Commuting-adjusted short-term health impact assessment of airborne fine particles with uncertainty quantification via Monte Carlo simulation

Abstract

Background: Exposure to air pollution is associated with a short-term increase in mortality, and this field has begun to focus on health impact assessment.

Objectives: Our aim was to estimate the impact of PM10 on mortality within 2 days from the exposure in the Italian region of Lombardy for the year 2007, at the municipality level, examining exposure entailed by daily intermunicipality commuting and accounting for uncertainty propagation.

Methods: We combined data from different sources to derive probabilistic distributions for all input quantities used to calculate attributable deaths (mortality rates, PM10 concentrations, estimated PM10 effects, and commuting flows) and applied a Monte Carlo procedure to propagate uncertainty and sample the distribution of attributable deaths for each municipality.

Results: We estimated that annual average PM10 concentrations above the World Health Organization-recommended threshold of 20 μg/m3 were responsible for 865 short-term deaths (80% credibility interval: 475, 1,401), 26% of which were attributable to PM10 above the European Union limit of 40 μg/m3. Reducing annual average PM10 concentrations > 20 μg/m3 by 20% would have reduced the number of attributable deaths by 36%. The largest estimated impacts were along the basin of the Po River and in the largest cities. Commuting contributed to the spatial distribution of the estimated impact.

Conclusions: Our estimates, which incorporated uncertainty quantification, indicate that the short-term impact of PM10 on mortality in Lombardy in 2007 was notable, and that reduction in air pollution would have had a substantial beneficial effect on population health. Using commuting data helped to identify critical areas for prioritizing intervention.

Figure 1

Subdivision of Lombardy by province in 2007. The provincial capitals and the cities of Busto Arsizio and Vigevano are indicated in white. Grayscale expresses altitude, with darker tones indicating mountain areas. Reproduced from Baccini et al. (2011), with permission of Oxford University Press.

Figure 2

Posterior means of annual average PM₁₀ concentrations (μg/m³) (A) and smoothed crude mortality rates (per 100,000 residents) (B); posterior medians of attributable deaths among residents (C) and posterior probabilities (%) of a non-null impact (D) under the RS0 scenario (for annual average PM₁₀ > 20 μg/m³), by municipality.

Figure 3

Log ratio between the posterior median of “exported” attributable deaths (C_i, attributable deaths among nonresidents due to exposure in city i) and the posterior median of the “imported” attributable deaths (B_i, attributable deaths among residents of city i due to exposure outside of city i), by municipality.