electricsheepafrica/ssa-breast-surgical-outcomes

SSA Breast Surgical Outcomes Dataset (Women, Multi-ancestry, Synthetic)

Dataset summary

This dataset provides a synthetic cohort of women undergoing primary breast cancer surgery across multiple ancestry groups, with emphasis on sub-Saharan Africa (SSA). It captures:

• Surgery type: breast-conserving surgery (BCS) versus mastectomy.
• Margin status: negative, close, or positive.
• Postoperative complication patterns: wound infection, seroma, hematoma, reoperation.

Parameters are qualitatively informed by population-based analyses (e.g., SEER/MSKCC) and surgical outcomes literature on BCS vs mastectomy, seroma formation, and positive margin rates, but all individuals and outcomes are fully synthetic.

Cohort design

Sample size and populations

• Total N: 10,000 synthetic women.

• Populations:

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

• Age: 18–85 years, with modest differences in mean age by population to approximate real-world breast cancer surgical cohorts (e.g., older in EUR/AAW).

Population labels are aligned with other Electric Sheep Africa synthetic datasets (e.g., stage, genomics, body composition) for multi-modal analyses.

Surgical variables

Surgery type

Variable:

• surgery_type – one of:
  • BCS – breast-conserving surgery.
  • mastectomy – total/modified radical mastectomy.

Surgery-type distributions are specified per population to reflect differences in access to radiotherapy, patient preferences, and resource levels:

• Higher BCS rates (~60–65%) in EUR/EAS/AAW.
• More mastectomy in some SSA settings, especially where radiotherapy access is limited or tumor burden is greater.

This structure is inspired by SEER-era estimates suggesting >60% BCS among eligible early-stage patients in high-income settings.

Margin status

Variable:

• margin_status – one of:
  • negative – no tumor at inked margin; adequate clearance.
  • close – tumor within a small distance from margin.
  • positive – tumor at inked margin.

Margin distributions differ by surgery type:

• BCS:

  • negative: ~80%
  • close: ~10%
  • positive: ~10%

  These values are qualitatively informed by systematic reviews of oncoplastic BCS, which report ~9–10% positive margins and an additional fraction of close margins.

• Mastectomy:

  • negative: ~93%
  • close: ~4%
  • positive: ~3%

Positive margins are less frequent after mastectomy but still possible (e.g., chest wall or skin involvement).

Postoperative complications

Variables (grade 2+ clinically relevant events):

• complication_wound_infection – surgical-site infection requiring treatment.
• complication_seroma – clinically significant seroma.
• complication_hematoma – hematoma requiring observation or intervention.
• complication_reoperation – unplanned return to OR for any surgical complication.
• any_complication – boolean flag (any of the above is True).

Complication probabilities are specified by surgery type, guided by reviews on seroma and wound complications:

• BCS:
  • Lower seroma and wound complication rates (e.g., seroma ~10%).
• Mastectomy (often with axillary surgery):
  • Higher seroma risk (mid-range ~30%) and modestly higher wound infection and reoperation rates.

Values are drawn from Bernoulli distributions per surgery type and complication type.

File and schema

surgical_outcomes_data.parquet / surgical_outcomes_data.csv

One row per woman-surgery:

• Demographics

  • sample_id
  • population, region, is_SSA, is_reference_panel
  • sex – Female
  • age

• Surgery

  • surgery_type – BCS or mastectomy.
  • margin_status – negative, close, positive.

• Complications

  • complication_wound_infection
  • complication_seroma
  • complication_hematoma
  • complication_reoperation
  • any_complication

Generation

The dataset is generated using:

• surgical_outcomes/scripts/generate_surgical_outcomes.py

with configuration in:

• surgical_outcomes/configs/surgical_outcomes_config.yaml

and literature curated in:

• surgical_outcomes/docs/LITERATURE_INVENTORY.csv

Key steps:

1. Sample generation – create a multi-ancestry breast cancer surgery cohort with age distribution.
2. Surgery type assignment – sample surgery_type per population according to configured BCS vs mastectomy probabilities.
3. Margin status assignment – sample margin_status per surgery type using distributions anchored in BCS margin literature.
4. Complication sampling – per surgery type, sample each complication as Bernoulli(p) and derive any_complication.

Validation

Validation is performed with:

• surgical_outcomes/scripts/validate_surgical_outcomes.py

and summarized in:

• surgical_outcomes/output/validation_report.md

Checks include:

• C01–C02 – Sample size and population counts vs config.
• C03 – Surgery-type distributions by population.
• C04 – Margin-status distributions by surgery type.
• C05 – Complication rates by surgery type.
• C06 – Missingness in key variables.

For the released version, the validator reports overall status PASS, indicating that modelled distributions match configured expectations within the specified tolerance.

Intended use

This dataset is intended for:

• Methods development in surgical outcomes modeling, quality metrics, and risk prediction.
• Educational use for demonstrating:
  • BCS vs mastectomy utilization patterns.
  • Margin status and re-excision risk proxies.
  • Common postoperative complication rates.

It is not suitable for:

• Clinical decision-making or individual risk counselling.
• Estimating real-world rates in any specific hospital or country.

All records and outcomes are synthetic.

Ethical considerations

• No real patient data were used.
• Population labels and surgery patterns are for simulation and methodological realism only.
• Analyses should be interpreted as methodological demonstrations, not as direct reflections of any specific health system.

License

• License: CC BY-NC 4.0.
• Free to use for non-commercial research, education, and methods development with attribution.

Citation

If you use this dataset, please cite:

Electric Sheep Africa. "SSA Breast Surgical Outcomes Dataset (Women, Multi-ancestry, Synthetic)." Hugging Face Datasets.

and, as appropriate, relevant breast-surgery outcomes literature (e.g., SEER/MSKCC analyses of BCS vs mastectomy rates, systematic reviews of margin positivity and seroma/complication rates after breast surgery).