Exposure to PM2.5 and Mortality in the Isle of Ireland

Talk by Babak Jahanshahi (Queen's University Belfast) as part of the Research Seminar Series of the IOS Economics Department.

This paper assesses the relationship between PM2.5 exposure and mortality in Northern Ireland, a setting where pollution is low compared with most other countries. The study used data from a new linkage between the Northern Ireland Longitudinal Study (NILS), matched to pollution data at the 1km grid-square level and death events. The NILS is a longitudinal study that tracks a 28% representative sample of the Northern Ireland population drawn from the NI Health Card Registration System, which contains address histories updated biannually. The NILS is linked to several other administrative datasets including Census records for 2001 and 2011, which provided rich information on socioeconomic and demographic characteristics and contexts for sample members. The pollution data, matched at the residential property level to NILS participants, provided annual 1km grid-square modelled pollution data from 2002-2019 for some pollutants including PM2.5. These data were produced by Ricardo Energy & Environment for the UK Government’s air quality assessments.

We apply two analytical methods to this data to better understand the relationship of interest: Cox proportional hazards models, and propensity score matching. Cox proportional hazards models reveal associations between mortality risk and PM2.5 exposure, though the magnitude of the effects depend on the specification. Without adjustment for measured characteristics, an interquartile range increase in the five-year moving average of exposure is associated with a relatively large increase in the hazard of mortality. This effect decreases after adjusting for prior individual-level demographic, socioeconomic and health-related factors and household-level factors; however, the hazard ratio remains above one and comparable with estimates from other studies. In further analysis, propensity score matching is used to assess whether, among those with similar levels of pre-move exposure, those who go on to move to areas with different pollution levels face different subsequent mortality risks. This analysis suggests that mortality risk is higher (lower) for those who move to a more (less) polluted area.  This study offers evidence of mortality effects of PM2.5 from a setting where annual mean PM2.5 concentrations averaged roughly 8 micrograms per cubic metre over the study period. Therefore, we conclude that, consistent with similar studies in England, Canada, and European countries, Individuals with higher long-term exposure face additional mortality risk even in the context of exposure to low pollution.