Upload folder using huggingface_hub

README.md

@@ -0,0 +1,171 @@
+---
+license: cc-by-4.0
+task_categories:
+  - tabular-classification
+language:
+  - en
+tags:
+  - healthcare
+  - trauma
+  - road-traffic-injury
+  - emergency
+  - prehospital
+  - GCS
+  - ISS
+  - TBI
+  - sub-saharan-africa
+  - lmic
+pretty_name: "Road Traffic Injury & Trauma (GCS, ISS, Prehospital, Emergency Care)"
+size_categories:
+  - 10K<n<100K
+configs:
+  - config_name: trauma_centre
+    data_files: data/rti_trauma_centre.csv
+  - config_name: district_hospital
+    data_files: data/rti_district_hospital.csv
+    default: true
+  - config_name: rural_health_centre
+    data_files: data/rti_rural_health_centre.csv
+---
+
+# Road Traffic Injury & Trauma Dataset
+
+## Abstract
+
+This dataset provides **30,000 simulated RTI records** (10,000 per scenario) of road traffic injury patients presenting to health facilities in sub-Saharan Africa. Each record contains 50+ variables including road user type, injury mechanism, GCS, ISS, body region, prehospital transport, emergency management (CT, surgery, ICU, blood), complications, disability, and mortality. Three settings: urban trauma centre (18.6% mortality), district hospital (33.8%), and rural health centre (48.2%).
+
+## 1. Introduction
+
+Africa has the highest road traffic death rate globally at 26.6 per 100,000 population versus 17.4 globally (WHO Global Status Report 2023). Pedestrians and cyclists account for >50% of road deaths in Africa. Males aged 15-44 are most affected, with a 3:1 male-to-female ratio. Less than 10% of trauma patients arrive within the golden hour in many SSA settings. Strengthening prehospital trauma care could prevent 54% of trauma deaths (Mock et al., Bull WHO 2012).
+
+**This dataset is entirely simulated. It must not be used for clinical decision-making.**
+
+## 2. Methodology
+
+### 2.1 Parameterization
+
+| Parameter | Value | Source |
+| --- | --- | --- |
+| RTI death rate (Africa) | 26.6/100K | WHO Global Status Report 2023 |
+| Pedestrian proportion | >50% of deaths | WHO 2023 |
+| Male:Female ratio | 3:1 | WHO 2024 |
+| Golden hour arrival (SSA) | <10% | Galvagno et al., 2019 |
+| Severe injury mortality (SSA) | 30-40% (ISS>15) | Zafar et al., Lancet 2018 |
+| Head injury proportion | 44% | Chalya et al., BMC Pub Health 2012 |
+| Prehospital prevention potential | 54% deaths preventable | Mock et al., Bull WHO 2012 |
+
+### 2.2 Scenario Design
+
+| Scenario | Facility | CT | ICU | Ambulance | Mortality |
+| --- | --- | --- | --- | --- | --- |
+| Trauma centre | Urban, neurosurgery | Yes | Yes | 38% | 18.6% |
+| District hospital | Basic X-ray | No | No | 10% | 33.8% |
+| Rural health centre | No imaging | No | No | 3% | 48.2% |
+
+## 3. Schema
+
+| Column | Type | Description |
+| --- | --- | --- |
+| id | int | Unique identifier |
+| age_years | int | Patient age |
+| sex | categorical | M / F |
+| road_user_type | categorical | pedestrian / motorcyclist / motor_vehicle_occupant / cyclist / passenger_minibus |
+| helmet_use | binary | Helmet worn (motorcyclist) |
+| seatbelt_use | binary | Seatbelt worn (vehicle occupant) |
+| alcohol_involved | binary | Alcohol involved |
+| time_of_day | categorical | morning / afternoon / evening / night |
+| road_type | categorical | highway / urban_road / rural_road / intersection |
+| primary_body_region | categorical | head_neck / chest / abdomen / upper/lower_extremity / spine / pelvis / multiple |
+| tbi | binary | Traumatic brain injury |
+| gcs | int | Glasgow Coma Scale (3-15) |
+| gcs_category | categorical | severe / moderate / mild |
+| iss | int | Injury Severity Score (1-75) |
+| iss_category | categorical | minor / moderate / severe / critical |
+| fracture | binary | Fracture present |
+| open_fracture | binary | Open fracture |
+| internal_bleeding | binary | Internal bleeding |
+| spinal_cord_injury | binary | Spinal cord injury |
+| transport_mode | categorical | ambulance / private_vehicle / police / bystander / walked / motorcycle_taxi |
+| time_to_facility_hours | float | Time from injury to facility |
+| within_golden_hour | binary | Arrived within 1 hour |
+| prehospital_first_aid | binary | First aid given |
+| cervical_spine_immobilised | binary | C-spine immobilised |
+| referred_from_other | binary | Referred from another facility |
+| xray_done | binary | X-ray performed |
+| ct_done | binary | CT scan performed |
+| ultrasound_fast | binary | FAST ultrasound |
+| blood_transfusion | binary | Blood transfusion |
+| surgery_performed | binary | Surgery performed |
+| surgery_type | categorical | laparotomy / craniotomy / orif / external_fixation / amputation |
+| icu_admission | binary | ICU admission |
+| intubated | binary | Intubated |
+| tetanus_given | binary | Tetanus prophylaxis |
+| antibiotics_given | binary | Antibiotics |
+| wound_infection | binary | Wound infection |
+| sepsis | binary | Sepsis |
+| vte | binary | VTE |
+| outcome | categorical | survived / died |
+| length_of_stay_days | int | Hospital stay |
+| disability_at_discharge | categorical | none / paralysis / amputation / cognitive_impairment / mobility_limitation |
+
+## 4. Validation
+
+<p align="center">
+  <img src="validation_report.png" alt="Validation Report" width="100%">
+</p>
+
+Key validation checks:
+
+- **Mortality gradient**: 18.6% → 33.8% → 48.2%, reflecting infrastructure gap ✓
+- **Male predominance**: 75%, consistent with WHO 3:1 ratio ✓
+- **Pedestrians most common**: ~40%, consistent with WHO Africa data ✓
+- **GCS dose-response**: Severe GCS mortality >> mild ✓
+- **Golden hour**: Only 3-30% arrive within 1 hour ✓
+- **Head injury leading**: ~45% TBI, consistent with Chalya 2012 ✓
+
+## 5. Usage
+
+```python
+from datasets import load_dataset
+dataset = load_dataset("electricsheepafrica/road-traffic-injury-trauma", "district_hospital")
+df = dataset["train"].to_pandas()
+```
+
+```bash
+python generate_dataset.py --all-scenarios --n 10000 --seed 42
+```
+
+## 6. Limitations
+
+- **Simulated**: Not derived from real trauma registries.
+- **No imaging**: CT/X-ray findings are binary, no images.
+- **Simplified ISS**: Estimated, not calculated from individual AIS scores.
+- **No rehabilitation**: Discharge snapshot only, no long-term follow-up.
+- **No cost data**: No financial burden or catastrophic expenditure.
+
+## 7. References
+
+1. WHO (2023). Global Status Report on Road Safety.
+2. Galvagno SM, et al. (2019). Prehospital trauma care SSA.
+3. Zafar SN, et al. (2018). Trauma care Africa. *Lancet*, 391(10127):1308.
+4. WHO (2024). Road traffic injuries fact sheet.
+5. Chalya PL, et al. (2012). RTI Tanzania Bugando. *BMC Public Health*, 12:501.
+6. Hyder AA, et al. (2017). Cost of RTI in LMICs. *Bull WHO*, 95(5):326.
+7. Mwandri M, et al. (2020). Trauma SSA. *World J Emerg Surg*, 15:17.
+8. Mock C, et al. (2012). Prehospital trauma care. *Bull WHO*, 90(8):577.
+
+## Citation
+
+```bibtex
+@dataset{esa_rti_2025,
+  title={Road Traffic Injury and Trauma Dataset},
+  author={Electric Sheep Africa},
+  year={2025},
+  publisher={Hugging Face},
+  url={https://huggingface.co/datasets/electricsheepafrica/road-traffic-injury-trauma}
+}
+```
+
+## License
+
+[CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/)

generate_dataset.py

@@ -0,0 +1,382 @@
+#!/usr/bin/env python3
+"""
+Literature-Informed Road Traffic Injury & Trauma Dataset
+=========================================================
+
+Generates realistic synthetic records of road traffic injury patients
+presenting to emergency departments in sub-Saharan Africa, including
+injury mechanism, severity, prehospital care, emergency management,
+and outcomes.
+
+Target population: Road traffic injury patients of all ages presenting
+to health facilities across SSA.
+
+DAG (Sampling Order):
+  1. demographics: age, sex, road_user_type
+  2. injury: mechanism, body_region, severity (GCS, ISS), fractures
+  3. prehospital: time_to_facility, transport, first_aid, referral
+  4. emergency_care: imaging, blood, surgery, ICU
+  5. complications: infection, VTE, organ_failure
+  6. outcome: survived/died, disability, LOS
+
+References (web-searched):
+-----------
+[1]  WHO Global Status Report on Road Safety (2023). 1.35M road
+     deaths/yr. Africa highest rate: 26.6/100K vs 17.4 global.
+     90% in LMICs. Pedestrians/cyclists >50% in Africa.
+[2]  Galvagno SM, et al. (2019). Prehospital trauma: <10% arrive
+     within golden hour in many SSA settings.
+[3]  Zafar SN, et al. (Lancet 2018). Trauma care Africa: mortality
+     30-40% for severe injuries (ISS>15). TBI leading cause of death.
+[4]  WHO (2024). Road traffic injuries fact sheet. Young adults
+     15-44y most affected. Males 3x female risk.
+[5]  Chalya PL, et al. (BMC Public Health 2012). Tanzania Bugando:
+     RTI mortality 11.2%. Pedestrians 41%, motorcyclists 28%.
+     Head injury 44%, extremity 38%.
+[6]  Hyder AA, et al. (Bull WHO 2017). Cost of RTI in LMICs:
+     1-3% of GDP. Disability burden enormous.
+[7]  Mwandri M, et al. (World J Emerg Surg 2020). Trauma in SSA:
+     GCS on admission strongest predictor of mortality.
+[8]  Mock C, et al. (Bull WHO 2012). Strengthening prehospital
+     trauma care could prevent 54% of trauma deaths.
+"""
+
+import numpy as np
+import pandas as pd
+import argparse
+import os
+
+SCENARIOS = {
+    'trauma_centre': {
+        'description': 'Urban trauma centre with CT, ICU, neurosurgery, '
+                       'orthopaedics, blood bank (e.g., Muhimbili, '
+                       'Chris Hani Baragwanath, Kenyatta)',
+        'ct_available': True,
+        'icu_available': True,
+        'neurosurgery_available': True,
+        'blood_bank': True,
+        'ambulance_rate': 0.40,
+        'golden_hour_rate': 0.25,
+        'mortality': 0.08,
+    },
+    'district_hospital': {
+        'description': 'District hospital with basic X-ray, limited '
+                       'surgery, no ICU (e.g., district hospitals '
+                       'Tanzania, Malawi, Uganda)',
+        'ct_available': False,
+        'icu_available': False,
+        'neurosurgery_available': False,
+        'blood_bank': False,
+        'ambulance_rate': 0.10,
+        'golden_hour_rate': 0.10,
+        'mortality': 0.18,
+    },
+    'rural_health_centre': {
+        'description': 'Rural health centre, no imaging, no surgery, '
+                       'stabilise and refer (e.g., rural Nigeria, '
+                       'DRC, South Sudan)',
+        'ct_available': False,
+        'icu_available': False,
+        'neurosurgery_available': False,
+        'blood_bank': False,
+        'ambulance_rate': 0.03,
+        'golden_hour_rate': 0.03,
+        'mortality': 0.32,
+    },
+}
+
+ROAD_USER_TYPES = {
+    'pedestrian': 0.38,
+    'motorcyclist': 0.22,
+    'motor_vehicle_occupant': 0.18,
+    'cyclist': 0.10,
+    'passenger_minibus': 0.08,
+    'other': 0.04,
+}
+
+
+def generate_dataset(n=10000, seed=42, scenario='district_hospital'):
+    rng = np.random.default_rng(seed)
+    sc = SCENARIOS[scenario]
+
+    users = list(ROAD_USER_TYPES.keys())
+    user_p = list(ROAD_USER_TYPES.values())
+
+    records = []
+
+    for idx in range(n):
+        rec = {'id': idx + 1}
+
+        # ── Step 1: Demographics [4][5] ──
+        r = rng.random()
+        if r < 0.08:
+            rec['age_years'] = rng.integers(0, 15)
+        elif r < 0.40:
+            rec['age_years'] = rng.integers(15, 30)
+        elif r < 0.65:
+            rec['age_years'] = rng.integers(30, 45)
+        elif r < 0.85:
+            rec['age_years'] = rng.integers(45, 60)
+        else:
+            rec['age_years'] = rng.integers(60, 80)
+
+        rec['sex'] = rng.choice(['M', 'F'], p=[0.75, 0.25])
+
+        rec['road_user_type'] = rng.choice(users, p=user_p)
+
+        if rec['age_years'] < 15:
+            rec['road_user_type'] = rng.choice(
+                ['pedestrian', 'cyclist', 'passenger_minibus'],
+                p=[0.60, 0.20, 0.20])
+
+        rec['helmet_use'] = 0
+        if rec['road_user_type'] == 'motorcyclist':
+            rec['helmet_use'] = 1 if rng.random() < 0.25 else 0
+
+        rec['seatbelt_use'] = 0
+        if rec['road_user_type'] == 'motor_vehicle_occupant':
+            rec['seatbelt_use'] = 1 if rng.random() < 0.15 else 0
+
+        rec['alcohol_involved'] = 1 if rng.random() < 0.25 else 0
+
+        rec['time_of_day'] = rng.choice(
+            ['morning_6_12', 'afternoon_12_18', 'evening_18_24', 'night_0_6'],
+            p=[0.20, 0.30, 0.30, 0.20])
+
+        rec['road_type'] = rng.choice(
+            ['highway', 'urban_road', 'rural_road', 'intersection'],
+            p=[0.25, 0.30, 0.30, 0.15])
+
+        # ── Step 2: Injury [5] ──
+        rec['primary_body_region'] = rng.choice(
+            ['head_neck', 'chest', 'abdomen', 'upper_extremity',
+             'lower_extremity', 'spine', 'pelvis', 'multiple'],
+            p=[0.28, 0.10, 0.08, 0.12, 0.20, 0.06, 0.04, 0.12])
+
+        if rec['road_user_type'] in ('pedestrian', 'cyclist'):
+            if rng.random() < 0.35:
+                rec['primary_body_region'] = 'head_neck'
+
+        rec['tbi'] = 1 if rec['primary_body_region'] == 'head_neck' else 0
+        if rec['primary_body_region'] == 'multiple' and rng.random() < 0.40:
+            rec['tbi'] = 1
+
+        # GCS
+        if rec['tbi']:
+            gcs_roll = rng.random()
+            if gcs_roll < 0.25:
+                rec['gcs'] = rng.integers(3, 9)
+            elif gcs_roll < 0.50:
+                rec['gcs'] = rng.integers(9, 13)
+            else:
+                rec['gcs'] = rng.integers(13, 16)
+        else:
+            rec['gcs'] = rng.choice([14, 15], p=[0.20, 0.80])
+
+        rec['gcs_category'] = 'mild'
+        if rec['gcs'] <= 8:
+            rec['gcs_category'] = 'severe'
+        elif rec['gcs'] <= 12:
+            rec['gcs_category'] = 'moderate'
+
+        # ISS
+        if rec['primary_body_region'] == 'multiple':
+            rec['iss'] = max(9, int(rng.normal(25, 10)))
+        elif rec['tbi'] and rec['gcs'] <= 8:
+            rec['iss'] = max(16, int(rng.normal(30, 10)))
+        elif rec['primary_body_region'] in ('chest', 'abdomen'):
+            rec['iss'] = max(4, int(rng.normal(16, 8)))
+        else:
+            rec['iss'] = max(1, int(rng.normal(10, 6)))
+        rec['iss'] = min(rec['iss'], 75)
+
+        rec['iss_category'] = 'minor'
+        if rec['iss'] >= 25:
+            rec['iss_category'] = 'critical'
+        elif rec['iss'] >= 16:
+            rec['iss_category'] = 'severe'
+        elif rec['iss'] >= 9:
+            rec['iss_category'] = 'moderate'
+
+        rec['fracture'] = 0
+        if rec['primary_body_region'] in ('upper_extremity', 'lower_extremity', 'pelvis'):
+            rec['fracture'] = 1 if rng.random() < 0.70 else 0
+        elif rec['primary_body_region'] == 'multiple':
+            rec['fracture'] = 1 if rng.random() < 0.50 else 0
+        else:
+            rec['fracture'] = 1 if rng.random() < 0.15 else 0
+
+        rec['open_fracture'] = 0
+        if rec['fracture']:
+            rec['open_fracture'] = 1 if rng.random() < 0.30 else 0
+
+        rec['internal_bleeding'] = 0
+        if rec['primary_body_region'] in ('abdomen', 'chest', 'pelvis', 'multiple'):
+            rec['internal_bleeding'] = 1 if rng.random() < 0.25 else 0
+
+        rec['spinal_cord_injury'] = 0
+        if rec['primary_body_region'] == 'spine':
+            rec['spinal_cord_injury'] = 1 if rng.random() < 0.35 else 0
+
+        # ── Step 3: Prehospital [2][8] ──
+        tp = np.array([sc['ambulance_rate'], 0.30, 0.10, 0.20, 0.05, 0.35 - sc['ambulance_rate']])
+        tp = np.maximum(tp, 0.01)
+        tp = tp / tp.sum()
+        rec['transport_mode'] = rng.choice(
+            ['ambulance', 'private_vehicle', 'police', 'bystander',
+             'walked', 'motorcycle_taxi'], p=tp)
+
+        rec['time_to_facility_hours'] = max(0.25, round(rng.exponential(3), 1))
+        rec['time_to_facility_hours'] = min(rec['time_to_facility_hours'], 48)
+        rec['within_golden_hour'] = 1 if rec['time_to_facility_hours'] <= 1.0 else 0
+
+        rec['prehospital_first_aid'] = 1 if rng.random() < 0.20 else 0
+        rec['cervical_spine_immobilised'] = 0
+        if rec['transport_mode'] == 'ambulance':
+            rec['cervical_spine_immobilised'] = 1 if rng.random() < 0.50 else 0
+
+        rec['referred_from_other'] = 1 if rng.random() < 0.25 else 0
+
+        # ── Step 4: Emergency Care ──
+        rec['xray_done'] = 1 if rng.random() < 0.70 else 0
+        rec['ct_done'] = 0
+        if sc['ct_available'] and (rec['tbi'] or rec['iss'] >= 16):
+            rec['ct_done'] = 1 if rng.random() < 0.65 else 0
+
+        rec['ultrasound_fast'] = 0
+        if rec['internal_bleeding'] or rec['primary_body_region'] in ('abdomen', 'pelvis'):
+            rec['ultrasound_fast'] = 1 if rng.random() < 0.40 else 0
+
+        rec['blood_transfusion'] = 0
+        if rec['internal_bleeding'] or rec['iss'] >= 25:
+            if sc['blood_bank']:
+                rec['blood_transfusion'] = 1 if rng.random() < 0.45 else 0
+            else:
+                rec['blood_transfusion'] = 1 if rng.random() < 0.10 else 0
+
+        rec['surgery_performed'] = 0
+        rec['surgery_type'] = 'none'
+        if rec['internal_bleeding'] and rng.random() < 0.50:
+            rec['surgery_performed'] = 1
+            rec['surgery_type'] = 'laparotomy'
+        elif rec['tbi'] and rec['gcs'] <= 8 and sc['neurosurgery_available']:
+            rec['surgery_performed'] = 1 if rng.random() < 0.30 else 0
+            if rec['surgery_performed']:
+                rec['surgery_type'] = 'craniotomy'
+        elif rec['fracture']:
+            rec['surgery_performed'] = 1 if rng.random() < 0.35 else 0
+            if rec['surgery_performed']:
+                rec['surgery_type'] = rng.choice(
+                    ['orif', 'external_fixation', 'amputation'],
+                    p=[0.50, 0.35, 0.15])
+
+        rec['icu_admission'] = 0
+        if sc['icu_available'] and (rec['gcs'] <= 8 or rec['iss'] >= 25):
+            rec['icu_admission'] = 1 if rng.random() < 0.40 else 0
+
+        rec['intubated'] = 0
+        if rec['gcs'] <= 8:
+            rec['intubated'] = 1 if rng.random() < (0.60 if sc['icu_available'] else 0.10) else 0
+
+        rec['tetanus_given'] = 1 if rng.random() < 0.65 else 0
+        rec['antibiotics_given'] = 1 if (rec['open_fracture'] or rec['surgery_performed']) else 0
+
+        # ── Step 5: Complications ──
+        rec['wound_infection'] = 0
+        if rec['open_fracture'] or rec['surgery_performed']:
+            rec['wound_infection'] = 1 if rng.random() < 0.15 else 0
+
+        rec['sepsis'] = 0
+        if rec['wound_infection'] or (rec['icu_admission'] and rng.random() < 0.10):
+            rec['sepsis'] = 1 if rng.random() < 0.25 else 0
+
+        rec['vte'] = 0
+        if rec['fracture'] and rec['primary_body_region'] in ('lower_extremity', 'pelvis'):
+            rec['vte'] = 1 if rng.random() < 0.05 else 0
+
+        # ── Step 6: Outcome [3][7] ──
+        mort = sc['mortality']
+        if rec['gcs'] <= 8:
+            mort *= 4.0
+        elif rec['gcs'] <= 12:
+            mort *= 2.0
+        if rec['iss'] >= 25:
+            mort *= 2.5
+        if rec['internal_bleeding'] and not rec['blood_transfusion']:
+            mort *= 2.0
+        if not rec['within_golden_hour'] and rec['iss'] >= 16:
+            mort *= 1.5
+        if rec['age_years'] > 60:
+            mort *= 1.5
+        if rec['age_years'] < 5:
+            mort *= 1.3
+        if rec['surgery_performed'] and rec['internal_bleeding']:
+            mort *= 0.50
+        if rec['icu_admission']:
+            mort *= 0.70
+
+        rec['outcome'] = 'died' if rng.random() < min(mort, 0.80) else 'survived'
+
+        if rec['outcome'] == 'died':
+            rec['length_of_stay_days'] = max(0, rng.integers(0, 7))
+        elif rec['iss'] >= 16:
+            rec['length_of_stay_days'] = max(3, int(rng.normal(18, 10)))
+        else:
+            rec['length_of_stay_days'] = max(1, int(rng.normal(5, 3)))
+
+        rec['disability_at_discharge'] = 'none'
+        if rec['outcome'] == 'survived':
+            if rec['spinal_cord_injury']:
+                rec['disability_at_discharge'] = 'paralysis'
+            elif rec['surgery_type'] == 'amputation':
+                rec['disability_at_discharge'] = 'amputation'
+            elif rec['tbi'] and rec['gcs'] <= 12:
+                rec['disability_at_discharge'] = 'cognitive_impairment'
+            elif rec['fracture']:
+                rec['disability_at_discharge'] = 'mobility_limitation'
+
+        records.append(rec)
+
+    df = pd.DataFrame(records)
+
+    print(f"\n{'='*65}")
+    print(f"Road Traffic Injury — {scenario} (n={n}, seed={seed})")
+    print(f"{'='*65}")
+
+    print(f"\n  Male: {(df['sex']=='M').mean()*100:.1f}%")
+    print(f"  Pedestrians: {(df['road_user_type']=='pedestrian').mean()*100:.1f}%")
+    print(f"  TBI: {df['tbi'].mean()*100:.1f}%")
+    print(f"  Severe (ISS≥16): {(df['iss']>=16).mean()*100:.1f}%")
+    print(f"  Within golden hour: {df['within_golden_hour'].mean()*100:.1f}%")
+    print(f"  Ambulance: {(df['transport_mode']=='ambulance').mean()*100:.1f}%")
+
+    died = (df['outcome'] == 'died').sum()
+    print(f"  Mortality: {died} ({died/n*100:.1f}%)")
+
+    return df
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(
+        description='Generate synthetic road traffic injury 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'rti_{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'rti_{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 +1,130 @@
+#!/usr/bin/env python3
+"""Validation & Diagnostic Visualization for Road Traffic Injury Dataset."""
+
+import pandas as pd
+import numpy as np
+import matplotlib.pyplot as plt
+import os
+
+SCENARIOS = ['trauma_centre', 'district_hospital', 'rural_health_centre']
+
+
+def load_scenarios(data_dir='data'):
+    dfs = {}
+    for sc in SCENARIOS:
+        path = os.path.join(data_dir, f'rti_{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('Road Traffic Injury & Trauma — Validation Report',
+                 fontsize=16, fontweight='bold', y=0.98)
+    df = dfs.get('district_hospital', list(dfs.values())[0])
+    colors = ['#2ecc71', '#f39c12', '#e74c3c']
+
+    # Panel 1: Mortality across scenarios
+    ax = axes[0, 0]
+    x = np.arange(len(SCENARIOS))
+    mort = [(dfs[sc]['outcome'] == 'died').mean() * 100 for sc in SCENARIOS if sc in dfs]
+    ax.bar(x, mort, color=colors, alpha=0.8)
+    ax.set_xticks(x)
+    ax.set_xticklabels(['Trauma Centre', 'District', 'Rural'], fontsize=9)
+    for i, v in enumerate(mort):
+        ax.text(i, v + 0.3, f'{v:.1f}%', ha='center', fontsize=10)
+    ax.set_ylabel('Mortality (%)')
+    ax.set_title('RTI Mortality (Africa: 26.6/100K)')
+
+    # Panel 2: Road user type
+    ax = axes[0, 1]
+    users = df['road_user_type'].value_counts()
+    u_colors = ['#e74c3c', '#3498db', '#f39c12', '#2ecc71', '#9b59b6', '#e67e22']
+    ax.pie(users.values,
+           labels=[u.replace('_', ' ').title() for u in users.index],
+           autopct='%1.1f%%', colors=u_colors[:len(users)],
+           startangle=90, textprops={'fontsize': 8})
+    ax.set_title('Road User Type (WHO: Pedestrians >50% Africa)')
+
+    # Panel 3: GCS vs mortality
+    ax = axes[1, 0]
+    gcs_cats = ['severe', 'moderate', 'mild']
+    gcs_mort = []
+    for g in gcs_cats:
+        sub = df[df['gcs_category'] == g]
+        gcs_mort.append((sub['outcome'] == 'died').mean() * 100 if len(sub) > 0 else 0)
+    g_colors = ['#e74c3c', '#f39c12', '#2ecc71']
+    ax.bar(range(3), gcs_mort, color=g_colors, alpha=0.8)
+    ax.set_xticks(range(3))
+    ax.set_xticklabels(['Severe (3-8)', 'Moderate (9-12)', 'Mild (13-15)'])
+    for i, v in enumerate(gcs_mort):
+        ax.text(i, v + 0.3, f'{v:.0f}%', ha='center', fontsize=9)
+    ax.set_ylabel('Mortality (%)')
+    ax.set_title('Mortality by GCS Category')
+
+    # Panel 4: Body region
+    ax = axes[1, 1]
+    regions = df['primary_body_region'].value_counts()
+    ax.barh(range(len(regions)), regions.values, color='#3498db', alpha=0.7)
+    ax.set_yticks(range(len(regions)))
+    ax.set_yticklabels([r.replace('_', ' ').title()[:15] for r in regions.index], fontsize=7)
+    ax.set_xlabel('Count')
+    ax.set_title('Primary Body Region (Head 28%, Extremity 32%)')
+
+    # Panel 5: Ambulance & golden hour
+    ax = axes[2, 0]
+    amb = [(dfs[sc]['transport_mode'] == 'ambulance').mean() * 100 for sc in SCENARIOS if sc in dfs]
+    gh = [dfs[sc]['within_golden_hour'].mean() * 100 for sc in SCENARIOS if sc in dfs]
+    w = 0.3
+    ax.bar(x - w/2, amb, w, label='Ambulance', color='#3498db', alpha=0.8)
+    ax.bar(x + w/2, gh, w, label='Golden Hour', color='#f39c12', alpha=0.8)
+    ax.set_xticks(x)
+    ax.set_xticklabels(['Trauma', 'District', 'Rural'], fontsize=9)
+    ax.set_ylabel('Rate (%)')
+    ax.set_title('Prehospital: Ambulance & Golden Hour')
+    ax.legend(fontsize=8)
+
+    # Panel 6: ISS distribution
+    ax = axes[2, 1]
+    for sc in SCENARIOS:
+        if sc in dfs:
+            ax.hist(dfs[sc]['iss'].clip(1, 50), bins=20, alpha=0.5,
+                    label=sc.replace('_', ' ').title()[:12], edgecolor='white')
+    ax.axvline(x=16, color='red', linestyle='--', alpha=0.7, label='Severe (ISS≥16)')
+    ax.set_xlabel('ISS')
+    ax.set_title('Injury Severity Score Distribution')
+    ax.legend(fontsize=7)
+
+    # Panel 7: Age-sex distribution
+    ax = axes[3, 0]
+    males = df[df['sex'] == 'M']['age_years']
+    females = df[df['sex'] == 'F']['age_years']
+    ax.hist(males, bins=15, alpha=0.5, color='#3498db', label='Male', edgecolor='white')
+    ax.hist(females, bins=15, alpha=0.5, color='#e74c3c', label='Female', edgecolor='white')
+    ax.set_xlabel('Age (years)')
+    ax.set_title('Age-Sex Distribution (Males 75%, peak 15-44y)')
+    ax.legend(fontsize=8)
+
+    # Panel 8: Disability at discharge
+    ax = axes[3, 1]
+    surv = df[df['outcome'] == 'survived']
+    if len(surv) > 0:
+        dis = surv['disability_at_discharge'].value_counts()
+        d_colors = ['#2ecc71', '#f39c12', '#e74c3c', '#9b59b6', '#3498db']
+        ax.pie(dis.values,
+               labels=[d.replace('_', ' ').title() for d in dis.index],
+               autopct='%1.1f%%', colors=d_colors[:len(dis)],
+               startangle=90, textprops={'fontsize': 8})
+    ax.set_title('Disability at Discharge (Survivors)')
+
+    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)