electricsheepafrica/africa-synth-mortality-heat-related-illness-all

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

Heat-Related Illness & Mortality in Sub-Saharan Africa

Abstract

A synthetic, literature-grounded dataset modelling heat-related morbidity, mortality, and productivity loss across three distinct climatic and health-system scenarios in sub-Saharan Africa (SSA). Each record represents an individual-level heat-health encounter, capturing environmental exposure (daily maximum temperature, Wet-Bulb Globe Temperature, humidity, heatwave status), demographic vulnerability (age, sex, occupation, chronic conditions), health outcomes (diagnosis, severity, hospitalisation, death), labour productivity impacts (work hours lost), and health-system factors (early warning systems, facility disruption, healthcare access).

The dataset is motivated by a critical evidence gap: while SSA bears an estimated 34% of global climate-attributable DALYs, a 2016 systematic review found only 23 studies across the entire region examining temperature–health associations (Azongo et al., 2016), and a 2019 global review found only 10 of 146 heat–health studies were conducted in SSA (Green et al., 2019). Meanwhile, the Lancet Countdown (2024) reports a record 512 billion potential work hours lost globally to heat in 2023, with agricultural workers in tropical regions most affected.

Three scenarios span the SSA heat-health gradient:

• Sahel Extreme (Northern Nigeria, Niger, Mali, Burkina Faso, Chad): Peak temperatures routinely exceeding 45 °C, agriculture-dominant economy with 70% outdoor workers, minimal air-conditioning access (3%), very low early-warning coverage (5%), and limited healthcare access (30%).
• Tropical Humid (Coastal West Africa, Kenya, Tanzania lowlands): Moderate peak temperatures (~33 °C) but high humidity amplifying effective heat stress (WBGT ~29 °C), 55% outdoor workers, emerging early-warning systems (12%), moderate healthcare access (50%).
• Southern Urban (South Africa, Botswana, Namibia cities): Lower baseline temperatures with wider seasonal variation, urban heat island effects, higher NCD burden in an older population, better infrastructure including 20% AC access and 35% early-warning coverage.

Dataset Structure

Each scenario contains 10,000 records (30,000 total across all three). Key columns include:

Demographics

• age, sex, urban, occupation, outdoor_worker, has_chronic_condition, elderly

Environmental Exposure

• daily_tmax_c — Daily maximum temperature (°C)
• daily_tmin_c — Daily minimum temperature (°C)
• wbgt_c — Wet-Bulb Globe Temperature (°C), the gold-standard occupational heat metric
• relative_humidity_pct — Relative humidity (%)
• heat_index_c — Apparent temperature (°C)
• heatwave_day — Binary: part of a heatwave event
• consecutive_hot_days — Duration of heatwave streak
• nighttime_heat — Binary: overnight temperature >25 °C (critical for recovery)
• uhi_effect — Binary: urban heat island amplification

Protective Factors

• ac_access — Air-conditioning access
• shade_available, water_accessible, protective_behaviour

Health Outcomes

• diagnosis — One of: heatstroke, heat_exhaustion, dehydration, heat_cramps, heat_syncope, acute_kidney_injury, cvd_exacerbation, respiratory_exacerbation, heat_rash, none
• heat_illness — Binary: any heat-related diagnosis
• severity — mild / moderate / severe / na
• hospitalised, icu_admission, died

Productivity

• work_hours_lost — Hours of work capacity lost on that day (outdoor workers)
• work_capacity_pct — Remaining work capacity (%)

Health System

• early_warning_received — Received heat-health alert
• health_facility_disrupted — Facility disrupted by heat/power outage
• healthcare_accessible, power_outage

Parameterization Evidence

Parameter	Value Used	Source	Year
34% of global climate-DALYs in SSA	Scenario framing	Azongo et al. Environ Int 91:133-149	2016
Only 23 SSA studies in systematic review of 6,745	Evidence gap justification	Azongo et al.	2016
Only 10 of 146 LMIC heat–health studies in SSA	Evidence gap justification	Green et al. Environ Int 132:105107	2019
92.9% of studies found positive heat–health association	Outcome direction	Green et al.	2019
~489,000 heat-related deaths/year globally	Mortality framing	WHO Heat & Health Fact Sheet	2024
85% increase in heat-related mortality (≥65 y) since 2000	Age vulnerability	WHO / Lancet Countdown	2024
Heatstroke = medical emergency, high CFR	CFR parameterization (10-50% untreated)	WHO	2024
512 billion work hours lost to heat globally in 2023	Productivity loss framing	Lancet Countdown 2024 report	2024
13.8 heatwave-exposure days/person record in 2023	Heatwave frequency	Lancet Countdown	2024
Nighttime heat driving escalated mortality in SSA	Nighttime heat variable	Chersich et al. Sci Adv	2024
WBGT >26 °C: rest 15 min/hr; >28 °C: 30 min; >30 °C: 45 min	WBGT thresholds	OSHA occupational guidelines	—
Elderly, women, low-SES most vulnerable	Vulnerability modifiers	Green et al.	2019
Urban heat island amplifies exposure	UHI variable	Lancet Countdown	2024

Validation Summary

The 8-panel validation report (validation_report.png) confirms:

1. Scenario gradient: Sahel shows highest heat illness (8.0%), hospitalisation (1.4%), and mortality (0.24%); Southern Urban shows lowest (0.9%, 0.3%, 0.01%).
2. Temperature distributions: Sahel mean Tmax ≈ 40 °C; Tropical ≈ 33 °C; Southern ≈ 30 °C — consistent with climatological norms.
3. WBGT thresholds: Sahel distribution centred above the 30 °C OSHA "45-min rest" threshold; Southern centred near the 26 °C first-alert threshold.
4. Diagnosis spectrum: Heatstroke and dehydration dominate in Sahel; lower incidence across all diagnoses in Southern Urban.
5. Occupation profiles: Subsistence farming dominant across all scenarios but highest in Sahel.
6. Protective factors: AC access ranges from ~3% (Sahel) to ~20% (Southern); nighttime heat prevalence highest in Sahel.
7. Productivity loss: Outdoor workers in Sahel lose most work hours; Southern urban workers minimally affected.
8. Correlations: WBGT positively correlated with heat illness and work hours lost; AC access negatively correlated with illness.

Usage

from datasets import load_dataset

# Load the tropical humid scenario (default)
ds = load_dataset("electricsheepafrica/heat-related-illness", name="tropical_humid")
df = ds["train"].to_pandas()

# Explore heat illness by occupation
print(df.groupby("occupation")["heat_illness"].mean().sort_values(ascending=False))

# Filter to heatwave days only
heatwave_df = df[df["heatwave_day"] == 1]
print(f"Heatwave illness rate: {heatwave_df['heat_illness'].mean()*100:.1f}%")

Intended Uses

• Training climate–health risk prediction models for SSA contexts
• Exploring occupational heat stress and productivity loss in agricultural economies
• Evaluating the potential impact of early warning systems and cooling interventions
• Teaching and capacity-building on planetary health and climate epidemiology
• Benchmarking heat vulnerability indices across different African climatic zones

Limitations

• Synthetic data: Generated from literature-derived parameter distributions, not from real patient records. Statistical relationships are modelled, not observed.
• Parameter uncertainty: SSA-specific heat–health evidence is sparse (only 23 studies in the Azongo 2016 review). Parameters for heatstroke incidence and CFR are extrapolated from global literature.
• Simplified climate model: Daily temperatures are drawn from normal distributions; real climate exhibits autocorrelation, seasonality, and extreme-value behaviour not fully captured here.
• No spatial granularity: Records are not geolocated; within-scenario heterogeneity (e.g., rural vs. urban Sahel) is modelled probabilistically, not spatially.
• Single-day encounters: Each record represents one day; cumulative multi-day heat exposure effects are approximated via the consecutive_hot_days variable but not mechanistically modelled.

References

1. Azongo DK, Awine T,'; et al. (2016). Temperature-related morbidity and mortality in Sub-Saharan Africa: A systematic review of the empirical evidence. Environ Int 91:133-149. DOI: 10.1016/j.envint.2016.02.027
2. Green H, Bailey J, Schwarz L, et al. (2019). Impact of heat on mortality and morbidity in low and middle income countries: A review of the epidemiological evidence. Environ Int 132:105107. DOI: 10.1016/j.envint.2019.105107
3. WHO (2024). Heat and health. Fact sheet. https://www.who.int/news-room/fact-sheets/detail/climate-change-heat-and-health
4. Romanello M, di Napoli C, Green C, et al. (2024). The 2024 report of the Lancet Countdown on health and climate change. Lancet 404:2072-2098. DOI: 10.1016/S0140-6736(24)01822-1
5. Lancet Countdown (2024). Heat and health indicators. https://lancetcountdown.org/heat-and-health/
6. Chersich MF, et al. (2024). Escalated heatwave mortality risk in sub-Saharan Africa under recent warming. Sci Adv. DOI: 10.1126/sciadv.ady7379

Citation

If you use this dataset, please cite:

@dataset{electricsheepafrica_heat_related_illness_2025,
  title={Heat-Related Illness and Mortality in Sub-Saharan Africa},
  author={Electric Sheep Africa},
  year={2025},
  publisher={HuggingFace},
  url={https://huggingface.co/datasets/electricsheepafrica/heat-related-illness}
}

License

CC-BY-4.0