electricsheepafrica/maternal-health-subsaharan-africa-2026

---
dataset_info:
features:
- name: age
dtype: float64
- name: parity
dtype: int64
- name: gravida
dtype: int64
- name: anc_visits
dtype: int64
- name: gestational_age_weeks
dtype: float64
- name: haemoglobin_gdl
dtype: float64
- name: systolic_bp
dtype: float64
- name: diastolic_bp
dtype: float64
- name: bmi
dtype: float64
- name: birth_weight_kg
dtype: float64
- name: distance_to_facility_km
dtype: float64
- name: household_income_usd_monthly
dtype: float64
- name: n_previous_pregnancies
dtype: int64
- name: n_previous_complications
dtype: int64
- name: country
dtype: string
- name: location_type
dtype: string
- name: facility_type
dtype: string
- name: education_level
dtype: string
- name: marital_status
dtype: string
- name: skilled_birth_attendant
dtype: string
- name: delivery_mode
dtype: string
- name: complication_type
dtype: string
- name: risk_tier
dtype: string
- name: adverse_outcome
dtype: int64
splits:
- name: train
num_examples: 499
- name: test
num_examples: 99
task_categories:
- tabular-classification
language:
- en
tags:
- africa
- nigeria
- kenya
- maternal-health
- synthetic
- machine-learning
- electric-sheep-africa
license: other
---

# Maternal & Pregnancy Health Risk Bundle — Teaser Dataset

This is the **public teaser** of the Maternal & Pregnancy Health Risk Bundle dataset bundle.
It contains the full schema, documentation, and a **499-row sample**.

**The complete bundle** — including the full dataset (35,000 rows), trained xgboost model (AUC-ROC: 0.990), and fully-executed notebook, and full Paper— is available on Gumroad:

👉 **[Get the full bundle on Gumroad for $30](https://kossisoro.gumroad.com/l/mphrsb)**

---

## Abstract

This pack provides a research-grade, ML-ready dataset for maternal and pregnancy health risk prediction in Sub-Saharan Africa, with a focus on Nigeria and Kenya. The dataset comprises 35,000 individual-level records (15,000 real-base + 20,000 synthetic augmentation) across 23 features spanning demographics, obstetric history, clinical measurements, care access indicators, and delivery outcomes. Every distribution parameter is traceable to a verified data source: WHO Global Health Observatory API, DHS Program API, World Bank API, or peer-reviewed publications (all verified March 2026). Key verified statistics anchoring the dataset: Nigeria MMR 993/100k [718–1,540] vs Kenya 379/100k [267–547] (WHO 2023); Nigeria facility delivery 41.0% vs Kenya 88.1% (DHS API); Nigeria anaemia in pregnancy 56.0% vs Kenya 40.3% (WHO GHO). The pack includes a baseline XGBoost classifier, ONNX export, inference wrapper, and full paper-style documentation.

---

## Dataset Card

| Attribute | Value |
|---|---|
| **Full dataset rows** | 35,000 (15,000 real + 20,000 synthetic) |
| **Teaser rows** | 598 (this download) |
| **Features** | 23 |
| **Target** | `adverse_outcome` |
| **Geography** | Nigeria, Kenya |
| **Model AUC-ROC** | 0.990 (on held-out test set, real data only) |

---

## Methodology Summary

All synthetic distribution parameters are grounded in peer-reviewed sources. Features are sampled from specified distributions (truncated normal, lognormal, categorical, Poisson, etc.) with parameters extracted from published literature. Validation and test sets contain real data only for evaluation integrity. See the full README in the Gumroad bundle for complete methodology.

---

## Limitations

- Geographic scope limited to Nigeria, Kenya
- Synthetic data may not capture complex multivariate interactions
- Not intended for direct production deployment without live data validation
- See full README in the Gumroad bundle for comprehensive limitations

---

## Citation

```bibtex
@dataset{esa_maternal_health_subsaharan_africa_2024_2026,
author = {{Electric Sheep Africa}},
title = {Maternal & Pregnancy Health Risk Bundle},
year = {2026},
version = {1.0.0},
publisher = {Gumroad},
}
```

---

*Electric Sheep Africa — Building Africa's AI data layer.*