electricsheepafrica/africa-synth-malaria-chw-iccm-triage-all

⚠️ Synthetic dataset — Parameterized from published SSA literature, not real observations. Not suitable for empirical analysis or policy inference.

Synthetic Community Health Worker iCCM Triage Dataset (2–59 months)

Abstract

This dataset provides 30,000 synthetic records (10,000 per scenario) of sick child assessments by community health workers (CHWs) using integrated Community Case Management (iCCM) protocols. Each record contains 29 variables covering demographics, symptoms and signs (fever, cough, fast breathing, diarrhoea, dehydration, danger signs), RDT result, MUAC-based nutrition screening, CHW classification, treatment decisions (ACT, amoxicillin, ORS/zinc), and referral actions. The iCCM algorithm covers malaria, pneumonia, and diarrhoea—the three leading causes of under-5 mortality in LMICs.

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

2. Methodology

2.1 iCCM Protocol

Based on WHO/UNICEF "Caring for the Sick Child in the Community" (2012):

Condition	Assessment	Classification	Treatment
Malaria	Fever + RDT	RDT+ → malaria	ACT
Pneumonia	Cough + fast breathing	Fast breathing → pneumonia	Amoxicillin
Diarrhoea	Loose stools + dehydration	Diarrhoea ± dehydration	ORS + Zinc
Danger signs	Unable to drink, convulsions, lethargy, chest indrawing	Any → refer urgently	Pre-referral + refer
Malnutrition	MUAC	<11.5cm SAM, 11.5-12.4cm MAM	Refer

2.2 Scenario Design

Scenario	Malaria	Pneumonia	Diarrhoea	Danger Signs	SAM	Referral
Low burden	19.9%	15.2%	24.1%	6.5%	2.3%	18.2%
Moderate burden	35.1%	17.4%	21.8%	9.6%	4.3%	24.8%
High burden	44.5%	19.9%	18.5%	14.0%	8.1%	32.2%

3. Schema

Column	Type	Description
age_months	int	Age (2-59 months)
sex	categorical	M/F
fever, cough, diarrhoea	binary	Presenting symptoms
temperature_c	float	Axillary temperature
fever_duration_days	int	Days of fever
respiratory_rate_bpm	int	Respiratory rate
fast_breathing	binary	WHO age-specific threshold
chest_indrawing	binary	Lower chest wall indrawing
diarrhoea_duration_days	int	Days of diarrhoea
blood_in_stool	binary	Dysentery indicator
dehydration_status	categorical	none/some/severe
rdt_result	binary	Malaria RDT result
muac_cm	float	Mid-upper arm circumference
nutrition_status	categorical	normal/MAM/SAM
unable_to_drink, vomiting_everything, convulsions, lethargic_unconscious	binary	WHO/UNICEF danger signs
any_danger_sign	binary	Any danger sign present
true_diagnosis	categorical	malaria/pneumonia/diarrhoea/mixed/other_febrile
chw_classification	categorical	iCCM algorithm classification
chw_action	categorical	treat_at_community/treat_and_refer/refer/refer_urgently
act_given, amoxicillin_given, ors_given, zinc_given	binary	Treatments administered

4. Validation

5. Usage

from datasets import load_dataset
dataset = load_dataset("electricsheepafrica/synthetic-chw-iccm-triage-WHO-UNICEF-2-59months", "moderate_burden")
df = dataset["train"].to_pandas()

6. Limitations

• Synthetic: Not for clinical use or programme evaluation.
• Simplified diagnostics: Real CHW assessments involve subjective judgement not fully captured.
• No follow-up: Single encounter; no outcome tracking after treatment/referral.
• No medication stockouts: Real iCCM faces frequent commodity shortages.

7. References

1. WHO/UNICEF (2012). Caring for the sick child in the community. Geneva.
2. WHO/UNICEF (2014). Integrated Community Case Management: Evidence review. Geneva.
3. Marsh DR, et al. (2012). Introduction to iCCM. Am J Trop Med Hyg, 87(5 Suppl):1-5.
4. Druetz T, et al. (2015). Impact of iCCM on child mortality. Paediatrics & Int Child Health, 35(1):18-29.
5. WHO (2014). Revised classification and treatment of pneumonia in children. Geneva.

Citation

@dataset{esa_iccm_2025,
  title={Synthetic Community Health Worker iCCM Triage Dataset},
  author={Electric Sheep Africa},
  year={2025},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/electricsheepafrica/synthetic-chw-iccm-triage-WHO-UNICEF-2-59months}
}

License

CC-BY-4.0