Rokhaya10/renaltrack-senegal

# RenalTrack Senegal — v0.2 ### MIMIC-Inspired Longitudinal EHR & Clinical Research Platform **Cardio-Renal Health in African Populations** > **Research prototype — synthetic data only. Not validated for clinical use.** --- ## What This Is RenalTrack Senegal is a **MIMIC-inspired longitudinal EHR and survival modeling platform** designed for low-resource African clinical settings, starting with Senegal. It is explicitly **not** a simple CKD risk calculator. The architecture is designed from the beginning to support: - Longitudinal patient tracking across multiple visits - Multi-outcome survival analysis (kidney, cardiovascular, hospitalization, death) - Population health analytics and Kaplan-Meier stratification - Research-grade model benchmarking and data export - A clear path to publication-quality clinical research The current version uses **5,000 synthetic patients** with **43,000+ visits** generated by a clinically-grounded longitudinal simulator. The data schema is designed so that real hospital data can replace the synthetic cohort without changing any downstream code. --- ## Project Structure ``` RenalTrack/ ├── app.py # Streamlit UI — 8 tabs, no business logic ├── synthetic_data_generator.py # Longitudinal EHR data generator (MIMIC-style) ├── model_training.py # Modular model registry (Cox active, 4 stubs) ├── requirements.txt # Streamlit Cloud-compatible dependencies └── README.md # This file ``` ### Module separation (why it matters) Each file has a single responsibility and **zero cross-dependencies on Streamlit**: | File | Responsibility | Can run standalone? | |---|---|---| | `synthetic_data_generator.py` | Generate normalized EHR tables | ✅ Yes | | `model_training.py` | Train, predict, evaluate models | ✅ Yes | | `app.py` | Render the Streamlit UI | Requires Streamlit | This means you can run `python model_training.py` to validate model performance on a server, or run `pyt …