electricsheepafrica/ssa-cardiovascular-metrics

SSA Cardiovascular Metrics Dataset (Multi-ancestry, Synthetic)

Dataset summary

This dataset provides a synthetic cardiovascular metrics cohort of 10,000 adults across multiple ancestry groups with a focus on sub-Saharan Africa (SSA). It includes:

• Blood pressure profiles – systolic/diastolic BP and hypertension categories.
• Left ventricular (LV) structure – LV mass index, septal wall thickness, LV hypertrophy (LVH) flag.
• ECG patterns – normal, LVH pattern, T-wave inversions, ST-segment elevation.

The design is informed by echocardiography reference values, SSA hypertension studies, and ECG morphology reviews, but all individuals and measurements are fully synthetic and non-identifiable.

Cohort design

Sample size and populations

• Total N: 10,000 synthetic adults.

• Populations:

  • SSA_West: 2,000
  • SSA_East: 2,000
  • SSA_Central: 1,500
  • SSA_Southern: 1,500
  • AAW (African American, admixed): 1,500
  • EUR (European reference): 1,000
  • EAS (East Asian reference): 500

• Sex distribution:

  • Male: ~45%
  • Female: ~55%

• Age range: 18–80 years, with population-specific means and standard deviations tuned to resemble adult cardiovascular cohorts.

These population labels align with other Electric Sheep Africa datasets (SNP arrays, structural variation, body composition, pharmacogenomics) for multi-modal method development.

Cardiovascular metrics

Blood pressure

Variables:

• sbp – systolic blood pressure (mmHg).
• dbp – diastolic blood pressure (mmHg).
• bp_category – categorical BP class:
  • normal
  • elevated
  • stage1 hypertension
  • stage2 hypertension

Category prevalences are set per population, with higher hypertension burden in SSA and AAW than in EUR/EAS, reflecting community-based SSA studies that report 30–40% hypertension prevalence in adults.

Target SBP/DBP means and SDs by category are loosely based on guideline thresholds and SSA BP profiles, for example:

• normal: ~115/75 mmHg.
• elevated: ~125/80 mmHg.
• stage1: ~140/90 mmHg.
• stage2: ~160/100 mmHg.

Sampling enforces category-specific thresholds so BP values remain consistent with their labels.

Left ventricular structure

Variables:

• lv_mass_index_g_m2 – LV mass indexed to body surface area (g/m²).
• septal_thickness_mm – interventricular septal wall thickness (mm).
• lvh_present – boolean flag indicating LV hypertrophy.

Design anchors:

• Normal LV mass index around:
  • ~70 g/m² in men.
  • ~61 g/m² in women. (from 3D echocardiography reference values).
• LV mass index increases with higher BP categories (normal → elevated → stage1 → stage2).
• LVH thresholds approximate ASE/EACVI cutoffs:
  • Men: LVMI ≥ 115 g/m².
  • Women: LVMI ≥ 95 g/m².

Septal thickness is modeled with means increasing from ~8 mm in normotensive to ~12 mm in stage 2 hypertension, with bounds ensuring plausible values (~6–18 mm).

ECG patterns

Variables:

• ecg_pattern – categorical:
  • normal
  • LVH (voltage/strain-like pattern)
  • T_wave_inversion
  • ST_elevation

ECG patterns are tied to BP category:

• Normotensive individuals have predominantly normal ECGs with a small fraction of LVH/T-wave inversions and very rare ST elevation.
• Stage 1 and stage 2 hypertensives have progressively higher LVH and T-wave inversion fractions, and modestly higher ST_elevation prevalence, reflecting higher probability of structural heart disease and ischemia.

Design is guided qualitatively by ECG morphology reviews (e.g., StatPearls on T waves and STEMI) that emphasize:

• T-wave inversions can be benign or ischemic, but are relatively uncommon in healthy adults.
• ST-elevation patterns consistent with STEMI are rare in the general population but critical when present.

File and schema

cardiovascular_metrics_data.parquet / cardiovascular_metrics_data.csv

One row per synthetic individual, with:

• Demographics / ancestry

  • sample_id
  • population – SSA_West, SSA_East, SSA_Central, SSA_Southern, AAW, EUR, EAS.
  • region – SSA subregion or Non_SSA.
  • is_SSA – boolean.
  • is_reference_panel – True for AAW/EUR/EAS.
  • sex – Male or Female.
  • age – years (18–80).

• Blood pressure

  • sbp – systolic BP (mmHg).
  • dbp – diastolic BP (mmHg).
  • bp_category – BP class as above.

• LV structure

  • lv_mass_index_g_m2 – LV mass index.
  • septal_thickness_mm – septal thickness.
  • lvh_present – LVH flag.

• ECG

  • ecg_pattern – normal, LVH, T_wave_inversion, or ST_elevation.

Generation

The dataset is generated using:

• cardiovascular_metrics/scripts/generate_cardiovascular_metrics.py

with configuration in:

• cardiovascular_metrics/configs/cardiovascular_metrics_config.yaml

and literature curated in:

• cardiovascular_metrics/docs/LITERATURE_INVENTORY.csv

Key modeling steps:

1. Sample table – ages and sexes per population drawn from truncated normal distributions.
2. Blood pressure assignment – BP category sampled using population-specific prevalences, then SBP/DBP drawn from category-specific distributions and constrained by threshold ranges.
3. LV structure – LV mass index and septal thickness drawn by sex and BP category; LVH flag set using sex-specific LVMI thresholds.
4. ECG pattern – chosen stochastically from per-BP category pattern distributions, increasing LVH/ischemic patterns with higher BP stages.

Validation

Validation follows the GENOMICS Synthetic Data Playbook and is implemented in:

• cardiovascular_metrics/scripts/validate_cardiovascular_metrics.py

Major checks include:

• C01–C02 – Sample size and population counts
  • Confirm N = 10,000 and population counts close to configuration (±10% tolerance).
• C03 – BP category distributions by population
  • Compare observed BP category proportions to configured targets for each population.
• C04 – BP values vs thresholds
  • Quantify the fraction of individuals whose SBP/DBP lie outside the configured bounds for their BP category; require this to be very low.
• C05 – LV mass index means by sex and BP
  • Check that LVMI means by sex and BP category align with configuration.
• C06 – ECG pattern distributions by BP
  • Validate that pattern frequencies per BP category match configured expectations.
• C07 – Missingness in key variables
  • Ensure negligible missingness across demographics, BP, LV metrics, and ECG pattern.

The validator writes a Markdown report:

• cardiovascular_metrics/output/validation_report.md

For the released version, all checks complete with an overall status of PASS.

Intended use

This dataset is intended for:

• Methods development in cardiovascular risk modeling, echocardiography/ECG analytics, and multi-ancestry BP profiling.
• Teaching and demonstration of:
  • Hypertension staging distributions across populations.
  • Relationships between BP, LV mass, and LVH.
  • ECG pattern variation with cardiovascular risk.

It is not suitable for:

• Clinical decision-making or patient care.
• Deriving real-world incidence/prevalence estimates.
• Individual-level inference.

All data are synthetic and non-identifiable.

Ethical considerations

• No real patient data are used.
• Population labels are for simulation realism and are not tied to specific countries or real-world cohorts.
• Analyses should be interpreted as methodological rather than epidemiological statements.

License

• License: CC BY-NC 4.0.
• Free for non-commercial research, method development, and teaching with attribution.

Citation

If you use this dataset, please cite:

Electric Sheep Africa. "SSA Cardiovascular Metrics Dataset (Multi-ancestry, Synthetic)." Hugging Face Datasets.

and consider citing relevant underlying cardiovascular and echocardiography literature used to guide the design (e.g., ASE/EACVI chamber quantification guidelines, LV mass index reference studies, SSA hypertension burden papers, and ECG morphology reviews).