Local approach to attributable disease burden: a case study on air pollution and mortality in Belgium

Burden of disease estimation and the attribution to risk factors are commonly done on national or regional scale. This research proposes a novel approach, where air pollution-related mortality in Belgium was estimated locally, and compares the results to those of the common 'global' approach.

In the local approach, mortality attributable to long-term exposure to particulate matter < 2.5 μm (PM_2.5) and nitrogen dioxide (NO₂) is derived at the level of census tracts. Relying on a statistical concentration-response function suggests potential bias when applied to such small scale. Therefore, the local method is validated by comparing aggregated results to estimates derived with a global approach. In a sensitivity analysis, the difference between the global and local approach is compared to the impact of other methodological choices and sources of uncertainty.

The local method estimates (95% confidence interval) 12,276 (6,695; 17,826) deaths for PM_2.5 and 7,944 (4,725; 11,181) for NO₂ in Belgium. For both pollutants, these national estimates never deviate more than 2% from those obtained with the global method, and never more than 4% in the individual provinces. The sensitivity analysis demonstrates the concentration-response function as having the largest contribution to overall uncertainty, while the global-local discrepancy is slightly larger compared to the exposure uncertainty.

Aggregated local burden estimates prove to be accurate compared to the global approach. This means the local method shows potential for comparing areas and population groups at subnational level, where estimates can be generated in a flexible manner depending on research or policy needs.