<<<<<<< HEAD
To simulate a modular Clinical Decision Support (CDS) pipeline using real-world ICU data via an integrated framework that spans:
- Cohort selection from MIMIC-IV
- Risk prediction using ML models (e.g., sepsis, readmission)
- Transparent reasoning via model interpretability
- (Optional) ClinicalBERT-based summarization of notes
- Export to standardized FHIR-compatible formats
This system models how modern CDS platforms can bridge data science and clinical workflows while preserving transparency, modularity, and alignment with ethical AI principles.
| Notebook | Description |
|---|---|
1_data_preparation.ipynb |
Cohort creation and merging MIMIC-IV tables |
2_risk_model_training.ipynb |
ICU risk model training + prediction export |
3_fairness_analysis.ipynb |
Fairness metrics across gender, race, and age |
4_model_explainability.ipynb |
Global interpretability using Permutation Importance |
5_model_card.ipynb |
Structured model card (see below) for responsible ML reporting |
This project foregrounds explainability and ethical alignment:
- ❌ SHAP was evaluated but excluded due to known incompatibilities with class-weighted ensembles
- ✅ Permutation Importance was used as a model-agnostic alternative
- 📈 Fairness metrics computed for race, gender, and age groups
- 📦 Dataset: MIMIC-IV v2.2
- ✅ Fully de-identified under the official data use agreement
- 🔒 No PHI stored, transmitted, or processed outside local environment
- ⚖️ Built in line with Responsible AI principles:
- Fairness
- Transparency
- Nonmaleficence
# Clone the repository
git clone https://github.com/sobcza11/mimiccds.git
cd mimiccds
# (Optional) Create a virtual environment
python -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate
# Install dependencies
pip install -r requirements.txt
# Launch notebooks
jupyter lab
---
## 📚 Citation
If using this project in academic or research work, please cite the MIMIC-IV dataset per [MIT-LCP guidelines](https://mimic.mit.edu/docs/iv/modules/data-reference/).
=======
# mimiccds
>>>>>>> 1c16d4339fe2c84d008b1cab57beb65d5bcc679f