Health Reports
Socioeconomic differences in nitrogen dioxide ambient air pollution exposure among children in the three largest Canadian citiesHealth Reports
Socioeconomic differences in nitrogen dioxide ambient air pollution exposure among children in the three largest Canadian cities

View the most recent version.

Archived Content

Information identified as archived is provided for reference, research or recordkeeping purposes. It is not subject to the Government of Canada Web Standards and has not been altered or updated since it was archived. Please "contact us" to request a format other than those available.

Archived

This page has been archived on the Web.

by Lauren Pinault, Daniel Crouse, Michael Jerrett, Michael Brauer and Michael Tjepkema

Environmental exposure to nitrogen dioxide (NO₂), which is largely a by-product of vehicular exhaust and incomplete combustion of fossil fuels, has been associated with emergency hospitalizations for respiratory conditions among adults in lower income quartiles in large Canadian cities.^{Note 1} Researchers have hypothesized that children may be even more vulnerable to the effects of air pollution because of their greater susceptibility to respiratory damage and faster respiratory rate, both of which exacerbate the effects of exposure.^{Note 2Note 3} A time-series study of school-aged children in Vancouver observed that boys had a greater risk of asthma hospitalization after increases in ambient NO₂ exposure if they were of low rather than high socioeconomic status (SES).^{Note 2}

NO₂ exposure among children has been associated with an increased risk of asthma, wheezing, ear, nose and throat infections, influenza, and serious colds^{Note 4Note 5Note 6Note 7}; reduced lung function^{Note 8}; and more severe respiratory tract infections for those with asthma.^{Note 9} A California study found that each increase of 5.7 ppb in NO₂ exposure was associated with almost a doubling of asthma prevalence.^{Note 5} According to a similar Vancouver study, exposure to NO₂ increases of 10 μg/m³ during the first year of life significantly increased the odds of developing asthma.^{Note 6} In addition to these health outcomes, air pollution exposure has been related to reduced school attendance and academic performance within the northeastern United States.^{Note 3}

Previous Canadian studies have reported significant within-city variations in residential exposure to NO₂ among adults,^{Note 10} and have correlated neighbourhood socioeconomic characteristics such as low income, lower educational attainment, and lower dwelling value with higher exposures.^{Note 10Note 11}

It is possible that NO₂ exposure varies within cities, and that children in lower income neighbourhoods have greater exposure than do those in higher income neighbourhoods.^{Note 2} The purpose of the present study was to determine if NO₂ exposure in the three largest Canadian cities—Toronto, Montreal, and Vancouver—was higher among children of lower SES and who were members of visible minorities.

Data and methods

Nitrogen dioxide exposure data were taken from published land use regression (LUR) models for Toronto, Montreal, and Vancouver, which had been created based on a combination of extensive field-measured samples of NO₂ along with datasets describing local land use characteristics, such as traffic counts, land use patterns, and population and road densities.^{Note 12Note 13Note 14} The Montreal LUR model pertains to mean annual concentrations in 2006. The Toronto and Vancouver NO₂ models were based on data from 2003 and 2010, respectively, and were adjusted with observations from Environment Canada’s National Air Pollution Surveillance network to be consistent with city-wide mean values for 2006.^{Note 15} This adjustment for year is supported by analyses in Canada and Europe demonstrating that NO₂ spatial models are consistent over time.^{Note 14Note 16Note 17}

Socioeconomic, visible minority, and demographic data for children were derived from the 2006 Canadian Census long-form questionnaire, representing a random sample of 20% of the three urban populations.^{Note 18} Children were included in the study if they lived within the boundaries of the LUR models for Toronto, Montreal or Vancouver,^{Note 12Note 13Note 14} were younger than age 18, and did not live in an institution. The LURs do not include all the suburban communities in the greater Toronto area, or areas beyond the island of Montreal. The variables derived for each child were before-tax household income, residence in a lone-parent family, visible minority status, age, and sex. Visible minorities are persons (other than Aboriginal), who are non-Caucasian (race) or non-white (colour).^{Note 18}

Respondent data were aggregated to the Dissemination Area (DA) level. DAs consist of one or more Census dissemination blocks and have populations of about 400 to 700.^{Note 18} A Geographic Information System (ArcGIS v.10, ESRI 2010) was used to map respondents to their approximate location of residence using Statistics Canada’s Postal Code Conversion File (PCCF+) version 5K.^{Note 19} The PCCF+ program uses a population-weighted random allocation algorithm to assign approximate postal code centroid locations, with an error rate of 1.4% in Census Metropolitan Areas.^{Note 19} These locations were used to spatially match respondents to a mean ambient NO₂ estimate at the DA level. Thus, the DA-level NO₂ mean was calculated as the mean value of all residential NO₂ estimates (point estimates) from all children in each DA. DA-level estimates are, in this sense, weighted by residential locations of the population. Descriptive statistics, such as mean NO₂ values for specific subpopulations, were calculated from this geographically linked dataset. Although household income was used in regression models, mean NO₂ values are provided by income quintile, calculated based on before-tax income for all households enumerated in the 2006 Census in each city’s metropolitan area.

Multiple Ordinary Least Squares (OLS) regression models were used to explore associations between NO₂ air pollution estimates and the variables described above at the DA level. Separate analyses were conducted for each city. Models were fit for all children younger than age 18, and also, for the age groups of infant (0 or 1), child (2 to 12), and adolescent (13 to 17). These age groups were chosen based on preliminary analyses of relationships between age and ambient air pollution exposure in the three cities. In all cities and age group combinations, residuals in the OLS models demonstrated spatial autocorrelation (as indicated by Moran’s I statistic); therefore, simultaneous autoregressive (SAR) models were fit concurrently to account for spatial autocorrelation and provide more conservative significance levels for the variables.^{Note 20Note 21} Spatial weights for the SAR model were derived using a Queen’s contiguity matrix, which accounts for all neighbouring DAs in creating the spatial weights matrix.^{Note 21} Lagrange Multiplier test statistics were used to determine if the spatial lag or spatial error models for SAR would be more appropriate in each case.^{Note 21}

Results

Mean NO₂ values across the cities ranged from 19.1 ppb in Montreal to 23.3 ppb in Toronto (Table 1). Mean NO₂ exposures generally rose as household income quintile declined, with children in the lowest quintile having the greatest mean exposure in Toronto and Vancouver (Table 1, Figure 1). In all cities, NO₂ exposure was highest among infants and lowest among adolescents (Table 1). NO₂ exposure of children in lone-parent families and visible minority children generally exceeded city means; the exception was visible minority children in Toronto (Table 1).

The SAR models were generally similar to the OLS models, but provided more conservative estimates of significance, and were a better fit, as evidenced by the greater log-likelihood and lower Akaike Information Criterion values (Tables 2 to 4). In each city, the mean age of children in a DA was negatively associated with ambient air pollution exposure—NO₂ exposure was lower in DAs with older children (Tables 2 to 4). Sex was non-significant in all SAR models (Tables 2 to 4), consistent with the observation that boys’ and girls’ NO₂ exposure was comparable to city means (Table 1).

Except for one model, air pollution in all cities was negatively associated with increasing household income within a DA in OLS and SAR models (Tables 2 to 4). However, for Montreal, household income in the all-ages SAR model was non-significant (p = 0.087), although it was significant in the OLS model and in each age group.

DAs with larger percentages of children in lone-parent families had higher NO₂ exposure in the all-ages models for Toronto and Vancouver. This relationship was also significant for children in the Toronto and Vancouver SAR models, and adolescents in the Vancouver SAR model (Tables 2 and 4). However, in the SAR model, the percentage of children in lone-parent families in a DA was not significantly associated with NO₂ exposure in Montreal (Table 3).

The relationship between household income quintile and air pollution exposure was similar in each city, regardless of visible minority status. In Montreal and Vancouver, however, visible minority children’s NO₂ exposure was generally greater than that of children in the same income quintile who were not members of a visible minority; the exception was children in the lowest income quintile in Montreal (Figure 1). In these two cities, DAs with higher percentages of visible minority children had more exposure to ambient NO₂; this significant relationship (in SAR models) was observed among children overall and children aged 2 to 12 in both cities, and among adolescents aged 13 to 17 in Vancouver (Tables 3 and 4). In Toronto, no consistent difference was apparent between visible minority and white children within income quintiles (Figure 1). Similarly, visible minority status was not significant in the Toronto SAR model for children overall (Table 2).

Discussion

In Canada’s three largest cities, children in lower income DAs were exposed to more NO₂ than were those in higher income DAs. Additionally, in Toronto and Vancouver, DAs with larger percentages of children in lone-parent families had greater exposure, although the statistical significance of this difference varied by age group. These results support the hypothesis that differences in respiratory health between children of low and high SES may, at least in part, be related to differences in exposure to NO₂.^{Note 2} In addition, children in lower income households may be more vulnerable to the effects of air pollution owing to greater innate susceptibility through poor general health or specific health conditions related to low SES.^{Note 22} The combination of the effects of exposure and susceptibility to the effects of air pollution among low and high SES adults are difficult to assess quantitatively together.^{Note 23} Alternatively, some differences in asthma prevalence between children of low and high SES in previous studies may be explained by other, well-documented contributing factors related to lower SES, namely, lower health services use and greater exposure to second-hand smoke in the home, or earlier use of cigarettes^{Note 24Note 25}—factors not assessed in this analysis.

A secondary finding was that ambient NO₂ exposure decreased with increasing age in each city. This decline in exposure may be related to changes in household SES that influence where families choose to live. In general, household income (while employed) rises over time^{Note 26} as children age. As well, home ownership rates increase with parental age.^{Note 27} These trends may allow older families to live in more desirable neighbourhoods. Indeed, a study of the three cities found that families with children younger than age 6 moved away from low income neighbourhoods less frequently than did families without young children.^{Note 28} Desirable neighbourhoods may be farther from major highways and thoroughfares, and therefore, have lower ambient NO₂ than other parts of the city.

A 2014 national study in the United States found greater exposure to ambient NO₂ among visible minority adults than white adults (4.6 ppb); income was less important.^{Note 22} The present analysis found that children in low income households in each city had greater exposure to NO₂, but the role of visible minority status in exposure differed between cities. In Vancouver, lower income visible minority children were exposed to greater concentrations of NO₂ than were white children in all household income quintiles. The relationship between visible minority status and NO₂ exposure was less clear in Toronto and Montreal. Additional research using a larger sample is necessary to determine differences in exposure among specific visible minority groups.

The disparity in NO₂ exposure between children in the lowest and highest household income quintiles (from 1.5 ppb in Toronto to 1.9 ppb in Montreal) was greater than that observed for the other variables. This within-city disparity is close to that reported for school-aged children Malmö, Sweden, where mean residential exposure to NO₂ ranged from 7.6 ppb in the highest income group to 10.2 ppb in the lowest—a difference of 2.6 ppb.^{Note 29}

A 2015 study of adults in the 10 largest Canadian cities found that increments in exposure of 5 ppb in ambient NO₂ were associated with a 5% increased risk of mortality from non-accidental causes and ischemic heart disease, and a 4% increased risk of mortality from respiratory diseases.^{Note 15} Given children’s potentially greater susceptibility to the effects of air pollution, it is possible that differences of this magnitude may have a measurable effect on health outcomes, such as asthma and other respiratory illnesses. A study of children aged 5 to 9 in Toronto reported associations between NO₂ exposure at birth and the risk of developing asthma and wheeze,^{Note 30} although the association was stronger among children with allergic disease.^{Note 31}

Limitations

Despite the significance of the statistical models, the exposure estimates relied on estimates of NO₂ derived from LUR models. Any error introduced during creation of the LUR models, including interpolation error and measurement error (for example, of traffic counts), would be reflected in the point measurements assigned to children in this study. Most of this error was likely mediated by the use of aggregate, rather than individual, point data for the regression models, as well as the use of large samples of children (20% of the population).

Locating the residences of children within the GIS relied on the PCCF+ model, which is highly accurate in urban centres, but does introduce some error through the placement of postal code centroids. Here, residences were mapped by postal code, which is only accurate to the number of residences within each postal code. Use of aggregate (DA) datasets in creating the regression models should mediate some of these errors in locating residences.

The analysis relied on ecological, ambient (outdoor) exposures at a child’s residence to assess differences in exposure. However, exposure to air pollution occurs throughout the day in different environments, such as at school, in traffic, and indoors, thereby contributing to misclassification of exposure. Nonetheless, ambient exposure estimates at a person’s residence have been regularly used in air pollution research to investigate associations with mortality and other health outcomes.^{Note 15Note 32}

Conclusion

In general, the findings of this analysis support the observation that in major Canadian cities, exposure to traffic-related air pollution is higher among children in lower income than in higher income neighbourhoods—ranging from 1 to 2 ppb. Research on adults suggests that differences in exposure of this magnitude may be medically significant, but additional studies are needed to determine if this translates into differences in health outcomes among children. Future studies of asthma prevalence and respiratory infection hospitalizations among children may be warranted to examine within-city differences in air pollution-related health problems.