A data engineering pipeline that predicts referral dropout risk in NGO-powered community eye health screening programs — built for Kenya and scaled globally.
# 👁️ Eye Health Dropout Risk Predictor
> A data engineering pipeline that predicts referral dropout risk in
> NGO-powered community eye health screening programs — built for Kenya
> and scaled globally across 10 countries.
---
## 🌍 The Problem
Over **1 billion people** live with unaddressed vision impairment globally.
The barrier is not medical technology — it is the gap between:
> *"You were screened and told to go to a clinic"*
> and
> *"You actually arrived and received treatment."*
NGOs like **Peek Vision**, **CBM**, **Fred Hollows Foundation**, and
**Sightsavers** run community eye screening programs across Africa, Asia,
and Latin America — but have no early warning system for which patients
will drop out before completing their referral journey.
This project builds that early warning system.
---
## 🎯 What This Project Does
| Layer | What Was Built |
|---|---|
| **Data Simulation** | Synthetic patient records for Kenya (50K) + 10 countries (100K) |
| **ETL Pipeline** | Extract → Validate → Transform → Load to SQLite |
| **Feature Engineering** | Risk tiers, vulnerability scores, distance bands, age groups |
| **ML Model** | Random Forest dropout classifier · ROC-AUC 0.71 |
| **Risk Scoring** | Every patient scored and ranked by dropout probability |
| **Dashboard** | Interactive HTML visualization across all regions |
---
## 📊 Key Findings
### Global Dropout Rates by Country
| Country | Region | Dropout Rate | UHC Index |
|---|---|---|---|
| Papua New Guinea | Pacific | 90.5% | 32 |
| Ethiopia | Sub-Saharan Africa | 78.4% | 34 |
| Uganda | Sub-Saharan Africa | 67.6% | 37 |
| Myanmar | Southeast Asia | 62.4% | 42 |
| Tanzania | Sub-Saharan Africa | 59.7% | 39 |
| Cambodia | Southeast Asia | 52.6% | 44 |
| Bangladesh | South Asia | 46.6% | 48 |
| Bolivia | Latin America | 39.4% | 52 |
| India | South Asia | 37.2% | 56 |
| Peru | Latin America | 23.4% | 65 |
### Kenya County Breakdown (Vision Impact Project)
| County | Dropout Rate | Needs Outre …