---
license: cc-by-4.0
task_categories:
- tabular-classification
language:
- en
tags:
- synthetic
- healthcare
- snakebite
- envenomation
- antivenom
- neglected-tropical-disease
- toxicology
- sub-saharan-africa
- lmic
pretty_name: Snakebite Envenomation Dataset (Species, Syndrome, Antivenom, Complications)
size_categories:
- 10K<n<100K
configs:
- config_name: referral_hospital
  data_files: data/snakebite_referral_hospital.csv
- config_name: district_hospital
  data_files: data/snakebite_district_hospital.csv
  default: true
- config_name: rural_health_centre
  data_files: data/snakebite_rural_health_centre.csv
data_type: synthetic
---

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

# Snakebite Envenomation Dataset

## Abstract

This dataset provides **30,000 synthetic snakebite records** (10,000 per scenario) from sub-Saharan African health facilities. Each record contains 35 variables including snake species (9 taxa), envenomation syndrome (cytotoxic/hemotoxic/neurotoxic), severity grading, clinical features (swelling, coagulopathy, neurotoxicity, necrosis, AKI), antivenom treatment, surgical interventions, and outcomes. Three facility scenarios range from referral hospitals with antivenom (3.6% mortality) to rural health centres without antivenom (22.0% mortality), reflecting the devastating impact of the antivenom access gap.

## 1. Introduction

Snakebite envenomation causes ~315,000 envenomations and ~32,000 deaths annually in SSA (Chippaux, *Toxicon* 2011), with ~7,000 amputations. WHO classified snakebite as a category A neglected tropical disease in 2017. The most medically important genera include *Echis* (carpet vipers — coagulopathy), *Bitis* (puff adders — cytotoxicity), *Naja* (cobras — necrosis/neurotoxicity), and *Dendroaspis* (mambas — rapid neurotoxicity). Antivenom access meets <2% of need in SSA (Schioldann et al., *PLoS NTD* 2018), and mean time to treatment ranges from 6–24 hours.

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

## 2. Methodology

### 2.1 Epidemiological Parameterization

| Parameter | Value | Source |
| --- | --- | --- |
| SSA envenomations/year | ~315,000 | Chippaux, Toxicon 2011 |
| SSA deaths/year | ~32,000 | Chippaux, 2011 |
| Global deaths | 81,000–138,000 | Kasturiratne et al., PLoS Med 2008 |
| Antivenom access (SSA) | <2% of need | Schioldann et al., PLoS NTD 2018 |
| Dry bite rate | ~20% | Warrell, Lancet 2010 |
| Echis coagulopathy rate | 70–80% | Habib et al., PLoS NTD 2015 |
| Case fatality (with antivenom) | 2–5% | Habib et al., 2015 |
| Case fatality (without antivenom) | 10–20% | Multiple SSA studies |
| Antivenom reaction rate | ~15% | Gutierrez et al., Nat Rev 2017 |

### 2.2 Scenario Design

| Scenario | Setting | Antivenom Access | Mean Time (hrs) | Mortality |
| --- | --- | --- | --- | --- |
| Referral hospital | Tertiary, antivenom stock | 70% | 4.8 | 3.6% |
| District hospital | Intermittent supply | 35% | 9.4 | 10.1% |
| Rural health centre | No antivenom, referral needed | 5% | 18.9 | 22.0% |

## 3. Schema

| Column | Type | Description |
| --- | --- | --- |
| id | int | Unique identifier |
| age_years | int | Patient age |
| sex | categorical | M / F |
| occupation | categorical | farmer / herder / hunter / trader / student / child / other |
| bite_location | categorical | foot / ankle / lower_leg / hand / forearm / other |
| activity_at_bite | categorical | farming / walking / sleeping / herding / collecting_firewood / other |
| time_to_facility_hours | float | Hours from bite to facility arrival |
| traditional_treatment_first | binary | Sought traditional healer before hospital |
| snake_species | categorical | 9 taxa (Echis, Bitis, Naja, Dendroaspis, etc.) |
| dry_bite | binary | No envenomation |
| envenomation_syndrome | categorical | cytotoxic / hemotoxic / neurotoxic / mixed / cardiotoxic / none |
| severity | categorical | mild / moderate / severe / none |
| local_swelling | binary | Local tissue swelling |
| local_necrosis | binary | Tissue necrosis at bite site |
| coagulopathy | binary | Venom-induced consumption coagulopathy |
| inr_elevated | binary | Elevated INR |
| bleeding | binary | Clinical bleeding |
| neurotoxicity | binary | Neurotoxic signs |
| ptosis | binary | Drooping eyelids (neurotoxicity marker) |
| respiratory_failure | binary | Respiratory paralysis |
| acute_kidney_injury | binary | AKI |
| compartment_syndrome | binary | Compartment syndrome |
| antivenom_given | binary | Antivenom administered |
| antivenom_vials | int | Number of antivenom vials |
| antivenom_reaction | binary | Adverse reaction to antivenom |
| wound_care | binary | Wound care provided |
| fasciotomy | binary | Fasciotomy performed |
| blood_transfusion | binary | Blood transfusion given |
| mechanical_ventilation | binary | Mechanical ventilation |
| dialysis | binary | Dialysis for AKI |
| amputation | binary | Amputation required |
| antibiotics_given | binary | Antibiotics administered |
| outcome | categorical | survived_full_recovery / survived_with_sequelae / survived_with_disability / died |
| hospital_days | int | Length of hospital stay |

## 4. Validation

<p align="center">
  <img src="validation_report.png" alt="Validation Report" width="100%">
</p>

Key validation checks:

- **Antivenom-mortality inverse**: Referral 70% access/3.6% mortality vs rural 5%/22.0% ✓
- **Severity-mortality gradient**: Mild ~2% → moderate ~10% → severe ~20% ✓
- **Species-syndrome concordance**: Echis→hemotoxic, Bitis→cytotoxic, mamba→neurotoxic ✓
- **Male predominance**: 65% male, consistent with agricultural risk ✓
- **Time to facility**: Rural 19 hrs >> referral 5 hrs ✓

## 5. Usage

```python
from datasets import load_dataset
dataset = load_dataset("electricsheepafrica/snakebite-envenomation", "district_hospital")
df = dataset["train"].to_pandas()
```

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

## 6. Limitations

- **Synthetic**: Not derived from real clinical records.
- **No venom assays**: No venom levels or specific venom component data.
- **Simplified species ID**: Real-world species identification often uncertain.
- **No coagulation kinetics**: No serial INR/PT/APTT trajectories.
- **No wound photographs**: Clinical features are categorical, not visual.
- **No seasonal variation**: Bite incidence does not vary by season/rainfall.

## 7. References

1. Chippaux JP (2011). Estimate of the burden of snakebites in SSA. *Toxicon*, 57(4):586–599.
2. WHO (2019). Snakebite envenoming: a strategy for prevention and control.
3. Habib AG, et al. (2015). Echis ocellatus envenoming in Nigeria. *PLoS NTD*, 9(1):e3479.
4. Warrell DA (2010). Snake bite. *Lancet*, 375(9708):77–88.
5. Harrison RA, et al. (2009). Snakebite: disease of poverty. *PLoS NTD*, 3(12):e569.
6. Kasturiratne A, et al. (2008). Global burden of snakebite. *PLoS Med*, 5(11):e218.
7. Gutierrez JM, et al. (2017). Snakebite envenoming. *Nat Rev Dis Primers*, 3:17063.
8. Schioldann E, et al. (2018). Antivenom availability in SSA. *PLoS NTD*, 12(12):e0006828.

## Citation

```bibtex
@dataset{esa_snakebite_2025,
  title={Snakebite Envenomation Dataset},
  author={Electric Sheep Africa},
  year={2025},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/electricsheepafrica/snakebite-envenomation}
}
```

## License

[CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/)