Background

The adverse effects of air pollution on premature mortality are well known, yet methodological challenges in assessing the health benefits of environmental interventions remain. This study applies a g-computation approach to estimate the impact of hypothetical air quality interventions on premature mortality in Belgium, using nationally representative survey data.

Methods

Data from the Belgian Health Interview Surveys (2008, 2013, 2018; n = 27,536) were linked to environmental exposure data based on residential addresses and mortality follow-up through 2023. Premature mortality was defined as death occurring before age 70. Cox proportional hazards models were used to estimate the association between long-term air pollution exposure and premature mortality, adjusting for socio-economic, environmental, and lifestyle factors. A g-computation approach simulated scenarios where air pollution exposure was reduced to WHO air quality guideline levels. Potential impact fractions quantified the proportion of premature deaths that could be prevented.

Results

During follow-up, 3,995 deaths occurred and 771 participants were lost due to registry deletion. Mean follow-up was 8 years. Overall, 2.4% of the study population experienced premature mortality. Premature mortality was significantly associated with NO2 and BC exposure. A quadratic term for BC improved model fit, revealing a flattening effect at higher exposures. Hazard ratios (HRs) per interquartile range (IQR) were: PM2.5: 1.13 (0.83;1.53), PM10: 1.13 (0.93;1.38), NO2: 1.30 (1.01;1.69), and BC: 1.45 (1.29;1.75). No interactions were found with age or sex. Meeting WHO guideline levels for NO2 would reduce premature mortality risk by 0.55 percentage points (95% CI: 0.06;1.04), preventing 24% of cases. A clear dose-response was observed between NO2 and BC reductions and premature mortality.

Conclusions

Reducing traffic-related air pollution could substantially lower premature mortality in Belgium.

Key messages

• Air quality interventions could effectively decrease the prevalence of premature mortality in Belgium.

• The g-computation approach to assess PIF of interventions represents a straightforward approach for drawing causal inferences from observational data.