Inequality in the Distribution of Air Pollution Attributable Mortality Within Canadian Cities

Abstract

Recent studies have identified inequality in the distribution of air pollution attributable health impacts, but to our knowledge this has not been examined in Canadian cities. We evaluated the extent and sources of inequality in air pollution attributable mortality at the census tract (CT) level in seven of Canada's largest cities. We first regressed fine particulate matter (PM_2.5) and nitrogen dioxide (NO₂) attributable mortality against the neighborhood (CT) level prevalence of age 65 and older, low income, low educational attainment, and identification as an Indigenous (First Nations, Métis, Inuit) or Black person, accounting for spatial autocorrelation. We next examined the distribution of baseline mortality rates, PM_2.5 and NO₂ concentrations, and attributable mortality by neighborhood (CT) level prevalence of these characteristics, calculating the concentration index, Atkinson index, and Gini coefficient. Finally, we conducted a counterfactual analysis of the impact of reducing baseline mortality rates and air pollution concentrations on inequality in air pollution attributable mortality. Regression results indicated that CTs with a higher prevalence of low income and Indigenous identity had significantly higher air pollution attributable mortality. Concentration index, Atkinson index, and Gini coefficient values revealed different degrees of inequality among the cities. Counterfactual analysis indicated that inequality in air pollution attributable mortality tended to be driven more by baseline mortality inequalities than exposure inequalities. Reducing inequality in air pollution attributable mortality requires reducing disparities in both baseline mortality and air pollution exposure.

Keywords: air pollution; environmental injustice; environmental racism.

Figure 1

Locations of included cities.

Figure 2

Distribution of demographic characteristics, baseline mortality rates, air pollution concentrations and air pollution attributable mortality rates by census tract in Calgary; quintiles as cutpoints. LICO = Low Income Cut‐Off. Created with R tmap package.

Figure 3

Percent increase in PM_2.5 and NO₂ attributable mortality per interquartile range increase in census tract prevalence of age ≥ 65, income < Low Income Cut‐Off and identifying as an Indigenous person, by city and pooled across cities. Based on conditional autoregressive multivariate models including the three variables and accounting for spatial autocorrelation. Figure created with R metafor package using results generated by the CARBayes package.

Figure 4

Heatmap of inequality index values for baseline mortality rate/100,000 population, PM_2.5 concentration, NO₂ concentration, PM_2.5 attributable mortality/100,000 population and NO₂ attributable mortality/100,000 population by city and population characteristic. GTE = greater than or equal to, LTLICO = less than Low Income Cut‐Off, LTHS = less than High School. Only values significantly different from zero are shown. Figure created with R pheatmap package using inequality index values generated by the rineq package.

Figure 5

Heatmap of inequality index values for PM_2.5 and NO₂ attributable mortality/100,000, comparing base case, pollutant counterfactual and mortality counterfactual. All analyses including the base case employed income quintile specific hazard ratios for the association between air pollution and mortality. LTLICO = less than Low Income Cut‐Off. Only values significantly different from zero are shown. Figure created with R pheatmap package using inequality index values generated by the rineq package.

Figure 6

Observed and counterfactual spatial distribution of air pollution attributable mortality per 100,000 population in Calgary. Only census tracts above the median prevalence of income less than the low income cut‐off (LICO) and identifying as an Indigenous person are shaded. Counterfactual scenarios set air pollution concentrations or baseline mortality rates to the mean observed in census tracts below the median prevalence of income less than the low income cut‐off (LICO) and Indigenous identity. Created with R tmap package.