Skip to article frontmatterSkip to article content
Site not loading correctly?

This may be due to an incorrect BASE_URL configuration. See the MyST Documentation for reference.

Data Inventory

Data Collection Summary

Smoking/Tobacco Use Indicators (Downloaded)

Three smoking prevalence indicators have been downloaded from WHO GHO API:

  1. M_Est_smk_curr_std - Age-standardized current tobacco smoking prevalence (%)

    • Records: 5,181 (includes projections)

    • Years: 2000-2030 (observed: 2000, 2005, 2007, 2010, 2015, 2018, 2020, 2021, 2022; projected: 2025, 2030)

    • Countries: 172

    • Sex categories: Both sexes, Female, Male

    • File: data/who_smoking_data.csv

    • Status: ✅ Recommended - age-standardized, good temporal coverage

    • Note: Years 2025 and 2030 are projections (marked in Comments field). Filter these out for analysis of observed data only.

  2. M_Est_cig_curr_std - Age-standardized current cigarette smoking prevalence (%)

    • Records: 4,950 (likely includes projections)

    • Years: 2000-2030 (includes projected years 2025, 2030)

    • Countries: 165

    • Sex categories: Both sexes, Female, Male

    • File: data/who_smoking_cigarette_std.csv

    • Status: ✅ Good - cigarette-specific, age-standardized

    • Note: Check Comments field to identify projected vs. observed data

  3. Adult_curr_tob_smoking - Current tobacco smoking among adults (%)

    • Records: 570

    • Years: 2001-2022

    • Countries: 190 (most countries)

    • Sex categories: Both sexes, Female, Male

    • File: data/who_smoking_adult.csv

    • Status: ✅ Good country coverage but fewer records and shorter time span

Recommendation: Use M_Est_smk_curr_std as the primary smoking predictor - it has the best combination of temporal coverage, age-standardization, and sufficient country coverage.

Suicide Rate Indicators (Identified)

Five suicide-related indicators have been identified from WHO GHO API:

  1. MH_12 - Age-standardized suicide rates (per 100,000 population)

    • Records: 12,936

    • Years: 2000-2021

    • Countries: 196

    • Sex categories: Both sexes, Female, Male

    • Status: ✅ Recommended - age-standardized, excellent country coverage, good temporal coverage

    • Note: Age-standardized rates are preferred for HALE analysis since HALE is also age-standardized

  2. SDGSUICIDE - Crude suicide rates (per 100,000 population)

    • Records: 19,041

    • Years: 2000-2021

    • Countries: 196

    • Sex categories: Both sexes, Female, Male

    • Status: ✅ Good - crude rates, excellent country coverage

    • Note: Crude rates may be less comparable across countries with different age structures

  3. SDG_SH_STA_SCIDEN - Number of suicide deaths

    • Status: ⚠️ Less useful - absolute numbers rather than rates

  4. PRISON_D3_DEATHS_SUICIDE_MRATE - In-prison suicide mortality rate

    • Status: ⚠️ Not relevant - prison-specific, not general population

  5. PRISON_B16_SUICIDERISK - In-prison standardized protocol for suicide

    • Status: ⚠️ Not relevant - protocol indicator, not a rate

Recommendation: Use MH_12 as the primary suicide rate predictor - it has age-standardized rates (matching HALE methodology), excellent country coverage (196 countries), gender breakdowns, and good temporal coverage (2000-2021).

Alcohol-Attributable Death Rate Indicators (Identified)

Multiple alcohol-related death rate indicators have been identified from WHO GHO API:

  1. SA_0000001832 - Alcohol-attributable all-cause deaths per 100,000, age standardized

    • Records: 540

    • Years: 2019

    • Countries: 180

    • Sex categories: Both sexes, Female, Male

    • Status: ⚠️ Downloaded but not used in final model - replaced with IHME Alcohol Use Disorders (B.7.1) for better temporal coverage

    • Note: Age-standardized rates match HALE methodology. This indicator uses Population Attributable Fraction (PAF) methodology to estimate all deaths where alcohol is a contributing factor, including:

      • Direct alcohol-related deaths (alcohol poisoning, alcohol dependence syndrome, alcohol withdrawal)

      • Indirect alcohol-related deaths where alcohol is a contributing factor:

        • Liver disease (cirrhosis, alcoholic liver disease)

        • Some cancers (oral, pharyngeal, esophageal, liver, colorectal, breast)

        • Accidents and injuries (road traffic crashes, falls, drownings) where alcohol was involved

        • Violence (homicide, suicide) where alcohol was a contributing factor

        • Cardiovascular diseases where alcohol contributed

        • Other conditions where alcohol is a risk factor

    • Definitional Difference from IHME: WHO’s “alcohol-attributable” definition is much broader than IHME’s “alcohol use disorders” definition. WHO includes indirect alcohol-related deaths (e.g., liver disease deaths attributable to alcohol, even if liver disease is listed as the primary cause), while IHME only includes deaths where alcohol use disorders are the primary cause of death. This explains why WHO alcohol gap values are much higher than IHME values (e.g., USA: 38.8 vs 5.54, an 86% difference). See alcohol_data_comparison.md for detailed explanation.

    • Limitation: Only has data for 2019, which limits temporal analysis but provides a good snapshot for cross-country comparison.

  2. SA_0000001437 - Age-standardized death rates, alcohol use disorders, per 100,000

    • Records: 714

    • Years: 2002, 2004 (only 2 years)

    • Countries: 186

    • Sex categories: Both sexes, Female, Male

    • Status: ⚠️ Limited temporal coverage - only 2 years of data, older years (2002, 2004)

    • Note: More specific than SA_0000001832 (focuses on alcohol use disorders rather than all alcohol-attributable deaths), but limited temporal coverage makes it less useful for analysis.

  3. SA_0000001833 - Alcohol-attributable DALYs per 100,000 people (age standardized)

    • Years: 2019

    • Countries: 182

    • Records: 1,092

    • Sex categories: Both sexes, Female, Male

    • Status: ✅ Good - DALYs (Disability-Adjusted Life Years) provide a measure of both mortality and morbidity, but death rates are more directly comparable to HALE

  4. SA_0000001457_AA - Liver cirrhosis, alcohol-attributable, age-standardized death rates

    • Years: 2019

    • Countries: 180

    • Records: 1,080

    • Sex categories: Both sexes, Female, Male

    • Status: ✅ Good - specific cause of death, but narrower scope than all-cause alcohol-attributable deaths

Recommendation: The model uses IHME Alcohol Use Disorders (B.7.1) instead of WHO SA_0000001832 because IHME provides much better temporal coverage (1990-2023 vs 2019 only) and consistent methodology with other IHME indicators. However, it’s important to note that IHME’s definition is much narrower (only direct alcohol use disorder deaths) compared to WHO’s broader “alcohol-attributable” definition (which includes indirect alcohol-related deaths like liver disease, some cancers, and accidents where alcohol was involved). This definitional difference explains why alcohol gap values are much lower in IHME data (e.g., USA: 5.54 vs 38.8 in WHO, an 86% difference) and why alcohol importance decreased when switching from WHO to IHME data. See alcohol_data_comparison.md for detailed explanation of these definitional differences.

Unintentional Poisoning Mortality Rate Indicators (Identified)

Multiple unintentional poisoning-related indicators have been identified from WHO GHO API:

  1. SDGPOISON - Mortality rate attributed to unintentional poisoning (per 100,000 population)

    • Records: 12,936

    • Years: 2000-2021 (22 years)

    • Countries: 196

    • Sex categories: Both sexes, Female, Male

    • Status: ✅ Recommended - excellent temporal coverage, excellent country coverage, gender breakdowns, includes confidence intervals

    • Note: This is a crude rate (not explicitly age-standardized), but has excellent temporal and country coverage. Unintentional poisoning includes accidental poisonings from chemicals, drugs, and other substances, which can contribute to the gender gap in mortality. Men often have higher rates of accidental deaths, including poisonings.

  2. SA_0000001450 - Age-standardized death rates, poisoning, per 100,000

    • Records: 731

    • Years: 2002, 2004 (only 2 years)

    • Countries: 185

    • Sex categories: Both sexes, Female, Male

    • Status: ⚠️ Limited temporal coverage - only 2 years of data, older years (2002, 2004)

    • Note: Age-standardized rates are preferred for HALE analysis, but limited temporal coverage makes it less useful than SDGPOISON.

  3. SA_0000001458 - Age-standardized death rates (15+ years), poisoning, per 100,000

    • Years: 2002, 2004 (only 2 years)

    • Status: ⚠️ Limited temporal coverage - similar to SA_0000001450 but for ages 15+

  4. SA_0000001837 - Alcohol poisoning deaths, per 100,000 population

    • Status: ⚠️ Narrow scope - only alcohol-related poisonings, not all unintentional poisonings

Recommendation: Use SDGPOISON as the primary unintentional poisoning mortality rate predictor - it has excellent temporal coverage (2000-2021), excellent country coverage (196 countries), gender breakdowns, and includes confidence intervals. While it’s not explicitly age-standardized, the comprehensive temporal and country coverage make it more valuable for analysis than the age-standardized indicators with only 2 years of data. Unintentional poisoning is relevant to the gender gap as men often have higher rates of accidental deaths.

IHME Global Burden of Disease Data

Data downloaded from IHME Global Burden of Disease (GBD) Compare tool: https://vizhub.healthdata.org/gbd-compare/

Note: IHME data provides separate male and female files, allowing for gender gap analysis. Data is downloaded for OECD countries only. All indicators use “All ages” to match HALE methodology (calculated from birth). All IHME indicators include separate male and female values, allowing for gender gap analysis. Country names in IHME data use “Republic of Korea” and “United States of America” which are mapped to “South Korea” and “United States” respectively for compatibility with WHO country name mappings.

Drug Use Disorders Death Rates (Downloaded)

Indicator: B.7.2 Drug Use Disorders
Measure: Deaths
Metric: Rate (per 100,000 population)
Locations: OECD countries
Age: All ages
Sex: Separate files for Male and Female

GBD Compare Tool Settings:

Files:

Status: ✅ Downloaded and integrated into model
Relevance: Drug overdoses, particularly opioid overdoses, are a major cause of death in some OECD countries (especially the US) and may contribute significantly to the HALE gender gap. This indicator captures overdose deaths that may not be fully captured in the WHO poisoning indicator.
Model Results: Drug Use Disorders has importance = 0 in both models, meaning it is not selected by Elastic Net and does not contribute to explaining gender gaps. This suggests that drug-related mortality may not be a major factor in explaining gender gaps in Life Expectancy or HALE, at least with the current data and model structure. The WHO poisoning indicator (SDGPOISON) was removed from the model, and Drug Use Disorders remains but is not selected. See validation.md section “Removing WHO Poisoning: Keeping Only IHME DrugDisorder” for detailed analysis.

Alcohol Use Disorders Death Rates (Downloaded)

Indicator: B.7.1 Alcohol use disorders
Measure: Deaths
Metric: Rate (per 100,000 population)
Locations: OECD countries
Age: All ages
Sex: Separate files for Male and Female

GBD Compare Tool Settings:

Files:

Status: ✅ Downloaded and integrated into model
Relevance: Alcohol use disorders are a significant cause of death and may contribute to the HALE gender gap. Men typically have higher rates of alcohol-related mortality than women. This indicator provides comprehensive alcohol use disorder death rates with excellent temporal coverage (1990-2023, 34 years) and good country coverage (40 countries). This is used in the model instead of the WHO alcohol-attributable death rate indicator (SA_0000001832) which only has data for 2019. IHME data provides much better temporal coverage, allowing for more recent data and temporal analysis.

Definitional Difference from WHO: IHME uses “alcohol use disorders” which refers to deaths where alcohol use disorders are the primary or direct cause of death (ICD-10 F10 codes). This includes:

What IHME Excludes (that WHO includes):

Why the Difference Matters: The IHME definition is much narrower than WHO’s “alcohol-attributable” definition, which explains why IHME alcohol gap values are much lower than WHO values (e.g., USA: 5.54 vs 38.8, an 86% difference). This narrower definition also explains why Alcohol dropped from #1 to lower importance when switching from WHO to IHME data - the IHME definition captures a much smaller subset of alcohol-related mortality. However, IHME’s better temporal coverage (1990-2023 vs 2019 only) and consistent methodology with other IHME indicators make it preferable for the current analysis. See alcohol_data_comparison.md for detailed explanation of these definitional differences.

Self-Harm (Suicide) Death Rates (Downloaded)

Indicator: B.7.3 Self-harm
Measure: Deaths
Metric: Rate (per 100,000 population)
Locations: OECD countries
Age: All ages
Sex: Separate files for Male and Female

GBD Compare Tool Settings:

Files:

Status: ✅ Downloaded and integrated into model
Relevance: Self-harm (suicide) is a significant cause of death and contributes to the HALE gender gap. Men typically have much higher suicide rates than women in most countries. This indicator provides comprehensive self-harm death rates with excellent temporal coverage (1990-2023, 34 years) and good country coverage (40 countries). This is used in the model instead of the WHO suicide rate indicator (MH_12) which has data for 2000-2021. IHME data provides better temporal coverage (starting from 1990) and consistent methodology with other IHME indicators.
Model Results: Suicide importance increased substantially when switching from WHO to IHME data (+139% for Life Expectancy, +42% for HALE), suggesting IHME data captures suicide-related mortality more effectively. Suicide ranks #4 in Life Expectancy model and #5 in HALE model. See validation.md section “Suicide: WHO → IHME” for detailed analysis.

Interpersonal Violence (Homicide) Death Rates (Downloaded)

Indicator: B.7.4 Interpersonal violence
Measure: Deaths
Metric: Rate (per 100,000 population)
Locations: OECD countries
Age: All ages
Sex: Separate files for Male and Female

GBD Compare Tool Settings:

Files:

Status: ✅ Downloaded and integrated into model
Relevance: Interpersonal violence (homicide) is a significant cause of death and contributes to the HALE gender gap. Men typically have much higher homicide rates than women in most countries. This indicator provides comprehensive interpersonal violence death rates with excellent temporal coverage (1990-2023, 34 years) and good country coverage (40 countries). This is used in the model instead of the WHO homicide rate indicator (VIOLENCE_HOMICIDERATE) which has data for 2000-2021. IHME data provides better temporal coverage (starting from 1990) and consistent methodology with other IHME indicators.
Model Results: Homicide importance decreased when switching from WHO to IHME data (-28% for HALE), and homicide was not selected by Elastic Net for the Life Expectancy model (importance = 0), meaning it does not contribute to explaining the Life Expectancy gap when using IHME data. For HALE, homicide ranks #5 with moderate importance (3.04). This suggests that IHME homicide data may be less predictive than WHO data, or that other indicators (particularly Suicide) capture similar variance. See validation.md section “Homicide: WHO → IHME” for detailed analysis.

Road Injuries (Road Traffic Crash) Death Rates (Downloaded)

Indicator: Road injuries
Measure: Deaths
Metric: Rate (per 100,000 population)
Locations: OECD countries
Age: All ages
Sex: Separate files for Male and Female

GBD Compare Tool Settings:

Files:

Status: ✅ Downloaded and integrated into model
Relevance: Road injuries (road traffic crashes) are a significant cause of death and contribute to the HALE gender gap. Men typically have 2-4 times higher road traffic death rates than women in most countries due to higher exposure to driving (including occupational exposure), occupational hazards, and potentially risk-taking behaviors. This indicator provides comprehensive road injury death rates with excellent temporal coverage (1990-2023, 34 years) and good country coverage (40 countries). This is used in the model instead of the WHO road traffic crash death rate indicator (SA_0000001459) which only has data for 2019. IHME data provides much better temporal coverage, allowing for temporal analysis and more recent data.
Model Results: Road traffic has very low importance in both models (0.111 for Life Expectancy, ranked #8; 0.633 for HALE, ranked #9), suggesting it is not a major predictive factor for gender gaps. For Life Expectancy, only the Mid component was selected (Gap component = 0), meaning the gender gap in road traffic deaths does not contribute to explaining the Life Expectancy gap. See validation.md section “Road Traffic: WHO → IHME” for detailed analysis.

Maternal Disorders Death Rates (Downloaded)

Indicator: Maternal disorders
Measure: Deaths
Metric: Rate (per 100,000 population)
Locations: OECD countries
Age: All ages
Sex: Female only (inherently female-specific)

GBD Compare Tool Settings:

Files:

Status: ⚠️ Downloaded but removed from final model - removed due to counterintuitive positive coefficient
Relevance: Maternal disorders (maternal mortality) are a significant cause of death for women and can contribute to the HALE gender gap, especially in lower-income countries. High maternal mortality can significantly reduce female life expectancy, explaining why some countries have smaller gender gaps. This indicator provides comprehensive maternal disorder death rates with excellent temporal coverage (1990-2023, 34 years) and good country coverage (40 countries). This is an alternative to the WHO maternal mortality ratio indicator (MDG_0000000026) which has data for 1985-2023. Note: WHO indicator uses ratio per 100,000 live births, while IHME uses rate per 100,000 population, so they measure slightly different things.
Why Removed: Maternal mortality had a counterintuitive positive coefficient in the models, which implies that higher maternal mortality is associated with a larger LE/HALE gap. This is counterintuitive because if something increases female mortality, it should close the gap (since gap = Female - Male). The positive coefficient suggests a spurious association, possibly because maternal mortality is capturing something about general healthcare quality rather than a direct causal relationship. Removing it had minimal impact on model performance but improved interpretability. After removal, Cardiovascular and Homicide gained substantial importance in the HALE model, suggesting maternal mortality may have been suppressing these indicators. See validation.md section “Removing Maternal Mortality Indicator” for detailed analysis.
Note: Maternal disorders are inherently female-specific (deaths during pregnancy, childbirth, or within 42 days of termination of pregnancy).

All-Cause Deaths Under 5 Years of Age (Downloaded)

Indicator: All causes (under 5 years)
Measure: Deaths
Metric: Rate (per 100,000 population)
Locations: OECD countries
Age: <5 years
Sex: Separate files for Male and Female

GBD Compare Tool Settings:

Files:

Status: ⚠️ Downloaded but not used in final model - removed due to methodological concerns
Relevance: All-cause mortality for children under 5 years of age is relevant to the HALE gender gap because HALE is calculated from birth, so early-life mortality directly affects HALE calculations. If child mortality differs by gender, it directly contributes to the HALE gender gap. Infant and child mortality is typically higher in males (biological vulnerability + some behavioral factors). This indicator provides comprehensive all-cause under-5 mortality rates with excellent temporal coverage (1990-2023, 34 years) and good country coverage (40 countries).

Definitional Difference from WHO: This indicator measures deaths per 100,000 population, which is fundamentally different from the WHO under-five mortality rate (MDG_0000000007) which measures deaths per 1,000 live births.

Methodological Concern - Confounding: The IHME indicator (deaths per 100,000 population) is confounded with age structure and fertility rates. Countries with:

will have more people under age 5 in the population, and therefore more deaths under 5, even if the underlying risk of death for children is the same. This confounding makes it difficult to interpret the IHME indicator as a pure measure of early-life mortality risk. The WHO indicator (deaths per 1,000 live births) controls for these factors by using live births as the denominator, making it a more direct measure of early-life mortality risk independent of demographic structure.

Why Removed: Both WHO and IHME under-five mortality indicators were removed from the final model because:

See validation.md section “Removing Childhood Indicator (Under-Five Mortality)” for detailed analysis of the removal and its effects.

Diabetes Type 2 Death Rates (Downloaded)

Indicator: B.8.1.2 Diabetes mellitus type 2
Measure: Deaths
Metric: Rate (per 100,000 population)
Locations: OECD countries
Age: All ages
Sex: Separate files for Male and Female

GBD Compare Tool Settings:

Files:

Status: ⚠️ Downloaded but not yet integrated into model
Note: This is an alternative to the WHO diabetes death rate indicator (SA_0000001440) which only has data for 2004. IHME data may have better temporal coverage.

Cardiovascular Diseases Death Rates (Downloaded)

Indicator: B.2 Cardiovascular diseases
Measure: Deaths
Metric: Rate (per 100,000 population)
Locations: OECD countries
Age: All ages
Sex: Separate files for Male and Female

GBD Compare Tool Settings:

Files:

Status: ⚠️ Downloaded but not yet integrated into model
Relevance: Cardiovascular diseases are a major cause of death and may contribute significantly to the HALE gender gap. This is an alternative to the WHO cardiovascular disease death rate indicators which only have data for 2004. IHME data may have better temporal coverage, allowing for more recent data to be used in the analysis.

Neoplasms (Cancer) Death Rates (Downloaded)

Indicator: B.1 Neoplasms
Measure: Deaths
Metric: Rate (per 100,000 population)
Locations: OECD countries
Age: All ages
Sex: Separate files for Male and Female

GBD Compare Tool Settings:

Files:

Status: ⚠️ Downloaded but not yet integrated into model
Relevance: Neoplasms (cancer) are a major cause of death and may contribute significantly to the HALE gender gap. Different types of cancer have different gender patterns (e.g., lung cancer is often higher in men, breast cancer is female-specific). This indicator provides comprehensive cancer death rates with better temporal coverage than WHO indicators.

Chronic Respiratory Diseases Death Rates (Downloaded)

Indicator: B.3 Chronic respiratory diseases
Measure: Deaths
Metric: Rate (per 100,000 population)
Locations: OECD countries
Age: All ages
Sex: Separate files for Male and Female

GBD Compare Tool Settings:

Files:

Status: ⚠️ Downloaded but not yet integrated into model
Relevance: Chronic respiratory diseases (including COPD, asthma, and other chronic lung conditions) are a major cause of death and may contribute significantly to the HALE gender gap. These diseases often have gender differences due to factors such as smoking patterns, occupational exposures, and environmental factors. This indicator provides comprehensive chronic respiratory disease death rates with better temporal coverage than WHO indicators.

Liver Disease (Cirrhosis and Other Chronic Liver Diseases) Death Rates (Downloaded)

Indicator: Cirrhosis and other chronic liver diseases
Measure: Deaths
Metric: Rate (per 100,000 population)
Locations: OECD countries
Age: All ages
Sex: Separate files for Male and Female

GBD Compare Tool Settings:

Files:

Status: ✅ Downloaded and integrated into model
Relevance: Liver disease (cirrhosis and other chronic liver diseases) is a significant cause of death and may contribute to the HALE gender gap. Men typically have higher rates of liver disease mortality than women, often due to higher alcohol consumption, hepatitis infections, and other risk factors. This indicator provides comprehensive liver disease death rates with excellent temporal coverage (1990-2023, 34 years) and good country coverage. Liver disease is often related to alcohol consumption, but also includes non-alcoholic causes such as viral hepatitis, non-alcoholic fatty liver disease, and other chronic liver conditions.

COVID-19 Death Rates (Downloaded)

Indicator: COVID-19
Measure: Deaths
Metric: Rate (per 100,000 population)
Locations: OECD countries
Age: All ages
Sex: Separate files for Male and Female

GBD Compare Tool Settings:

Files:

Status: ✅ Downloaded and integrated into model
Relevance: COVID-19 is a significant cause of death that emerged in 2020 and may contribute to the HALE gender gap. COVID-19 mortality patterns show gender differences, with men typically having higher death rates than women in most countries. This indicator provides comprehensive COVID-19 death rates with temporal coverage from 2020-2023. Note: Data includes zeros for all years before 2020 (1990-2019) since COVID-19 did not exist before 2020. This indicator is particularly relevant for understanding recent changes in the gender gap in life expectancy and HALE, as the pandemic had substantial impacts on mortality patterns.
Note: Years 1990-2019 contain zeros (COVID-19 did not exist), with actual data starting in 2020.

Unintentional Injuries Death Rates (Downloaded)

Indicator: Unintentional injuries
Measure: Deaths
Metric: Rate (per 100,000 population)
Locations: OECD countries
Age: All ages
Sex: Separate files for Male and Female

GBD Compare Tool Settings:

Files:

Status: ✅ Downloaded and integrated into model
Relevance: Unintentional injuries (including falls, drowning, fires, and other accidents) are a significant cause of death and may contribute to the HALE gender gap. These injuries often show gender differences due to occupational exposures, risk-taking behaviors, and activity patterns. This indicator provides comprehensive unintentional injury death rates with better temporal coverage (1990-2023) than many WHO indicators.

Road Traffic Crash Death Rate Indicators (Identified)

Multiple road traffic-related death rate indicators have been identified from WHO GHO API:

  1. SA_0000001459 - Road traffic crash deaths, age-standardized death rates (15+), per 100,000 population

    • Records: 1,080

    • Years: 2019

    • Countries: 180

    • Sex categories: Both sexes, Female, Male

    • Status: ✅ Recommended - age-standardized, good country coverage, gender breakdowns, recent data (2019)

    • Note: Age-standardized rates for ages 15+ match HALE methodology (HALE is also age-standardized). Road traffic deaths are a major contributor to the gender gap in mortality, as men typically have much higher rates due to higher exposure to driving (including occupational exposure), occupational hazards, and potentially risk-taking behaviors. The limitation is that it only has data for 2019, but this provides a good cross-sectional snapshot for the analysis.

  2. RS_198 - Estimated road traffic death rate (per 100,000 population)

    • Years: 2021 (only 1 year)

    • Countries: 204

    • Sex categories: None (no gender breakdown)

    • Status: ⚠️ Not suitable - no gender breakdown available

  3. SA_0000001452 - Age-standardized death rates, road traffic accidents, per 100,000

    • Years: 2002, 2004 (only 2 years)

    • Countries: 192

    • Sex categories: Both sexes, Female, Male

    • Status: ⚠️ Limited temporal coverage - only 2 years of older data (2002, 2004)

  4. SA_0000001459_AA - Road traffic crash deaths, alcohol-attributable, age-standardized death rates

    • Years: 2019

    • Countries: 180

    • Sex categories: Both sexes, Female, Male

    • Status: ⚠️ Narrow scope - only alcohol-attributable road traffic deaths, not all road traffic deaths

Recommendation: The model uses IHME Road Injuries instead of WHO SA_0000001459 because IHME provides much better temporal coverage (1990-2023 vs 2019 only) and consistent methodology with other IHME indicators. However, road traffic has very low importance in both models (0.111 for Life Expectancy, ranked #8; 0.633 for HALE, ranked #9), suggesting it is not a major predictive factor for gender gaps. For Life Expectancy, only the Mid component was selected (Gap component = 0), meaning the gender gap in road traffic deaths does not contribute to explaining the Life Expectancy gap. See validation.md section “Road Traffic: WHO → IHME” for detailed analysis.

Maternal Mortality Ratio Indicators (Identified)

Multiple maternal mortality indicators have been identified from WHO GHO API:

  1. MDG_0000000026 - Maternal mortality ratio (per 100,000 live births)

    • Records: 7,878 (full dataset), 4,848 (2000-2023)

    • Years: 1985-2023 (excellent temporal coverage)

    • Countries: 202

    • Sex categories: N/A (inherently female-specific)

    • Status: ⚠️ Downloaded but removed from final model - removed due to counterintuitive positive coefficient

    • Note: Maternal mortality is inherently female-specific (deaths during pregnancy, childbirth, or within 42 days of termination of pregnancy). This indicator was tested in the model but removed because it had a counterintuitive positive coefficient, which implies that higher maternal mortality is associated with a larger LE/HALE gap. This is counterintuitive because if something increases female mortality, it should close the gap (since gap = Female - Male). The positive coefficient suggests a spurious association, possibly because maternal mortality is capturing something about general healthcare quality rather than a direct causal relationship. Removing it had minimal impact on model performance but improved interpretability. See validation.md section “Removing Maternal Mortality Indicator” for detailed analysis.

  2. MDG_0000000032 - Maternal mortality ratio (per 100,000 live births) - Country reported estimates

    • Years: 1987, 2000, 2002-2009 (limited temporal coverage)

    • Countries: 169

    • Status: ⚠️ Limited temporal coverage - only 10 years of data, older years, fewer countries than MDG_0000000026

  3. MORT_MATERNALNUM - Number of maternal deaths

    • Status: ⚠️ Less useful - absolute numbers rather than rates (rates are more comparable across countries)

Recommendation: Maternal mortality was tested in the model but removed from the final model due to a counterintuitive positive coefficient. The indicator had moderate importance (1.89 in Life Expectancy model, ranked #5; 2.15 in HALE model, ranked #7), but the positive coefficient suggests a spurious association rather than a direct causal relationship. Removing it had minimal impact on model performance (R² decreased slightly for LE, improved slightly for HALE) but improved model interpretability. After removal, Cardiovascular and Homicide gained substantial importance in the HALE model, suggesting maternal mortality may have been suppressing these indicators. See validation.md section “Removing Maternal Mortality Indicator” for detailed analysis.

Homicide Rate Indicators (Identified)

Two homicide-related indicators have been identified from WHO GHO API:

  1. VIOLENCE_HOMICIDERATE - Estimates of rates of homicides per 100,000 population

    • Records: 12,936

    • Years: 2000-2021 (excellent temporal coverage)

    • Countries: 196

    • Sex categories: Both sexes, Female, Male

    • Status: ✅ Implemented - excellent temporal coverage, excellent country coverage, gender breakdowns, includes confidence intervals

    • Note: This is a crude rate (not explicitly age-standardized), but has excellent temporal and country coverage. Homicide rates are typically much higher in men than women across most countries, making it a major contributor to the gender gap in mortality. Homicide reflects violence, conflict, and social factors that differentially affect men and women.

  2. VIOLENCE_HOMICIDENUM - Estimates of number of homicides

    • Years: 2000-2019 (slightly less recent than rate indicator)

    • Countries: 194

    • Sex categories: Both sexes, Female, Male

    • Status: ⚠️ Less useful - absolute numbers rather than rates (rates are more comparable across countries), and has less recent data (up to 2019 vs 2021)

Recommendation: The model uses IHME Interpersonal Violence (B.7.4) instead of WHO VIOLENCE_HOMICIDERATE because IHME provides better temporal coverage (1990-2023 vs 2000-2021) and consistent methodology with other IHME indicators. However, homicide was not selected by Elastic Net for the Life Expectancy model (importance = 0) when using IHME data, meaning it does not contribute to explaining the Life Expectancy gap. For HALE, homicide has moderate importance (ranked #5). This suggests that IHME homicide data may be less predictive than WHO data, or that other indicators (particularly Suicide) capture similar variance. See validation.md section “Homicide: WHO → IHME” for detailed analysis.

Diabetes Death Rate Indicators (Identified)

Multiple diabetes-related indicators have been identified from WHO GHO API:

  1. SA_0000001440 - Age-standardized death rates, diabetes mellitus, per 100,000

    • Records: 573

    • Years: 2004 (only 1 year)

    • Countries: 191

    • Sex categories: Both sexes, Female, Male

    • Status: ✅ Implemented - age-standardized, good country coverage, gender breakdowns

    • Note: Age-standardized rates match HALE methodology (HALE is also age-standardized). Diabetes is a chronic condition that can contribute to the gender gap in mortality, though the relationship may vary by country and healthcare access. The limitation is that it only has data for 2004, similar to cardiovascular disease indicators, which limits temporal analysis but provides a good cross-sectional snapshot for the analysis.

  2. SA_0000001421 - Age-standardized DALYs, diabetes mellitus, per 100,000

    • Records: 573

    • Years: 2004 (only 1 year)

    • Countries: 191

    • Sex categories: Both sexes, Female, Male

    • Status: ⚠️ Limited temporal coverage - only 2004 data, DALYs (Disability-Adjusted Life Years) provide a measure of both mortality and morbidity, but death rates are more directly comparable to HALE

    • Note: DALYs capture both mortality and morbidity, but for HALE gender gap analysis, death rates are more directly relevant since HALE focuses on healthy life expectancy.

  3. NCDMORT3070 - Probability (%) of dying between age 30 and exact age 70 from any of cardiovascular disease, cancer, diabetes, or chronic respiratory disease

    • Records: 12,936

    • Years: 2000-2021 (excellent temporal coverage - 22 years)

    • Countries: 196

    • Sex categories: Both sexes, Female, Male

    • Status: ✅ Implemented - excellent temporal coverage, excellent country coverage, gender breakdowns

    • Note: Combines multiple causes of death (cardiovascular disease, cancer, diabetes, chronic respiratory disease), so it’s less specific than individual cause indicators. However, it has much better temporal coverage (2000-2021) than diabetes-specific indicators (which only have 2004 data). This makes it useful for model comparison - trading off specificity for temporal coverage. The combined indicator may capture overall NCD mortality patterns that contribute to the HALE gender gap.

  4. NCD_DIABETES_PREVALENCE_AGESTD - Prevalence of diabetes, age-standardized

    • Status: ⚠️ Not suitable - prevalence indicator (not a death rate), measures disease burden but not mortality

  5. NCD_DIABETES_TREATMENT_AGESTD - Diabetes treatment coverage, age-standardized

    • Status: ⚠️ Not suitable - treatment coverage indicator, not a mortality measure

  6. Other indicators - Multiple policy/registry indicators (NCD_CCS_DiabetesReg, NCD_CCS_DiabetesTest, etc.)

    • Status: ⚠️ Not suitable - policy/regulatory indicators, not mortality data

Recommendation:

Model Strategy: Consider testing both indicators in the regression model to compare:

Intimate Partner Violence (IPV) Indicators (Identified)

Multiple intimate partner violence indicators have been identified from WHO GHO API. Note: IPV is a prevalence indicator (percentage of women experiencing violence), not a direct death rate. It affects women’s health and mortality indirectly through mental health, injuries, and other health consequences.

  1. SDGIPV - Proportion of ever-partnered women and girls aged 15-49 years subjected to physical and/or sexual violence by a current or former intimate partner in the previous 12 months

    • Records: 577

    • Years: 2000-2017

    • Countries: 126

    • Sex categories: Female (inherently female-specific)

    • Status: ✅ Implemented - good temporal coverage, good country coverage, matches SDG indicator 5.2.1

    • Note: This is a prevalence indicator (percentage), not a death rate. IPV affects women’s health indirectly through mental health impacts, injuries, and other health consequences. It may contribute to the gender gap in HALE through its effects on women’s physical and mental health, though the relationship is complex and indirect.

  2. SDGIPV12M - Proportion of ever-partnered women and girls aged 15–49 years subjected to physical and/or sexual violence by a current or former intimate partner in the previous 12 months

    • Years: 2018 (only 1 year)

    • Countries: 163

    • Status: ⚠️ Limited temporal coverage - only 2018 data, but good country coverage

  3. SDGIPVLT - Proportion of ever-partnered women and girls aged 15–49 years subjected to physical and/or sexual violence by a current or former intimate partner in their lifetime

    • Years: 2018 (only 1 year)

    • Countries: 158

    • Status: ⚠️ Limited temporal coverage - only 2018 data, lifetime prevalence (broader than 12-month)

  4. RHR_IPV - Intimate partner violence prevalence among ever partnered women (%)

    • Years: 2010 (only 1 year)

    • Countries: 29

    • Status: ⚠️ Very limited coverage - only 2010, only 29 countries

  5. SA_0000001455 - Age-standardized death rates, violence, per 100,000

    • Years: 2002, 2004 (only 2 years)

    • Countries: 192

    • Sex categories: Both sexes, Female, Male

    • Status: ⚠️ Limited temporal coverage - only 2 years, but age-standardized and has gender breakdowns. This captures all violence-related deaths (not just IPV), which may include homicide and other forms of violence.

Recommendation: Use SDGIPV as the primary intimate partner violence indicator - it has the best temporal coverage (2000-2017) and good country coverage (126 countries). ✅ Implemented - Data download functionality added to who_data.py. However, note that IPV is a prevalence indicator affecting women’s health indirectly, not a direct cause of death. It may be less directly relevant to HALE gender gap analysis than direct mortality indicators, but could be useful for understanding broader health impacts on women. Consider whether the indirect relationship to mortality makes it suitable for the regression analysis, or if it should be analyzed separately.

Infant and Child Mortality Indicators (Explored)

Multiple indicators related to infant, neonatal, and under-five mortality have been identified from WHO GHO API. Note: These indicators measure mortality in early life (birth to age 5), which may be less directly relevant to HALE gender gap analysis since HALE focuses on adult health outcomes. However, early-life mortality patterns can reflect underlying health disparities and may be relevant for understanding population-level gender differences.

Infant Mortality Indicators (with gender breakdowns):

  1. imr - Infant mortality rate (deaths per 1000 live births)

    • Years: 1932-2023 (excellent temporal coverage)

    • Countries: 249

    • Sex categories: Both sexes, Female, Male

    • Total records: 43,513

    • Status: ✅ Excellent coverage - has gender breakdowns, very long temporal coverage, comprehensive country coverage

  2. MDG_0000000001 - Infant mortality rate (probability of dying between birth and age 1 per 1000 live births)

    • Years: 1932-2023 (excellent temporal coverage)

    • Countries: 249

    • Sex categories: Both sexes, Female, Male

    • Total records: 43,513

    • Status: ✅ Excellent coverage - similar to imr, has gender breakdowns, very long temporal coverage

  3. CM_02 - Number of infant deaths

    • Years: 1951-2023

    • Countries: 249

    • Sex categories: Both sexes, Female, Male

    • Total records: 42,716

    • Status: ⚠️ Less useful - absolute numbers rather than rates (rates are more comparable across countries), and has less recent historical data (starts 1951 vs 1932)

Under-Five Mortality Indicators (with gender breakdowns):

  1. u5mr - Under-five mortality rate (deaths per 1000 live births)

    • Years: 1932-2023 (excellent temporal coverage)

    • Countries: 249

    • Sex categories: Both sexes, Female, Male

    • Total records: 63,070

    • Status: ✅ Excellent coverage - has gender breakdowns, very long temporal coverage, comprehensive country coverage

  2. MDG_0000000007 - Under-five mortality rate (probability of dying by age 5 per 1000 live births)

    • Years: 1932-2023 (excellent temporal coverage)

    • Countries: 249

    • Sex categories: Both sexes, Female, Male

    • Total records: 63,070 (30,648 with sex dimension when filtered)

    • Status: ⚠️ Downloaded but removed from final model - Data download functionality added to who_data.py. Excellent coverage with clean gender breakdowns (5,976 Male, 5,976 Female records). Much better data quality than u5mr when filtered for sex dimension. However, removed from final model due to very low importance (0.0558 in Life Expectancy model, not in top 10 for HALE model) and minimal impact on model performance. See validation.md section “Removing Childhood Indicator (Under-Five Mortality)” for detailed analysis.

Recommendation:

For HALE gender gap analysis: These indicators SHOULD be considered for inclusion in the regression model because:

  1. HALE is calculated from birth - HALE (Healthy Life Expectancy) measures expected years of healthy life at birth, so it includes all mortality from birth to death. If infant/child mortality differs by gender, it directly affects the HALE calculation and contributes to the gender gap.

  2. Goal is to explain the gap - The purpose of the model is to estimate what portion of the HALE gender gap is explainable by each factor. If infant/child mortality contributes to the gap, it should be included to properly attribute its contribution.

  3. Gender differences exist - Infant mortality is typically higher in males, and this gender difference will affect HALE calculations. Under-five mortality also shows gender differences that should be accounted for.

However, note that:

Recommendation for HALE model:

Occupational Attributable Death Indicators (Identified)

Multiple occupational-related death indicators have been identified from WHO GHO API. Note: Occupational hazards are a major contributor to the gender gap in mortality, as men are more likely to work in dangerous occupations (construction, mining, manufacturing) with higher rates of workplace accidents, injuries, and exposure to hazardous materials.

Occupational Risk Factors Indicators:

  1. OCC_1 - Occupational risk factors attributable deaths

    • Years: 2004 (only 1 year)

    • Country/Regions: 8 (not actual countries)

    • Sex categories: None (no gender breakdown)

    • Status: ⚠️ Very limited - only 2004 data, only 8 country/regions, no gender breakdown

  2. OCC_3 - Occupational risk factors attributable deaths per 100’000 capita

    • Years: 2004 (only 1 year)

    • Country/Regions: 8 (not actual countries)

    • Sex categories: None (no gender breakdown)

    • Status: ⚠️ Very limited - only 2004 data, only 8 country/regions, no gender breakdown

  3. OCC_2 - Occupational risk factors attributable DALYs ('000)

    • Status: ⚠️ Less useful - DALYs rather than deaths, absolute numbers rather than rates

  4. OCC_4 - Occupational risk factors attributable DALYs per 100’000 capita

    • Status: ⚠️ Less useful - DALYs rather than deaths

Occupational Injuries Indicators:

  1. OCC_19 - Occupational injuries attributable deaths

    • Years: 2004 (only 1 year)

    • Country/Regions: 8 (not actual countries)

    • Sex categories: None (no gender breakdown)

    • Status: ⚠️ Very limited - only 2004 data, only 8 country/regions, no gender breakdown

  2. OCC_21 - Occupational injuries attributable deaths per 100’000 capita

    • Years: 2004 (only 1 year)

    • Country/Regions: 8 (not actual countries)

    • Sex categories: None (no gender breakdown)

    • Status: ⚠️ Very limited - only 2004 data, only 8 country/regions, no gender breakdown

  3. OCC_20 - Occupational injuries attributable DALYs ('000)

    • Status: ⚠️ Less useful - DALYs rather than deaths, absolute numbers rather than rates

  4. OCC_22 - Occupational injuries attributable DALYs per 100’000 capita

    • Status: ⚠️ Less useful - DALYs rather than deaths

Occupational Airborne Particulates Indicators:

  1. OCC_5 - Occupational airborne particulates attributable deaths

    • Status: ⚠️ Very limited coverage expected (similar to other OCC indicators)

  2. OCC_7 - Occupational airborne particulates attributable deaths per 100’000 capita

    • Status: ⚠️ Very limited coverage expected (similar to other OCC indicators)

  3. OCC_6 - Occupational airborne particulates attributable DALYs ('000)

    • Status: ⚠️ Less useful - DALYs rather than deaths, absolute numbers rather than rates

  4. OCC_8 - Occupational airborne particulates attributable DALYs per 100’000 capita

    • Status: ⚠️ Less useful - DALYs rather than deaths

Occupational Carcinogens Indicators:

  1. OCC_9 - Occupational carcinogens attributable deaths

    • Status: ⚠️ Very limited coverage expected (similar to other OCC indicators)

  2. OCC_11 - Occupational carcinogens attributable deaths per 100’000 capita

    • Status: ⚠️ Very limited coverage expected (similar to other OCC indicators)

  3. OCC_10 - Occupational carcinogens attributable DALYs ('000)

    • Status: ⚠️ Less useful - DALYs rather than deaths, absolute numbers rather than rates

  4. OCC_12 - Occupational carcinogens attributable DALYs per 100’000 capita

    • Status: ⚠️ Less useful - DALYs rather than deaths

Occupational Ergonomic Stressors Indicators:

  1. OCC_15 - Occupational ergonomic stressors attributable deaths

    • Status: ⚠️ Very limited coverage expected (similar to other OCC indicators)

  2. OCC_17 - Occupational ergonomic stressors attributable deaths per 100’000 capita

    • Status: ⚠️ Very limited coverage expected (similar to other OCC indicators)

  3. OCC_16 - Occupational ergonomic stressors attributable DALYs ('000)

    • Status: ⚠️ Less useful - DALYs rather than deaths, absolute numbers rather than rates

  4. OCC_18 - Occupational ergonomic stressors attributable DALYs per 100’000 capita

    • Status: ⚠️ Less useful - DALYs rather than deaths

Other Occupational Indicators:

Recommendation:

For HALE gender gap analysis: The occupational attributable death indicators identified are NOT suitable for the regression model because:

  1. No gender breakdowns - None of the occupational death indicators have sex categories (Male, Female, Both sexes), which is essential for analyzing gender gaps. This is a critical limitation since occupational hazards are known to differentially affect men and women.

  2. Very limited temporal coverage - All indicators checked (OCC_1, OCC_3, OCC_19, OCC_21) only have data for 2004 (1 year), which severely limits temporal analysis and cross-country comparison.

  3. Very limited country/region coverage - Only 8 country/regions have data (not actual countries), which is insufficient for a comprehensive cross-country analysis of HALE gender gaps.

  4. Missing key information - The indicators don’t provide the gender-specific data needed to calculate male vs. female differences or ratios, which are required for the regression model.

Alternative approaches:

Status: ⚠️ Not suitable for model - No gender breakdowns, very limited temporal and country coverage. Occupational hazards remain an important theoretical factor but cannot be directly measured with available WHO GHO data.

Drilldown Data

Data for investigating specific drivers of gender gaps within broader categories (separate from the primary cross-country regression models).

Cancer (Neoplasms) Drilldown (Downloaded)

Indicator: Drilldown into Neoplasms (Level 3 or Level 4 causes) Measure: Death rates per 100,000 Locations: United States, Iceland, and OECD (Total) Age: All ages Sex: Separate files for Male and Female

GBD Compare Tool Settings:

Files:

Status: ✅ Downloaded for specific drilldown analysis Relevance: Neoplasms are the biggest driver of gender gaps in LE and HALE. This data allows for identifying which specific types of cancer (e.g., lung cancer, liver cancer, colorectal cancer, etc.) contribute most to the gap, providing more granular insights than the top-level “Neoplasms” category. Comparing US data with the OECD total helps contextualize the findings.

Target Variables

HALE (Healthy Life Expectancy) - Primary Target Variable

Indicator: WHOSIS_000002 - Healthy life expectancy (HALE) at birth (years)
Source: WHO GHO API
File: data/who_hale_data.csv
Status: ✅ Implemented

Data Details:

Relevance: HALE measures the average number of years that a person can expect to live in “full health” by taking into account years lived in less than full health due to disease and/or injury. This is the primary target variable for the analysis. The gender gap (Female HALE - Male HALE) measures the difference in healthy life expectancy between women and men.

IHME HALE (Healthy Life Expectancy) - Alternative Target Variable Source

Indicator: HALE (Healthy life expectancy)
Source: IHME Global Burden of Disease (GBD) 2023
File: data/IHME-GBD_2023_DATA-fc42b373-1.csv
Status: ⚠️ Downloaded, exploration pending

Data Details:

Data Structure:

Advantages over WHO HALE data:

  1. More years: 1990-2023 (34 years) vs WHO 2000-2021 (22 years)

  2. More recent data: Includes 2022-2023 (WHO stops at 2021)

  3. More countries: 204 vs 196

  4. Consistent methodology: Same uncertainty bounds structure as other IHME indicators used in the model

  5. Earlier start date: 1990 vs 2000 allows for longer temporal analysis

Considerations:

Relevance: IHME HALE data could potentially replace WHO HALE data in the model, providing better temporal coverage (especially for recent years 2022-2023) and more countries. This would be particularly valuable for:

Citation: Global Burden of Disease Collaborative Network. Global Burden of Disease Study 2023 (GBD 2023) Results. Seattle, United States: Institute for Health Metrics and Evaluation (IHME), 2024. Available from https://vizhub.healthdata.org/gbd-results/.

Life Expectancy - Secondary Target Variable

Indicator: WHOSIS_000001 - Life expectancy at birth (years)
Source: WHO GHO API
File: data/who_life_expectancy_data.csv
Status: ✅ Implemented

Data Details:

Relevance: Life expectancy at birth measures the average number of years a person can expect to live, regardless of health status. This is the secondary target variable for the analysis, allowing comparison of which factors explain the gender gap in overall life expectancy versus healthy life expectancy. The gender gap (Female LE - Male LE) measures the difference in life expectancy between women and men.

Key Differences from HALE:

Note on Temporal Coverage Limitation: HALE and Life Expectancy data from WHO are currently only available through 2021. This limits the temporal coverage of analyses that include COVID-19 data, even though COVID-19 death rate data from IHME extends to 2023. When including COVID-19 as a predictor, the analysis cutoff year is set to 2021 (rather than 2023) to match the availability of the target variables. This ensures that all countries have complete data for both predictors and targets in the same years.

Life Expectancy (OWID) - Extended Temporal Coverage

Indicator: Life expectancy at birth by sex
Source: Our World in Data (OWID) - Combines Human Mortality Database (2025) and UN World Population Prospects (2024)
Files:

Data Details:

Key Advantages Over WHO Data:

  1. Extended Temporal Coverage: 2 additional years (2022-2023) beyond WHO’s 2021 cutoff

    • Allows analysis through 2023 when paired with IHME HALE (2000-2023)

    • Captures post-acute COVID recovery period (2022-2023)

  2. Complete OECD Coverage: All 38 OECD countries including Turkey (TUR) with complete data for 2020-2023

  3. High Data Quality: Combines authoritative sources (HMD pre-1950, UN WPP 1950+)

  4. 100% Completeness: No missing values in recent years

  5. Historical Context: Full historical series from 1751 for selected countries (HMD coverage)

Gender Gap Statistics (2020-2023, OECD):

Source Details:

Use Cases:

  1. Primary Use: Extended temporal analysis when paired with IHME HALE data (2000-2023)

  2. Secondary Use: Cross-validation with WHO LE data where they overlap (2000-2021)

  3. Historical Analysis: Long-term trends in gender gaps (if historical analysis desired)

Recommendation: Use OWID LE data for future analyses when extending beyond 2021 to match IHME HALE temporal coverage (2000-2023). For consistency with WHO HALE-based analyses, continue using WHO LE data. The two sources show high agreement where they overlap.

Promising Indicators Checklist

Based on the WHO GHO Indicators Index, the following indicators are most relevant for analyzing HALE gender gaps. They are likely to differ between men and women and are related to causes of death.

Already Implemented ✅

High Priority - To Investigate

Medium Priority - To Investigate

Lower Priority - May Be Useful

Notes:

WHO and IHME Indicator Correspondence

This table shows how WHO and IHME indicators correspond to each other, helping identify alternatives and complementary data sources.

Indicator CategoryWHO IndicatorWHO CodeWHO Temporal CoverageIHME IndicatorIHME CodeIHME Temporal CoverageRelationship
Target Variables
HALEHealthy Life ExpectancyWHOSIS_0000022000-2021———WHO only (primary source)
Life ExpectancyLife Expectancy at BirthWHOSIS_0000012000-2021———WHO only (primary source)
Alcohol-Related
Alcohol-attributable deathsAlcohol-attributable all-cause deathsSA_00000018322019 onlyAlcohol use disordersB.7.11990-2023IHME used in model (better temporal coverage, but narrower definition - see alcohol_data_comparison.md)
Suicide/Self-Harm
Suicide ratesAge-standardized suicide ratesMH_122000-2021Self-harmB.7.31990-2023IHME used in model (better temporal coverage, importance increased)
Violence/Homicide
Homicide ratesEstimates of homicide ratesVIOLENCE_HOMICIDERATE2000-2021Interpersonal violenceB.7.41990-2023IHME used in model (better temporal coverage, but dropped out of LE model)
Road Traffic
Road traffic crashesRoad traffic crash deaths (15+)SA_00000014592019 onlyRoad injuriesRoad injuries1990-2023IHME used in model (much better temporal coverage, but very low importance)
Maternal Mortality
Maternal mortality ratioMaternal mortality ratioMDG_00000000261985-2023Maternal disordersMaternal disorders1990-2023Both removed from model - counterintuitive positive coefficient (see validation.md)
Child Mortality
Under-five mortality rateUnder-five mortality rateMDG_00000000071932-2023All-cause deaths under 5All causes (<5 years)1990-2023Both removed from model - WHO had very low importance; IHME confounded with age structure/fertility (see validation.md)
Diabetes
Diabetes death ratesAge-standardized diabetes death ratesSA_00000014402004 onlyDiabetes type 2B.8.1.21990-2023IHME alternative (much better temporal coverage)
Cardiovascular Disease
Cardiovascular diseaseAge-standardized cardiovascular death ratesVarious (WHS2_161, etc.)2004 onlyCardiovascular diseasesB.21990-2023IHME alternative (much better temporal coverage)
Chronic Respiratory Disease
————Chronic respiratory diseasesB.31990-2023IHME only (no WHO equivalent with good coverage)
Liver Disease
————Cirrhosis and other chronic liver diseasesCirrhosis and other chronic liver diseases1990-2023IHME only (no WHO equivalent with good coverage)
COVID-19
————COVID-19COVID-192020-2023IHME only (no WHO equivalent with good coverage)
Cancer
————Neoplasms (cancer)B.11990-2023IHME only (no WHO equivalent with good coverage)
Injuries
Unintentional poisoningMortality rate from unintentional poisoningSDGPOISON2000-2021———WHO removed from model (not selected)
Unintentional injuries———Unintentional injuriesUnintentional injuries1990-2023IHME used in model (broader than WHO poisoning)
Drug Use
Unintentional poisoningMortality rate from unintentional poisoningSDGPOISON2000-2021Drug use disordersB.7.21990-2023IHME used in model (WHO poisoning removed, but IHME DrugDisorder has 0 importance - not selected)
Other
Smoking prevalenceAge-standardized tobacco smokingM_Est_smk_curr_std2000-2030———WHO only (prevalence indicator, not mortality)
NCD mortality (30-70)Probability of dying 30-70 from NCDsNCDMORT30702000-2021———WHO only (combined indicator)
Intimate partner violenceIPV prevalenceSDGIPV2000-2017———WHO only (prevalence indicator, not mortality)

Notes on Correspondence:

  1. IHME alternatives with better temporal coverage: Alcohol use disorders, Self-harm, Interpersonal violence, Road injuries, Diabetes type 2, and Cardiovascular diseases all have IHME versions with much better temporal coverage (1990-2023) compared to their WHO counterparts (often 2004 or 2019 only).

  2. Complementary indicators:

    • Maternal mortality: WHO uses ratio per 100,000 live births; IHME uses rate per 100,000 population

    • Under-five mortality: WHO uses rate per 1,000 live births; IHME all-cause under-5 uses rate per 100,000 population

  3. IHME-only indicators: Chronic respiratory diseases, Liver disease (cirrhosis and other chronic liver diseases), COVID-19, Neoplasms, Unintentional injuries, and Drug use disorders are available from IHME but have no good WHO equivalent with adequate temporal coverage.

  4. WHO-only indicators: Smoking prevalence, NCD mortality (30-70), Intimate partner violence, and Unintentional poisoning are available from WHO but have no IHME equivalent.

  5. Target variables: HALE and Life Expectancy are only available from WHO and serve as the primary target variables for the analysis.

Gender Gaps in Life Expectancy
Gender Gaps in Life Expectancy
Gender Gaps in Life Expectancy
Neoplasms Drilldown Analysis Report