What We Set Out to Answer

Our Core Research Question

"To what extent have ambient PM₂.₅ concentrations (2010–2019) influenced the burden of cardiovascular and respiratory diseases across countries with varying socio-demographic development levels, and how might this long-term exposure relate to COVID-19 mortality outcomes?"
"It was driven by a story from a group member: Her mother, who has asthma, experiences severe discomfort when staying in cities with poor air quality. Even short visits can mean weeks of relying on medications, constant coughing, and the simple wish to breathe without pain. For her and millions with similar and different health conditions, clean air isn't just a preference, it's a necessity."
But how feasible is it to find cities where the air is truly safe for vulnerable individuals?

Study Scope Note

This study focuses on ambient (outdoor) PM₂.₅ exposure. Indoor sources of pollution, such as cooking or heating fuels, are not included. While this means our estimates don't cover the full spectrum of exposure, outdoor air pollution remains the dominant and most consistently measured source across countries, making it a reliable basis for global comparison.

What Is PM₂.₅?

PM₂.₅ consists of ultrafine particles emitted from various sources. Because of its small size, PM₂.₅ can penetrate deep into the lungs, cross into the bloodstream, and contribute to serious health conditions.

Sources of PM₂.₅

  • • Vehicle exhaust
  • • Industrial activity
  • • Biomass burning
  • • Dust and construction
  • • Power plant emissions
  • • Agricultural activities

Health Effects

  • • Asthma and respiratory issues
  • • COPD (Chronic Obstructive Pulmonary Disease)
  • • Stroke and cardiovascular disease
  • • Heart disease
  • • Lung cancer
  • • Premature death

Statistical Methods & Techniques

We applied several statistical and machine learning techniques to uncover patterns in the data, each serving a specific purpose in our analysis.

Core Statistical Techniques

▶ Linear Regression
Our baseline for estimating average effects of PM₂.₅ exposure
▶ Random Forest Regression
To detect nonlinear patterns and complex relationships in the data
▶ Quantile Regression
To see if impacts differ between high-burden vs low-burden countries
▶ Interaction Models
To test if development level (SDI) changes how PM₂.₅ affects health
▶ Log-Transformed Regression
For skewed data like COVID-19 deaths to stabilize variance

Temporal Analysis

Rolling Averages

Multi-year averages to smooth out year-to-year fluctuations and capture longer-term trends.

Lagged Averages

Averages from previous years to estimate delayed health effects of pollution exposure.

Why This Matters

Health effects from air pollution often develop over years, not immediately. Our temporal analysis captures these delayed impacts.

Research methodology process

Research methodology and statistical analysis process

Key Findings

Our analysis reveals significant disparities in how air pollution affects different populations worldwide

PM₂.₅ Hits Harder in Low-SDI Countries

0.72 DALYs
per 100,000 people in low-SDI countries for 10 µg/m³ increase
0.22 DALYs
per 100,000 people in very high-SDI countries for same increase

The same air pollution is more dangerous if you live in a less-developed country.

SDI Is the Strongest Predictor

Socio-demographic development (SDI) consistently explains the largest share of variation in DALYs across countries.

Primary Factor
Development level determines health vulnerability more than pollution alone

Pollution's Effects Are Cumulative

Models that used 3 to 5 year rolling averages performed better than single-year exposure.

Long-term Impact
Suggests cumulative damage from pollution exposure over time

COVID-19: Weaker but Consistent Link

Long-term PM₂.₅ exposure showed a modest but consistent association with COVID-19 mortality in 2020.

Decade Average
Especially when averaged over the previous decade

Data Sources & Measurements

What We Measured

PM₂.₅ Exposure Data

Annual averages at country level from the WHO Global Health Observatory . This includes ambient (outdoor) particulate matter concentrations measured in μg/m³.

Health Burden Data (DALYs)

Disability-Adjusted Life Years from the IHME Global Burden of Disease study . Covers cardiovascular disease, stroke, chronic respiratory disease, and all-cause mortality.

SDI (Socio-Demographic Index)

A composite measure combining income per capita, average educational attainment, and total fertility rate, from the IHME SDI database . Represents a country's development level on a 0–1 scale.

Disease Categories: We grouped our results by disease type: Cardiovascular disease, Stroke, Chronic respiratory disease, All-cause DALYs, and COVID-19 mortality

Research Visualizations

Explore our complete collection of data visualizations and analysis figures

Vulnerability Analysis
Vulnerability Analysis by Region

How development level affects health vulnerability to air pollution

Top 10 Countries by PM2.5 Exposure
Top 10 Countries by PM₂.₅ Exposure

Countries with highest ambient air pollution levels

Respiratory DALYs Diagnostic Plots
Respiratory DALYs Model Diagnostics

Statistical validation of respiratory disease models

Cardiovascular DALYs Diagnostic Plots
Cardiovascular DALYs Model Diagnostics

Statistical validation of cardiovascular disease models

View All Visualizations on GitHub

Technical Glossary

Key terms and concepts used throughout our research and analysis

Term Definition
PM₂.₅ Fine particulate matter smaller than 2.5 micrometers in diameter. These airborne particles can penetrate deep into the lungs and enter the bloodstream, causing serious health effects.
DALYs Disability-Adjusted Life Years - a measure of overall disease burden, calculated as the sum of years of life lost due to premature death and years lived with disability.
SDI Socio-Demographic Index - a composite index combining income, education level, and fertility rate to represent a country's development level (0-1 scale).
Rolling Average A multi-year average of data used to smooth out year-to-year fluctuations and capture longer-term trends in pollution exposure.
Lagged Average An average of exposure values from previous years used to estimate delayed health effects of pollution (health impacts often occur years after exposure).
Linear Regression A basic statistical method that models the relationship between a continuous outcome and predictors by fitting a straight line. Used as baseline to estimate average effects.
Random Forest Regression A machine learning method using an ensemble of decision trees to model complex and nonlinear relationships between variables. Helps detect patterns missed by linear models.
Log-Transformed Regression Linear regression where the dependent variable is log-transformed to handle skewed data and stabilize variance, particularly useful for COVID-19 mortality data.
Interaction Models Regression models that test whether the effect of one variable (PM₂.₅) changes depending on another (SDI), revealing effect modification patterns.
Quantile Regression A statistical technique estimating relationships at different points of the outcome distribution (25th, 75th percentile), useful for understanding heterogeneous effects.

Ready to Explore the Data?

Now that you understand our methodology, try our interactive tool to see how these findings apply to specific countries.

Launch Data Tool