ODC Shift Prediction Tool

A web-based tool for predicting the rightward shift in the oxyhaemoglobin dissociation curve (ODC) using paired SpO₂ and inspired O₂ (PiO₂) measurements. Designed for bedside use in neonatal and pediatric care, especially for preterm infants.

---

Quick Start

No installation required!
Simply visit the website:

https://odcshift.sarahrebeccameyer.com/

---

What Does This Tool Do?

• Estimates the “shift” in the ODC using your SpO₂ and PiO₂ measurements.
• Visualizes the ODC curve and your entered datapoints.
• Provides confidence estimates for each prediction.
• Gives instant feedback on input quality and model limitations.

---

How to Use

1. Enter 1–5 pairs of oxygen saturation (SpO₂, %) and inspired O₂ pressure (PiO₂, kPa).
2. Submit to receive:
   • The predicted ODC shift (rounded to two decimals)
   • Uncertainty (± SD)
   • Confidence level (high, moderate, low)
   • Curve visualization
3. Review warnings and tips for best results:
   • Recommended: 1–3 datapoints
   • Best performance: SpO₂ <92.5% or <95%
   • Model is less reliable for SpO₂ >95%
   • PiO₂ must be between 13–53 kPa; SpO₂ between 72–99.9%

---

Model & Limitations

• Trained on: 219 preterm infants
• Best for: SpO₂ between 72% and 95%
• Not validated for: Term infants, adults, or patients with hemoglobinopathies
• Limitations:
  • Points above 95% SpO₂ contribute less information and may reduce accuracy
  • In very severely impaired infants, the model may overestimate the shift, especially with only 1–2 datapoints
  • Reliable input is essential; ensure values are stable and artifact-free

---

Data Privacy

• No data is stored, logged, or transmitted.
• All processing is in-memory and temporary.
• No patient-identifiable information is used or required.

---

For Developers & Contributors

If you wish to contribute or run the project locally, see below.

Project Structure

shift_prediction/
├── backend/
│   ├── app.py
│   ├── environment.yml
│   ├── models/
│   ├── predict/
│   ├── utils/
│   └── ...
├── frontend/
│   ├── src/
│   ├── public/
│   └── ...
├── Dockerfile
├── requirements.txt
└── README.md

Backend

• Python Flask API
• Machine learning models in backend/models/
• Input validation in backend/utils/validation.py

Frontend

• React app (frontend/src/components/PredictForm.js)
• Responsive UI, input validation, and ODC curve visualization

---

License

This project is for research and educational use. See LICENSE for details.

---

Contact

For questions, feedback, or collaboration, please contact the repository owner.