electricsheepafrica/africa-synth-malaria-synthetic-malaria-clinical-all

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

Synthetic Malaria Clinical & Diagnostic Dataset

Abstract

This dataset provides 30,000 synthetic records (10,000 per scenario) of patients presenting with suspected malaria at LMIC health facilities. Each record contains 24 variables including demographics, clinical symptoms (fever, headache, chills, vomiting, danger signs), diagnostic results (RDT, microscopy with parasite density), haematological data (hemoglobin), parasite species, WHO severity classification, and clinical outcome. RDT and microscopy results are generated with literature-informed sensitivity/specificity, creating realistic diagnostic discordance. Three endemicity scenarios (low, moderate, high) span test positivity rates from 12% to 58%.

1. Introduction

Malaria remains a leading cause of morbidity and mortality in tropical LMICs, with 249 million cases and 608,000 deaths in 2022 (WHO 2023). Accurate diagnosis—distinguishing malaria from other febrile illnesses—is critical for appropriate treatment. Despite massive research interest in ML-assisted malaria diagnosis, open-access tabular clinical datasets with both RDT and microscopy results are extremely rare.

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

2. Methodology

2.1 Diagnostic Test Performance

Test	Sensitivity (high density)	Sensitivity (low density)	Specificity	Source
RDT (HRP2)	95% (≥200 p/µL)	60% (<200 p/µL)	95%	Ochola et al., 2006; WHO 2023
Microscopy	90% (≥100 p/µL)	50% (<100 p/µL)	98%	Ochola et al., 2006

2.2 Scenario Design

Scenario	True Malaria+	RDT+	Severe	P. falciparum	Deaths
Low endemic	11.7%	15.2%	0.2%	68% of positives	0.02%
Moderate endemic	34.6%	35.0%	2.0%	90% of positives	0.19%
High endemic	57.6%	55.2%	5.9%	95% of positives	0.61%

3. Schema

Column	Type	Description
id	int	Unique identifier
age_years	float	Age (0.5-75)
sex	categorical	M/F
pregnant	binary	Pregnancy status
temperature_c	float	Axillary temperature (°C)
fever_reported	binary	Patient/caregiver reported fever
fever_duration_days	int	Days of fever
headache, chills_rigors, vomiting, diarrhea, body_aches	binary	Symptoms
convulsions, altered_consciousness, jaundice	binary	WHO danger signs
rdt_result	binary	Rapid diagnostic test result
microscopy_result	binary	Blood smear microscopy result
microscopy_parasite_density	int	Microscopy-estimated density (p/µL)
true_parasite_density	int	True parasite density (p/µL)
parasite_species	categorical	P_falciparum / P_vivax / P_malariae / none
hemoglobin_gdl	float	Hemoglobin (g/dL)
severe_anemia	binary	Hb < 5 g/dL
severity_classification	categorical	not_malaria / uncomplicated / severe
outcome	categorical	discharged / treated_outpatient / admitted / death

4. Validation

5. Usage

from datasets import load_dataset
dataset = load_dataset("electricsheepafrica/synthetic-malaria-clinical-diagnosis-RDT-microscopy", "moderate_endemic")
df = dataset["train"].to_pandas()

6. Limitations

• Synthetic: Not for clinical use or diagnostic algorithm validation on real populations.
• No co-infections: Does not model malaria-typhoid, malaria-pneumonia overlap.
• Simplified species: Three species only; no mixed infections modelled.
• No geographic/seasonal variation: Endemicity is static, not seasonal.

7. References

1. WHO (2023). World Malaria Report 2023. Geneva.
2. WHO (2022). Guidelines for malaria (3rd edition). Geneva.
3. Dalrymple U, et al. (2019). Non-P. falciparum malaria species. Lancet Infect Dis, 19(12):1260-1267.
4. Ochola LB, et al. (2006). Reliability of diagnostic techniques. Lancet Infect Dis, 6(10):582-588.
5. Severe Malaria (2014). Trop Med Int Health, 19(s1).

Citation

@dataset{esa_malaria_2025,
  title={Synthetic Malaria Clinical and Diagnostic Dataset},
  author={Electric Sheep Africa},
  year={2025},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/electricsheepafrica/synthetic-malaria-clinical-diagnosis-RDT-microscopy}
}

License

CC-BY-4.0