🏥 Hospital Deterioration — 12h Early Warning Baseline

A production-minded baseline notebook for predicting clinical deterioration in the next 12 hours from an ML-ready hospital cohort.

Notebook: hospital-deterioration-next-12h-early-warning-baseline.ipynb
Case study: CASE_STUDY.md

---

What this repository includes

• A single notebook with EDA, modeling, calibration, and threshold policies
• Exported artifacts under ./artifacts/ (models + threshold policy tables)

---

Dataset

Place your CSV files under data/raw/ (recommended):

Required

• hospital_deterioration_ml_ready.csv

Optional (enables temporal-window features)

• vitals_timeseries.csv
• labs_timeseries.csv

See data/raw/README.md.

The notebook also attempts to locate files under Kaggle inputs when running on Kaggle.

---

Quick start

1) Install

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

pip install -r requirements.txt

2) Run the notebook

Open and run:

• hospital-deterioration-next-12h-early-warning-baseline.ipynb

Outputs are exported to:

• artifacts/

See artifacts/README.md.

---

Outputs you get

The notebook produces decision-ready deliverables:

• baseline models (CPU-friendly)
• calibrated risk scores (when supported by the model)
• threshold policy tables for different alerting regimes:
  • balanced
  • high recall
  • low alerts

---

Disclaimer

This project is intended for educational and research use. It is not a certified medical device and must not be used for clinical decision-making without appropriate validation, governance, and regulatory review.