Datasetpublic

koushik1212/synthetic-diabetes-hypertension-NCD-screening-WHO-HEARTS

Synthetic Diabetes & Hypertension NCD Screening Dataset (Adults 18–80) Abstract This dataset provides 30,000 synthetic records (10,000 per scenario) of adults undergoing NCD screening at LMIC health facilities or community campaigns. Each record contains 27 variables including demographics, anthropometry (BMI, waist circumference), lifestyle risk factors (smoking, alcohol, physical activity), family history, clinical measurements (blood pressure, fasting glucose… See the full description on the dataset page: https://huggingface.co/datasets/koushik1212/synthetic-diabetes-hypertension-NCD-screening-WHO-HEARTS.

sourceHugging Facecc-by-4.0updated 7mo agoView on Hugging Face

0likes18downloads

Dataset Card

Synthetic Diabetes & Hypertension NCD Screening Dataset (Adults 18–80)

Abstract

This dataset provides 30,000 synthetic records (10,000 per scenario) of adults undergoing NCD screening at LMIC health facilities or community campaigns. Each record contains 27 variables including demographics, anthropometry (BMI, waist circumference), lifestyle risk factors (smoking, alcohol, physical activity), family history, clinical measurements (blood pressure, fasting glucose, HbA1c, lipid profile), NCD classifications (diabetes/prediabetes, hypertension staging), 10-year CVD risk estimation, and treatment status. All distributions are parameterized from IDF Diabetes Atlas, NCD-RisC global analyses, WHO HEARTS package, and WHO STEPS surveys. Three burden scenarios span diabetes prevalence from 13% to 40% and hypertension from 40% to 60%.

This dataset is entirely synthetic. It must not be used for clinical decision-making.

2. Methodology

2.1 Epidemiological Parameterization

Parameter	Value	Source
Diabetes prevalence (adults)	6-18% by setting	IDF Diabetes Atlas 2021
Hypertension prevalence	25-40%	Zhou et al., Lancet 2021
Obesity (BMI ≥30)	10-40%	WHO STEPS surveys
Treatment gap (HTN)	~75% untreated	WHO 2021 Global Hypertension Report
Treatment gap (DM)	~70% untreated	IDF 2021

2.2 Scenario Design

Scenario	Diabetes	Prediabetes	HTN (any)	Obesity	CVD Risk ≥20%
Low burden	13.1%	28.7%	40.2%	7.7%	7.0%
Moderate burden	24.3%	37.2%	51.7%	23.1%	12.4%
High burden	40.1%	29.8%	60.0%	40.8%	19.1%

2.3 Classification Criteria

Classification	Criteria	Source
Prediabetes	FPG 100-125 mg/dL or HbA1c 5.7-6.4%	ADA 2023
Diabetes	FPG ≥126 mg/dL or HbA1c ≥6.5%	ADA 2023 / WHO
HTN Stage 1	SBP 140-159 or DBP 90-99 mmHg	WHO HEARTS
HTN Stage 2	SBP ≥160 or DBP ≥100 mmHg	WHO HEARTS
Central obesity	Waist ≥94cm (M) or ≥80cm (F)	IDF

3. Schema

Column	Type	Description
age_years	int	Age (18-80)
sex	categorical	M/F
bmi	float	Body mass index
bmi_category	categorical	underweight/normal/overweight/obese
waistcircumferencecm	float	Waist circumference
central_obesity	binary	IDF central obesity
smoking_status	categorical	never/former/current
alcohol_use	categorical	none/moderate/heavy
physical_activity	categorical	inactive/moderate/active
familyhistorydiabetes, familyhistorycvd	binary	Family history
systolicbpmmhg, diastolicbpmmhg	int	Blood pressure
hypertension_status	categorical	normal/stage1/stage2
fastingglucosemgdl	int	Fasting plasma glucose
hba1c_pct	float	Glycated haemoglobin
diabetes_status	categorical	normal/prediabetes/diabetes
totalcholesterolmgdl, hdlmgdl, ldlmgdl, triglycerides_mgdl	int	Lipid profile
cvdrisk10yr_pct	float	Simplified 10-year CVD risk
cvdriskcategory	categorical	low/moderate/high/very_high
onantihypertensive, onantidiabetic, on_statin	binary	Treatment status

4. Validation

5. Usage

python

from datasets import load_dataset
dataset = load_dataset("electricsheepafrica/synthetic-diabetes-hypertension-NCD-screening-WHO-HEARTS", "moderate_ncd_burden")
df = dataset["train"].to_pandas()

6. Limitations

—Synthetic: Not for clinical use or population-level estimates.
—Simplified CVD risk: Uses a simplified scoring; real WHO/ISH charts are sex/region/age stratified.
—No longitudinal data: Single screening snapshot; no disease progression.
—Treatment simplification: Binary treatment status; no medication classes or adherence data.
—No renal function: Creatinine/eGFR not modelled despite relevance to diabetes/HTN.

7. References

1.IDF (2021). IDF Diabetes Atlas, 10th edition.
2.WHO (2021). Global report on hypertension. Geneva.
3.NCD-RisC (2024). Worldwide trends in diabetes. Lancet, 404(10467):2077-2093.
4.Zhou B, et al. (2021). Worldwide trends in hypertension. Lancet, 398(10304):957-980.
5.WHO (2020). HEARTS Technical package. Geneva.
6.ADA (2023). Standards of Medical Care in Diabetes. Diabetes Care, 46(S1).
7.WHO STEPS Surveys. NCD risk factor surveillance, multiple countries.

Citation

bibtex

@dataset{esa_ncd_2025,
  title={Synthetic Diabetes and Hypertension NCD Screening Dataset},
  author={Electric Sheep Africa},
  year={2025},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/electricsheepafrica/synthetic-diabetes-hypertension-NCD-screening-WHO-HEARTS}
}

License

CC-BY-4.0