electricsheepafrica/paediatric-burn-injuries

Paediatric Burn Injuries Dataset

Abstract

This dataset provides 30,000 simulated paediatric burn records (10,000 per scenario) of children 0–15 years presenting with burn injuries to health facilities in sub-Saharan Africa. Each record contains 40+ variables including mechanism (scald/flame/electrical), TBSA%, burn depth, body region, first aid, wound management, surgical intervention, complications (infection, sepsis, contracture), and outcomes. Three settings: dedicated burn centre (3.2% mortality), district surgical ward (11.3%), and rural health facility (31.4%).

1. Introduction

Burns are a leading cause of preventable childhood morbidity and mortality in SSA, with mortality averaging 17% — one in every five burn victims (Nthumba, Burns 2016). Children ≤10 years represent >80% of burn patients. Scalds account for 59% of burns and flame burns for 33%. Flame burns carry 2.6× higher mortality than scalds after controlling for TBSA and age (Gallaher et al., Burns 2021). The LA50 (lethal area 50%) in SSA burn units is only 30–40% TBSA versus >70% in HICs (Albertyn et al., 2014). WHO reports 180,000 burn deaths/year globally, with 2/3 in Africa and SE Asia.

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

2. Methodology

2.1 Parameterization

Parameter	Value	Source
Children ≤10y among burn patients	>80%	Nthumba, Burns 2016
Scald : flame ratio	59% : 33%	Nthumba, Burns 2016
Scald TBSA (median)	12% (IQR 7-20)	Gallaher et al., Burns 2021
Flame TBSA (median)	17.8% (IQR 10-28)	Gallaher et al., Burns 2021
Flame mortality vs scald	2.6× higher	Gallaher et al., Burns 2021
Average mortality (SSA)	17%	Nthumba, Burns 2016
LA50 TBSA (SSA burn units)	30–40%	Albertyn et al., Burns 2014

2.2 Scenario Design

Scenario	Facility	Surgery	Skin Graft	Sepsis	Mortality
Burn centre	Dedicated burn ICU	40%	Yes	8%	3.2%
District surgical	District surgical ward	21%	No	14%	11.3%
Rural health facility	Basic wound care only	4%	No	20%	31.4%

3. Schema

Column	Type	Description
id	int	Unique identifier
age_months	int	Age in months
age_years	int	Age in years
sex	categorical	M / F
caregiver_present	binary	Caregiver present at time of injury
nutritional_status	categorical	normal / moderate_malnutrition / severe_malnutrition
mechanism	categorical	scald / flame / electrical / chemical / contact
burn_agent	categorical	hot_water / kerosene_lamp / open_cooking_fire / etc.
tbsa_percent	int	Total body surface area burned (%)
burn_depth	categorical	superficial / partial_thickness / deep_partial / full_thickness
body_region_primary	categorical	upper_limb / lower_limb / trunk / head_neck / hands / feet / perineum
inhalation_injury	binary	Inhalation injury present
circumferential_burn	binary	Circumferential burn
suspected_non_accidental	binary	Suspected non-accidental injury
first_aid_applied	binary	First aid applied
first_aid_type	categorical	cool_water / traditional_remedy / toothpaste_egg / butter_oil / none
appropriate_first_aid	binary	Cool running water applied
time_to_facility_hours	float	Time from injury to facility
referred_from_other	binary	Referred from another facility
fluid_resuscitation	binary	IV fluid resuscitation
wound_dressing	categorical	silver_sulfadiazine / honey / saline_gauze / paraffin_gauze / traditional
surgery_performed	binary	Surgical intervention
surgery_type	categorical	debridement / skin_graft / escharotomy / amputation / contracture_release
antibiotics_given	binary	Antibiotics administered
pain_management	categorical	paracetamol / ibuprofen / morphine / ketamine / none
nutritional_support	binary	Nutritional support provided
physiotherapy	binary	Physiotherapy
blood_transfusion	binary	Blood transfusion
wound_infection	binary	Wound infection
sepsis	binary	Sepsis
contracture	binary	Burn contracture
pneumonia	binary	Pneumonia
outcome	categorical	survived / died
length_of_stay_days	int	Hospital stay
disability_at_discharge	categorical	none / functional_limitation / scarring_cosmetic / amputation

4. Validation

Key validation checks:

• Mortality gradient: Burn centre 3.2% → District 11.3% → Rural 31.4% ✓
• Scald dominant: ~62% scalds, consistent with Nthumba 59% ✓
• Flame mortality higher: 2–3× scald mortality across scenarios ✓
• TBSA distributions: Scald ~12%, flame ~18%, matching Gallaher 2021 ✓
• Age peak <5y: ~66%, consistent with literature ✓
• Appropriate first aid: Only 10%, reflecting SSA reality ✓

5. Usage

from datasets import load_dataset
dataset = load_dataset("electricsheepafrica/paediatric-burn-injuries", "district_surgical")
df = dataset["train"].to_pandas()

python generate_dataset.py --all-scenarios --n 10000 --seed 42

6. Limitations

• Simulated: Not derived from real burn registry data.
• No wound photographs: TBSA and depth are numerical, no imaging.
• Simplified contracture: Binary, no grading of functional limitation.
• No long-term follow-up: Discharge snapshot only, no reconstructive surgery tracking.
• No cost/psychosocial: No financial burden or PTSD assessment.

7. References

1. Nthumba PM (2016). Burns in SSA: a review. Burns, 42(5):1169–1176.
2. Gallaher JR, et al. (2021). Burn mechanism and paediatric mortality in Malawi. Burns, 47(3):583–590.
3. Forjuoh SN (2006). Burns in LMICs: a review. Burns, 32(5):529–537.
4. Albertyn R, et al. (2014). Burn care in developing countries. Burns, 40(Suppl 1):S3.
5. Kemp AM, et al. (2014). Non-accidental burns in children. Burns, 40(7):1276.
6. WHO (2018). Burns fact sheet.
7. Rode H, et al. (2017). Burn care in Africa. Lancet, 390(10106):1939.
8. Parbhoo A, et al. (2010). Burn wound sepsis in SSA. Burns, 36(7):1027–1031.

Citation

@dataset{esa_paediatric_burns_2025,
  title={Paediatric Burn Injuries Dataset},
  author={Electric Sheep Africa},
  year={2025},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/electricsheepafrica/paediatric-burn-injuries}
}

License

CC-BY-4.0