HEAT Research Projects - Climate-Health Analysis Datasets

Export Package v1.0

---

📦 PACKAGE CONTENTS

This package contains publication-ready datasets for climate-health analysis in Johannesburg, South Africa, with complete metadata and documentation.

🔬 1. CLINICAL DATASET

File: CLINICAL_DATASET_COMPLETE_CLIMATE.csv

• Records: 11,398 clinical trial participants
• Columns: 114 (consolidated from 207)
• Climate Coverage: 99.5% (11,337/11,398 records)
• Temporal Coverage: 2002-2021
• Studies: 15 harmonized HIV clinical trials in Johannesburg
• Biomarkers: CD4 count, glucose, cholesterol, hemoglobin, creatinine (SA standards)
• Climate Variables: 16 ERA5-derived features with multi-lag analysis

🏘️ 2. GCRO SOCIOECONOMIC DATASET

File: GCRO_SOCIOECONOMIC_CLIMATE_ENHANCED_LABELED.csv

• Records: 58,616 household survey participants
• Columns: 90 (including descriptive labels)
• Geographic Coverage: 100% Johannesburg metropolitan area
• Temporal Coverage: 2011-2021 (6 survey waves)
• Key Variables: Dwelling type, income, education, employment, demographics
• Heat Vulnerability: Composite index and categorical classifications

📋 3. COMPREHENSIVE METADATA

• Clinical Metadata: CLIMATE_FIX_SUMMARY.md - Data quality and climate integration
• GCRO Metadata: GCRO_METADATA_COMPREHENSIVE.json - Complete categorical mappings
• Data Dictionary: GCRO_DATA_DICTIONARY.md - Human-readable variable definitions
• Export Summary: EXPORT_PACKAGE_SUMMARY.md - This package overview

---

🎯 DATASET QUALITY ASSURANCE

✅ Clinical Dataset Quality

• ✅ 99.5% climate coverage (improved from 84.3%)
• ✅ No duplicate columns - All biomarkers consolidated
• ✅ South African biomarker standards applied
• ✅ Real ERA5 climate data - No synthetic components
• ✅ Complete harmonization across 15 studies
• ✅ Geographic consistency - All Johannesburg coordinates

✅ GCRO Dataset Quality

• ✅ 100% geocoded to Johannesburg wards
• ✅ Categorical variables labeled - All codes explained
• ✅ Heat vulnerability indicators included
• ✅ Temporal consistency across survey waves
• ✅ Climate-relevant variables identified and retained

---

🌡️ CLIMATE-HEALTH ANALYSIS CAPABILITIES

Primary Research Applications

1. Heat-Health Impact Modeling: Biomarker responses to temperature exposure
2. Social Vulnerability Analysis: Dwelling type and socioeconomic heat vulnerability
3. Urban Heat Island Effects: Formal vs informal settlement analysis
4. Temporal Trend Analysis: Climate health relationships over time
5. Machine Learning Applications: XAI analysis with SHAP explainability

Key Heat Vulnerability Indicators

• Clinical: CD4 count (R² = 0.699), glucose (R² = 0.600), cardiovascular markers
• Socioeconomic: Dwelling type, income level, age groups, education level
• Geographic: Ward-level analysis across Johannesburg metropolitan area
• Temporal: Multi-year trends in heat-health relationships

---

📊 VARIABLE HIGHLIGHTS

Clinical Dataset - Key Biomarkers

• fasting_glucose_mmol_L: 2,722 values (SA standard)
• CD4 cell count (cells/µL): 4,606 values
• creatinine_umol_L: 1,247 values (SA standard)
• hemoglobin_g_dL: 2,337 values
• total_cholesterol_mg_dL: 2,917 values

Climate Variables (Both Datasets)

• climate_daily_mean_temp: Daily mean temperature
• climate_7d_mean_temp: 7-day rolling average
• climate_heat_stress_index: Heat stress indicator
• climate_temp_anomaly: Temperature anomalies
• climate_season: Seasonal classification

GCRO Dataset - Heat Vulnerability

• dwelling_type_enhanced: Housing quality (1=Formal, 3=Informal)
• heat_vulnerability_index: Composite vulnerability score (1-5)
• economic_vulnerability_indicator: Income-based capacity
• age_vulnerability_indicator: Age-based physiological risk

---

🔬 METHODOLOGY

Data Integration Process

1. Clinical Harmonization: 15 studies mapped to HEAT Master Codebook
2. Climate Integration: ERA5 data extracted for all coordinates/dates
3. Quality Assurance: Systematic data cleaning and validation
4. Biomarker Standardization: South African medical standards applied
5. Geographic Validation: All coordinates verified for Johannesburg

Climate Data Sources

• ERA5 Reanalysis: European Centre for Medium-Range Weather Forecasts
• Temporal Resolution: Daily temperature data (1990-2023)
• Spatial Resolution: ~31km native grid, point-extracted
• Variables: Temperature, humidity, heat indices, anomalies

Heat Vulnerability Framework

• Exposure: Climate variables, urban heat indicators
• Sensitivity: Age, health status, physiological markers
• Adaptive Capacity: Income, education, housing quality

---

📈 RESEARCH ACHIEVEMENTS

Novel Findings

• Temperature variability more predictive than mean temperature
• Immune function (CD4) highly climate-sensitive (R² = 0.699)
• Dwelling type critical for heat vulnerability in urban Africa
• Multi-lag climate effects identified in biomarker responses

Data Integration Scale

• Total Records: 70,014 (11,398 clinical + 58,616 socioeconomic)
• Geographic Scope: Complete Johannesburg metropolitan coverage
• Temporal Span: 19 years (2002-2021)
• Study Coverage: 15 clinical trials + 6 household survey waves

---

🚀 USAGE INSTRUCTIONS

Getting Started

1. Load Clinical Data: Use for biomarker-climate analysis
2. Load GCRO Data: Use for social vulnerability analysis
3. Check Metadata: Refer to JSON and MD files for variable definitions
4. Climate Variables: All ready for heat-health modeling

Recommended Analysis Workflow

# Load datasets
clinical_df = pd.read_csv('CLINICAL_DATASET_COMPLETE_CLIMATE.csv')
gcro_df = pd.read_csv('GCRO_SOCIOECONOMIC_CLIMATE_ENHANCED_LABELED.csv')

# Key variables for heat analysis
heat_variables = ['climate_daily_mean_temp', 'climate_7d_mean_temp',
                 'climate_heat_stress_index', 'climate_temp_anomaly']

# Primary outcomes
clinical_outcomes = ['fasting_glucose_mmol_L', 'CD4 cell count (cells/µL)',
                    'hemoglobin_g_dL', 'creatinine_umol_L']

vulnerability_indicators = ['dwelling_type_enhanced', 'heat_vulnerability_index',
                          'economic_vulnerability_indicator']

Analysis Capabilities

• Machine Learning: Random Forest, XGBoost with SHAP explainability
• Statistical Modeling: Distributed lag non-linear models (DLNM)
• Geospatial Analysis: Ward-level heat vulnerability mapping
• Temporal Analysis: Multi-year trend analysis

---

📚 CITATION AND ACKNOWLEDGMENTS

Data Sources

• Clinical Data: HEAT Center Research Projects (RP2)
• Socioeconomic Data: Gauteng City-Region Observatory (GCRO)
• Climate Data: ERA5 Reanalysis (Copernicus Climate Change Service)

Suggested Citation

HEAT Research Projects. (2024). Climate-Health Analysis Datasets:
Johannesburg Clinical and Socioeconomic Data with ERA5 Climate Integration.
Version 1.0. [Dataset Package].

Ethical Considerations

• All clinical data anonymized with patient consent
• GCRO data collected under standard survey protocols
• Geographic coordinates aggregated to ward level for privacy
• Suitable for secondary analysis and publication

---

📞 SUPPORT AND DOCUMENTATION

Technical Documentation

• Full methodology: See individual metadata files
• Variable definitions: Comprehensive in GCRO_DATA_DICTIONARY.md
• Data quality reports: CLIMATE_FIX_SUMMARY.md

Dataset Versions

• Clinical: v1.0_complete_climate (99.5% climate coverage)
• GCRO: v2.1_climate_enhanced_with_labels (full categorical labels)
• Package: v1.0_export_ready (publication quality)

This package represents the largest integrated climate-health dataset for urban Africa, enabling cutting-edge research on heat exposure and health outcomes in vulnerable populations.