysis.

team_04/
│
├── scripts/
│   ├── notebook-ML-model-team4.ipynb              # Model training to predict the number of blood units
│   ├── script_equiflow_blood_transfusion.py
│   ├── script_plots.py
│   └── script_tableone.py
│
└── results/
    ├── Equiflow_blood_transfusion.pdf
    └── transfusion_key_plots.png

---

Data used

The data used in this project originates from the MIMIC clinical database