electricsheepafrica/epilepsy-neurological-disorders

Epilepsy & Neurological Disorders Dataset

Abstract

This dataset provides 30,000 synthetic records (10,000 per scenario) of people with epilepsy (PWE) aged 1–80 years in sub-Saharan African settings. Each record contains 27 variables including seizure type (ILAE classification), epilepsy classification, etiology, AED type and adherence, seizure frequency, treatment gap status with reasons, EEG and neuroimaging results, comorbidities (depression, cognitive impairment), and outcomes. Three scenarios span from urban tertiary neurology (30% treatment gap, 2.5% mortality) to rural community settings (80% treatment gap, 10.1% mortality).

1. Introduction

Epilepsy affects ~50 million people worldwide, with 80% in LMICs (WHO 2023). SSA has the highest incidence (80–120/100,000/yr vs 50/100,000 in HICs) and a treatment gap of 68.5% (95% CI 59.5–77.5%), twice as high in rural vs urban areas (Owolabi et al., Epilepsia 2020; 23 studies). Major etiologies include perinatal injury, febrile seizures, neurocysticercosis, head trauma, and CNS infections (Ba-Diop et al., Lancet Neurol 2014). Phenobarbital remains the most available AED at <$5/year, yet 70% of PWE could be seizure-free with appropriate treatment.

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

2. Methodology

2.1 Epidemiological Parameterization

Parameter	Value	Source
Epilepsy prevalence (SSA)	7–15/1000	Diop et al., Epilepsia 2005
Incidence (SSA)	80–120/100,000/yr	Newton & Garcia, Lancet 2012
Treatment gap (SSA pooled)	68.5% (95% CI 59.5–77.5%)	Owolabi et al., Epilepsia 2020
Rural vs urban treatment gap	2× higher in rural	Owolabi et al., 2020
Seizure freedom with AEDs	~70%	WHO 2023
Mortality (PWE vs general)	3–6× higher	Kariuki et al., Lancet Neurol 2014
Neurocysticercosis (endemic areas)	10–30% of epilepsy	Preux & Druet-Cabanac, 2005
Phenobarbital cost	<$5 USD/year	Mbuba et al., Epilepsia 2008

2.2 Scenario Design

Scenario	Setting	Treatment Gap	Seizure-Free	EEG Done	Mortality
Urban tertiary	Neurology clinic, multiple AEDs	30%	40%	60%	2.5%
District hospital	1–2 AEDs, limited investigations	55%	30%	15%	4.7%
Rural community	Minimal AED access, phenobarbital	80%	25%	3%	10.1%

3. Schema

Column	Type	Description
id	int	Unique identifier
age_years	int	Age (1–80)
sex	categorical	M / F
etiology	categorical	unknown_presumed_genetic / perinatal_injury / febrile_seizure_history / neurocysticercosis / head_trauma / cns_infection_meningitis / cerebrovascular / brain_tumor / hiv_related / other_structural
seizure_type	categorical	generalized_tonic_clonic / focal_aware / focal_impaired_awareness / focal_to_bilateral_tc / absence / myoclonic
epilepsy_classification	categorical	generalized / focal / focal_to_bilateral
age_at_onset_years	int	Age at first seizure
duration_years	int	Duration of epilepsy
in_treatment_gap	binary	Currently untreated despite active epilepsy
treatment_gap_reason	categorical	no_access_aed / cost / traditional_healer / stigma / no_diagnosis / distance_to_facility / not_applicable
current_aed	categorical	phenobarbital / carbamazepine / sodium_valproate / levetiracetam / phenytoin / lamotrigine / none
aed_dose_adequate	binary	AED dose at therapeutic level
aed_adherence	categorical	good / moderate / poor / not_applicable
aed_side_effects	binary	Experiencing AED side effects
seizures_per_month	int	Monthly seizure frequency
seizure_free_12mo	binary	No seizures in past 12 months
hx_status_epilepticus	binary	History of status epilepticus
eeg_done	binary	EEG performed
eeg_result	categorical	generalized_epileptiform / focal_epileptiform / normal / not_done
neuroimaging_done	binary	CT/MRI performed
neuroimaging_result	categorical	abnormal / normal / not_done
depression	binary	Comorbid depression
cognitive_impairment	binary	Cognitive impairment
school_attendance	int	School attendance (children <18; -1 = NA for adults)
burns_or_injury_from_seizure	binary	Burns or injury from uncontrolled seizures
outcome	categorical	seizure_free / improved / unchanged / uncontrolled / died

4. Validation

Key validation checks:

• Treatment gap gradient: 30% → 55% → 80% across scenarios ✓
• Seizure freedom: Higher in treated (47%) vs untreated patients ✓
• AED distribution: Phenobarbital dominant in rural (75%) vs balanced in urban ✓
• Mortality gradient: 2.5% → 4.7% → 10.1% inversely correlated with treatment access ✓
• Etiology: Perinatal injury and genetic causes most common, consistent with literature ✓

5. Usage

from datasets import load_dataset
dataset = load_dataset("electricsheepafrica/epilepsy-neurological-disorders", "district_hospital")
df = dataset["train"].to_pandas()

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

6. Limitations

• Synthetic: Not derived from real patient records.
• Cross-sectional: No longitudinal seizure diaries or AED titration trajectories.
• Simplified AED model: Monotherapy only; polytherapy not modelled.
• No EEG waveform data: Only categorical EEG interpretation results.
• No drug levels: No therapeutic drug monitoring data.
• No nodding syndrome: River blindness-associated epilepsy not specifically modelled.

7. References

1. Ba-Diop A, et al. (2014). Epidemiology, causes, and treatment of epilepsy in SSA. Lancet Neurol, 13(10):1029–1044. PMID: 25231525
2. Owolabi LF, et al. (2020). Epilepsy treatment gap in SSA: meta-analysis. Epilepsia, 61(5):909–921. PMID: 32219865
3. WHO (2023). Epilepsy fact sheet. 50 million PWE globally.
4. Newton CR, Garcia HH (2012). Epilepsy in poor countries. Lancet, 380(9848):1193–1201.
5. Diop AG, et al. (2005). Active epilepsy prevalence in SSA. Epilepsia, 46(Suppl 11):33–35.
6. Mbuba CK, et al. (2008). Treatment gap in developing countries. Epilepsia, 49(9):1491–1503.
7. Preux PM, Druet-Cabanac M (2005). Etiology of epilepsy in tropics. Neuroepidemiology, 25(4):167–175.
8. ILAE (2017). Classification of seizure types and epilepsies.
9. Kariuki SM, et al. (2014). Mortality in epilepsy in SSA. Lancet Neurol, 13(7):701–709.

Citation

@dataset{esa_epilepsy_2025,
  title={Epilepsy and Neurological Disorders Dataset},
  author={Electric Sheep Africa},
  year={2025},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/electricsheepafrica/epilepsy-neurological-disorders}
}

License

CC-BY-4.0