Upload folder using huggingface_hub

README.md

@@ -0,0 +1,138 @@
+---
+license: cc-by-4.0
+task_categories:
+  - tabular-classification
+language:
+  - en
+tags:
+  - healthcare
+  - obstetric-fistula
+  - vesicovaginal-fistula
+  - obstructed-labour
+  - maternal-health
+  - sub-saharan-africa
+  - lmic
+pretty_name: "Obstetric Fistula (VVF/RVF, Repair, Social Consequences, Outcomes)"
+size_categories:
+  - 10K<n<100K
+configs:
+  - config_name: specialized_fistula_centre
+    data_files: data/fistula_specialized_fistula_centre.csv
+  - config_name: district_hospital
+    data_files: data/fistula_district_hospital.csv
+    default: true
+  - config_name: no_surgical_access
+    data_files: data/fistula_no_surgical_access.csv
+---
+
+# Obstetric Fistula Dataset
+
+## Abstract
+
+This dataset provides **30,000 simulated obstetric fistula patient records** (10,000 per scenario) from sub-Saharan Africa. Each record contains 50+ variables including demographics, obstetric history, fistula characteristics, social consequences, surgical repair, rehabilitation, and outcomes. Three settings: specialized fistula centre (76% repair success), district hospital (59%), and no surgical access (0%).
+
+## 1. Introduction
+
+Obstetric fistula — a hole between the birth canal and bladder/rectum — is caused by prolonged, obstructed labour without timely EmONC. SSA has 50% of the global burden with 50,000-100,000 new cases annually (UNFPA 2024). Consequences include urinary/faecal incontinence, social isolation, divorce, depression, and economic devastation. Repair success is 85-90% at specialized centres, but most women lack access. Child marriage and short stature are key risk factors.
+
+**This dataset is entirely simulated. It must not be used for clinical decision-making.**
+
+## 2. Methodology
+
+### 2.1 Parameterization
+
+| Parameter | Value | Source |
+| --- | --- | --- |
+| SSA share of global burden | 50% | UNFPA 2024 |
+| VVF type | 80% | UNFPA |
+| Stillbirth rate | ~70% | PubMed 2018 |
+| Repair success (specialist) | 85-90% | PMC 2024 |
+| Divorce/separation | ~55% | UNFPA 2024 |
+| Depression | ~60% | UNFPA |
+| Child marriage risk | Major factor | PubMed 1999 |
+
+### 2.2 Scenario Design
+
+| Scenario | Repair | Specialist | Physio | Success |
+| --- | --- | --- | --- | --- |
+| Specialized centre | Yes | Yes | Yes | 76% |
+| District hospital | Yes | No | No | 59% |
+| No surgical access | No | No | No | 0% |
+
+## 3. Schema
+
+| Column | Type | Description |
+| --- | --- | --- |
+| id | int | Unique identifier |
+| age_at_presentation | int | Age at presentation |
+| age_at_fistula_onset | int | Age when fistula developed |
+| years_living_with_fistula | int | Duration before treatment |
+| child_marriage | binary | Married <18 |
+| fistula_type | categorical | vesicovaginal / rectovaginal / combined |
+| fistula_size | categorical | small / medium / large / extensive |
+| obstructed_labour | binary | Obstructed labour |
+| stillbirth | binary | Stillbirth in index pregnancy |
+| divorced_separated | binary | Divorced/separated |
+| social_isolation | binary | Social isolation |
+| depression | binary | Depression |
+| repair_performed | binary | Surgical repair done |
+| repair_successful | binary | Repair successful |
+| residual_incontinence | binary | Residual incontinence |
+| continence_at_3_months | binary | Continent at 3 months |
+| returned_to_community | binary | Community reintegration |
+
+## 4. Validation
+
+<p align="center">
+  <img src="validation_report.png" alt="Validation Report" width="100%">
+</p>
+
+Key validation checks:
+
+- **VVF**: 80% of fistula types ✓
+- **Stillbirth**: ~70% ✓
+- **Repair success gradient**: 76% → 59% → 0% ✓
+- **Social consequences**: >50% divorced, >60% depressed ✓
+- **Obstructed labour**: Major cause ✓
+- **Success decreases with fistula size** ✓
+
+## 5. Usage
+
+```python
+from datasets import load_dataset
+dataset = load_dataset("electricsheepafrica/obstetric-fistula", "district_hospital")
+df = dataset["train"].to_pandas()
+```
+
+## 6. Limitations
+
+- **Simulated**: Not from real fistula registries.
+- **No urodynamic data**: No bladder pressure studies.
+- **No imaging**: No fistulogram data.
+- **Simplified**: No detailed surgical notes.
+- **Cross-sectional**: No long-term follow-up trajectory.
+
+## 7. References
+
+1. UNFPA (2024). Obstetric fistula factsheet.
+2. UNFPA WCARO (2024). Fistula burden in West/Central Africa.
+3. PMC (2024). Fistula repair failure in SSA.
+4. PubMed (1999). Sociomedical risk factors for VVF.
+5. PubMed (2018). VVF in Uganda — Kitovu Hospital.
+6. PubMed (2024). Systematic review repair outcomes LMICs.
+
+## Citation
+
+```bibtex
+@dataset{esa_obstetric_fistula_2025,
+  title={Obstetric Fistula Dataset},
+  author={Electric Sheep Africa},
+  year={2025},
+  publisher={Hugging Face},
+  url={https://huggingface.co/datasets/electricsheepafrica/obstetric-fistula}
+}
+```
+
+## License
+
+[CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/)

generate_dataset.py

@@ -0,0 +1,252 @@
+#!/usr/bin/env python3
+"""
+Literature-Informed Obstetric Fistula Dataset
+==============================================
+
+Generates realistic synthetic records of obstetric fistula patients
+in sub-Saharan Africa, including risk factors, fistula characteristics,
+social consequences, surgical repair, and outcomes.
+
+References (web-searched):
+-----------
+[1]  UNFPA 2024. Obstetric fistula caused by prolonged
+     obstructed labour without timely EmONC. SSA has 50%
+     of global burden. 50K-100K new cases/year.
+[2]  UNFPA WCARO 2024. West/Central Africa highest rates.
+     Social isolation, divorce, depression common.
+[3]  PMC 2024. Fistula repair failure in SSA. Repair
+     success 85-90% at specialized centres.
+[4]  PubMed 1999. Early marriage, young adolescents,
+     premature pregnancy → VVF. Sociomedical risk factors.
+[5]  PubMed 2018. Kitovu Hospital Uganda, highest fistula
+     rates globally. Incidence and causative factors.
+[6]  PubMed 2024. Systematic review repair outcomes.
+     Proportions of successful repair in LMICs.
+"""
+
+import numpy as np
+import pandas as pd
+import argparse
+import os
+
+SCENARIOS = {
+    'specialized_fistula_centre': {
+        'description': 'Dedicated fistula repair centre with '
+                       'trained surgeon, physiotherapy, social '
+                       'support (e.g., Addis Ababa Fistula Hospital, '
+                       'Kitovu Uganda)',
+        'repair_available': True,
+        'specialist_surgeon': True,
+        'physiotherapy': True,
+        'social_support': True,
+        'repair_success': 0.90,
+    },
+    'district_hospital': {
+        'description': 'District hospital with visiting fistula '
+                       'surgeon, basic repair capability, limited '
+                       'follow-up (e.g., district hospitals DRC, '
+                       'Malawi, Tanzania)',
+        'repair_available': True,
+        'specialist_surgeon': False,
+        'physiotherapy': False,
+        'social_support': False,
+        'repair_success': 0.70,
+    },
+    'no_surgical_access': {
+        'description': 'Rural community with no surgical access, '
+                       'traditional remedies, referral barriers '
+                       '(e.g., rural Niger, Chad, South Sudan)',
+        'repair_available': False,
+        'specialist_surgeon': False,
+        'physiotherapy': False,
+        'social_support': False,
+        'repair_success': 0.0,
+    },
+}
+
+
+def generate_dataset(n=10000, seed=42, scenario='district_hospital'):
+    rng = np.random.default_rng(seed)
+    sc = SCENARIOS[scenario]
+
+    records = []
+
+    for idx in range(n):
+        rec = {'id': idx + 1}
+
+        # ── 1. Demographics ──
+        rec['age_at_presentation'] = max(14, min(65, int(rng.normal(28, 8))))
+        rec['age_at_fistula_onset'] = max(13, min(rec['age_at_presentation'],
+            int(rng.normal(20, 5))))
+        rec['years_living_with_fistula'] = max(0,
+            rec['age_at_presentation'] - rec['age_at_fistula_onset'])
+        rec['age_at_marriage'] = max(10, min(30, int(rng.normal(16, 3))))
+        rec['child_marriage'] = 1 if rec['age_at_marriage'] < 18 else 0
+        rec['age_at_first_pregnancy'] = max(rec['age_at_marriage'],
+            min(35, int(rng.normal(17, 3))))
+        rec['education'] = rng.choice(
+            ['none', 'primary', 'secondary', 'tertiary'],
+            p=[0.45, 0.35, 0.15, 0.05])
+        rec['rural'] = 1 if rng.random() < 0.80 else 0
+        rec['parity'] = max(0, min(12, int(rng.exponential(2.5))))
+        rec['height_cm'] = max(140, min(180, int(rng.normal(157, 6))))
+        rec['short_stature'] = 1 if rec['height_cm'] < 150 else 0
+        rec['bmi'] = round(max(14, min(35, rng.normal(20, 3))), 1)
+
+        # ── 2. Obstetric history ──
+        rec['labour_duration_hours'] = max(6, min(120, int(rng.exponential(24) + 12)))
+        rec['prolonged_labour'] = 1 if rec['labour_duration_hours'] > 24 else 0
+        rec['obstructed_labour'] = 1 if rec['labour_duration_hours'] > 18 and rng.random() < 0.85 else 0
+        rec['caesarean_performed'] = 0
+        if rec['obstructed_labour']:
+            rec['caesarean_performed'] = 1 if rng.random() < 0.15 else 0
+        rec['skilled_birth_attendant'] = 1 if rng.random() < 0.20 else 0
+        rec['place_of_delivery'] = rng.choice(
+            ['home', 'health_centre', 'hospital'],
+            p=[0.60, 0.25, 0.15])
+        rec['stillbirth'] = 1 if rng.random() < 0.70 else 0
+        rec['neonatal_death'] = 0
+        if not rec['stillbirth']:
+            rec['neonatal_death'] = 1 if rng.random() < 0.15 else 0
+
+        # ── 3. Fistula characteristics ──
+        rec['fistula_type'] = rng.choice(
+            ['vesicovaginal', 'rectovaginal', 'combined'],
+            p=[0.80, 0.10, 0.10])
+        rec['fistula_size'] = rng.choice(
+            ['small', 'medium', 'large', 'extensive'],
+            p=[0.20, 0.35, 0.30, 0.15])
+        rec['urethral_involvement'] = 1 if rng.random() < 0.25 else 0
+        rec['circumferential_defect'] = 1 if rec['fistula_size'] == 'extensive' and rng.random() < 0.40 else 0
+        rec['scarring_severity'] = rng.choice(
+            ['mild', 'moderate', 'severe'],
+            p=[0.25, 0.40, 0.35])
+        rec['vaginal_stenosis'] = 1 if rec['scarring_severity'] == 'severe' and rng.random() < 0.40 else 0
+        rec['foot_drop'] = 1 if rec['prolonged_labour'] and rng.random() < 0.10 else 0
+
+        # ── 4. Social consequences [1][2] ──
+        rec['divorced_separated'] = 1 if rng.random() < 0.55 else 0
+        rec['social_isolation'] = 1 if rng.random() < 0.65 else 0
+        rec['depression'] = 1 if rng.random() < 0.60 else 0
+        rec['anxiety'] = 1 if rng.random() < 0.45 else 0
+        rec['economic_impact'] = rng.choice(
+            ['none', 'mild', 'severe'],
+            p=[0.10, 0.30, 0.60])
+        rec['unable_to_work'] = 1 if rng.random() < 0.50 else 0
+        rec['skin_excoriation'] = 1 if rng.random() < 0.70 else 0
+        rec['recurrent_uti'] = 1 if rng.random() < 0.50 else 0
+        rec['malodour'] = 1 if rng.random() < 0.80 else 0
+
+        # ── 5. Care seeking ──
+        rec['delay_to_presentation_years'] = rec['years_living_with_fistula']
+        rec['previous_repair_attempts'] = max(0, min(5, int(rng.exponential(0.5))))
+        rec['traditional_remedy_tried'] = 1 if rng.random() < 0.35 else 0
+        rec['barrier_to_care'] = rng.choice(
+            ['cost', 'distance', 'awareness', 'stigma', 'no_service', 'none'],
+            p=[0.20, 0.20, 0.15, 0.15, 0.20, 0.10])
+        rec['referred_by'] = rng.choice(
+            ['self', 'ngo', 'health_worker', 'community', 'media'],
+            p=[0.30, 0.25, 0.20, 0.15, 0.10])
+
+        # ── 6. Surgical repair [3][6] ──
+        rec['repair_performed'] = 0
+        if sc['repair_available']:
+            rec['repair_performed'] = 1 if rng.random() < 0.85 else 0
+
+        rec['repair_technique'] = 'none'
+        if rec['repair_performed']:
+            rec['repair_technique'] = rng.choice(
+                ['transvaginal', 'transabdominal', 'combined'],
+                p=[0.75, 0.15, 0.10])
+
+        rec['graft_used'] = 0
+        if rec['repair_performed'] and rec['fistula_size'] in ('large', 'extensive'):
+            rec['graft_used'] = 1 if rng.random() < 0.20 else 0
+
+        rec['catheter_days'] = 0
+        if rec['repair_performed']:
+            rec['catheter_days'] = max(7, min(28, int(rng.normal(14, 3))))
+
+        rec['repair_successful'] = 0
+        if rec['repair_performed']:
+            success_prob = sc['repair_success']
+            if rec['fistula_size'] == 'extensive':
+                success_prob *= 0.60
+            elif rec['fistula_size'] == 'large':
+                success_prob *= 0.80
+            if rec['previous_repair_attempts'] > 0:
+                success_prob *= 0.85
+            if rec['circumferential_defect']:
+                success_prob *= 0.50
+            rec['repair_successful'] = 1 if rng.random() < success_prob else 0
+
+        rec['residual_incontinence'] = 0
+        if rec['repair_successful']:
+            rec['residual_incontinence'] = 1 if rng.random() < 0.15 else 0
+
+        rec['complication_post_repair'] = 0
+        if rec['repair_performed']:
+            rec['complication_post_repair'] = 1 if rng.random() < 0.10 else 0
+
+        # ── 7. Rehabilitation ──
+        rec['physiotherapy_received'] = 0
+        if rec['repair_performed'] and sc['physiotherapy']:
+            rec['physiotherapy_received'] = 1 if rng.random() < 0.70 else 0
+
+        rec['social_reintegration_support'] = 0
+        if rec['repair_performed'] and sc['social_support']:
+            rec['social_reintegration_support'] = 1 if rng.random() < 0.60 else 0
+
+        rec['livelihood_support'] = 0
+        if rec['social_reintegration_support']:
+            rec['livelihood_support'] = 1 if rng.random() < 0.40 else 0
+
+        rec['continence_at_3_months'] = 0
+        if rec['repair_successful'] and not rec['residual_incontinence']:
+            rec['continence_at_3_months'] = 1 if rng.random() < 0.90 else 0
+
+        rec['returned_to_community'] = 0
+        if rec['repair_successful']:
+            rec['returned_to_community'] = 1 if rng.random() < 0.75 else 0
+
+        records.append(rec)
+
+    df = pd.DataFrame(records)
+
+    print(f"\n{'='*65}")
+    print(f"Obstetric Fistula — {scenario} (n={n}, seed={seed})")
+    print(f"{'='*65}")
+    print(f"\n  Repair performed: {df['repair_performed'].mean()*100:.1f}%")
+    print(f"  Repair success: {df[df['repair_performed']==1]['repair_successful'].mean()*100:.1f}%" if df['repair_performed'].sum() > 0 else "  No repairs")
+    print(f"  Stillbirth: {df['stillbirth'].mean()*100:.1f}%")
+    print(f"  Divorced: {df['divorced_separated'].mean()*100:.1f}%")
+    print(f"  Depression: {df['depression'].mean()*100:.1f}%")
+    print(f"  Mean years with fistula: {df['years_living_with_fistula'].mean():.1f}")
+
+    return df
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(
+        description='Generate obstetric fistula dataset')
+    parser.add_argument('--scenario', type=str, default='district_hospital',
+                        choices=list(SCENARIOS.keys()))
+    parser.add_argument('--n', type=int, default=10000)
+    parser.add_argument('--seed', type=int, default=42)
+    parser.add_argument('--output', type=str, default=None)
+    parser.add_argument('--all-scenarios', action='store_true')
+    args = parser.parse_args()
+
+    os.makedirs('data', exist_ok=True)
+
+    if args.all_scenarios:
+        for sc_name in SCENARIOS:
+            df = generate_dataset(n=args.n, seed=args.seed, scenario=sc_name)
+            out = os.path.join('data', f'fistula_{sc_name}.csv')
+            df.to_csv(out, index=False)
+            print(f"  -> Saved to {out}\n")
+    else:
+        df = generate_dataset(n=args.n, seed=args.seed, scenario=args.scenario)
+        out = args.output or os.path.join('data', f'fistula_{args.scenario}.csv')
+        df.to_csv(out, index=False)
+        print(f"  -> Saved to {out}")

requirements.txt

@@ -0,0 +1,3 @@
+numpy>=1.24
+pandas>=2.0
+matplotlib>=3.7

validate_dataset.py

@@ -0,0 +1,132 @@
+#!/usr/bin/env python3
+"""Validation & Diagnostic Visualization for Obstetric Fistula Dataset."""
+
+import pandas as pd
+import numpy as np
+import matplotlib.pyplot as plt
+import os
+
+SCENARIOS = ['specialized_fistula_centre', 'district_hospital', 'no_surgical_access']
+
+
+def load_scenarios(data_dir='data'):
+    dfs = {}
+    for sc in SCENARIOS:
+        path = os.path.join(data_dir, f'fistula_{sc}.csv')
+        if os.path.exists(path):
+            dfs[sc] = pd.read_csv(path)
+    return dfs
+
+
+def make_report(dfs, output='validation_report.png'):
+    fig, axes = plt.subplots(4, 2, figsize=(16, 22))
+    fig.suptitle('Obstetric Fistula — Validation Report',
+                 fontsize=16, fontweight='bold', y=0.98)
+    df = dfs.get('district_hospital', list(dfs.values())[0])
+    colors = ['#2ecc71', '#f39c12', '#e74c3c']
+
+    ax = axes[0, 0]
+    x = np.arange(len(SCENARIOS))
+    repair = [dfs[sc]['repair_performed'].mean()*100 for sc in SCENARIOS if sc in dfs]
+    success = []
+    for sc in SCENARIOS:
+        if sc in dfs:
+            d = dfs[sc]
+            rep = d[d['repair_performed'] == 1]
+            success.append(rep['repair_successful'].mean()*100 if len(rep) > 0 else 0)
+    w = 0.3
+    ax.bar(x - w/2, repair, w, label='Repair Performed', color='#3498db', alpha=0.8)
+    ax.bar(x + w/2, success, w, label='Repair Success', color='#2ecc71', alpha=0.8)
+    ax.set_xticks(x)
+    ax.set_xticklabels(['Specialist', 'District', 'No Access'], fontsize=9)
+    ax.set_ylabel('Rate (%)')
+    ax.set_title('Repair Access & Success (85-90% at specialist)')
+    ax.legend(fontsize=8)
+
+    ax = axes[0, 1]
+    ft = df['fistula_type'].value_counts()
+    c_colors = ['#e74c3c', '#f39c12', '#9b59b6']
+    ax.pie(ft.values,
+           labels=[s.replace('_', ' ').upper() for s in ft.index],
+           autopct='%1.0f%%', colors=c_colors[:len(ft)],
+           startangle=90, textprops={'fontsize': 9})
+    ax.set_title('Fistula Type (VVF = 80%)')
+
+    ax = axes[1, 0]
+    social = ['divorced_separated', 'social_isolation', 'depression',
+              'unable_to_work', 'skin_excoriation', 'malodour']
+    s_labels = ['Divorced', 'Isolated', 'Depression', 'Unable Work', 'Excoriation', 'Malodour']
+    vals = [df[s].mean()*100 for s in social]
+    ax.barh(range(6), vals, color='#9b59b6', alpha=0.7)
+    ax.set_yticks(range(6))
+    ax.set_yticklabels(s_labels, fontsize=9)
+    for i, v in enumerate(vals):
+        ax.text(v + 0.5, i, f'{v:.0f}%', va='center', fontsize=9)
+    ax.set_xlabel('Prevalence (%)')
+    ax.set_title('Social Consequences (devastating)')
+
+    ax = axes[1, 1]
+    ax.hist(df['years_living_with_fistula'], bins=20, color='#e74c3c',
+            alpha=0.7, edgecolor='white')
+    ax.set_xlabel('Years Living with Fistula')
+    ax.set_title('Duration Before Presentation')
+
+    ax = axes[2, 0]
+    risks = ['child_marriage', 'short_stature', 'prolonged_labour',
+             'obstructed_labour', 'stillbirth']
+    r_labels = ['Child Marriage', 'Short Stature', 'Prolonged Labour',
+                'Obstructed Labour', 'Stillbirth']
+    vals = [df[r].mean()*100 for r in risks]
+    ax.bar(range(5), vals, color=['#e74c3c', '#f39c12', '#9b59b6', '#3498db', '#e67e22'], alpha=0.8)
+    ax.set_xticks(range(5))
+    ax.set_xticklabels(r_labels, fontsize=7, rotation=15)
+    for i, v in enumerate(vals):
+        ax.text(i, v + 0.5, f'{v:.0f}%', ha='center', fontsize=8)
+    ax.set_ylabel('Prevalence (%)')
+    ax.set_title('Risk Factors (obstructed labour = key cause)')
+
+    ax = axes[2, 1]
+    barriers = df[df['barrier_to_care'] != 'none']['barrier_to_care'].value_counts()
+    if len(barriers) > 0:
+        ax.barh(range(len(barriers)), barriers.values, color='#3498db', alpha=0.8)
+        ax.set_yticks(range(len(barriers)))
+        ax.set_yticklabels([s.replace('_', ' ').title() for s in barriers.index], fontsize=9)
+    ax.set_xlabel('Count')
+    ax.set_title('Barriers to Care')
+
+    ax = axes[3, 0]
+    sizes = ['small', 'medium', 'large', 'extensive']
+    rep_d = df[df['repair_performed'] == 1]
+    if len(rep_d) > 0:
+        size_success = []
+        for sz in sizes:
+            sz_d = rep_d[rep_d['fistula_size'] == sz]
+            size_success.append(sz_d['repair_successful'].mean()*100 if len(sz_d) > 0 else 0)
+        ax.bar(range(4), size_success,
+               color=['#2ecc71', '#f39c12', '#e74c3c', '#9b59b6'], alpha=0.8)
+        ax.set_xticks(range(4))
+        ax.set_xticklabels(sizes, fontsize=9)
+        for i, v in enumerate(size_success):
+            ax.text(i, v + 1, f'{v:.0f}%', ha='center', fontsize=9)
+    ax.set_ylabel('Success Rate (%)')
+    ax.set_title('Repair Success by Fistula Size')
+
+    ax = axes[3, 1]
+    ax.hist(df['age_at_fistula_onset'], bins=20, color='#f39c12',
+            alpha=0.7, edgecolor='white', label='Onset')
+    ax.hist(df['age_at_presentation'], bins=20, color='#3498db',
+            alpha=0.5, edgecolor='white', label='Presentation')
+    ax.set_xlabel('Age (years)')
+    ax.set_title('Age at Onset vs Presentation')
+    ax.legend(fontsize=8)
+
+    plt.tight_layout(rect=[0, 0, 1, 0.97])
+    plt.savefig(output, dpi=150, bbox_inches='tight')
+    print(f'Saved validation report to {output}')
+    plt.close()
+
+
+if __name__ == '__main__':
+    dfs = load_scenarios()
+    if dfs:
+        make_report(dfs)