Mortality rates in a cohort of patients undergoing pulmonary function testing in Hamilton, Ont., varied by location of residence. After adjustment for age, sex, lung function, BMI and the diagnosis of chronic diseases, we found that people living in lower-income neighbourhoods had higher mortality rates than those living in wealthier neighbourhoods.
Explanations for income-associated differences in mortality include biological and sociological factors. Surveys have found that lower-income Canadians have the highest prevalence of cardiovascular risk factors, particularly smoking and excess weight.14
We did not have information about individual smoking habits and attempted adjustment for biological risk factors by controlling for BMI, lung function and the diagnosis of chronic diseases. Dockery and associates15
reported that “smokers suffer an irreversible loss of FVC and FEV1
, which is described by a linear function of their cumulative cigarette smoking as measured in pack-years.”
We believe that inclusion of lung function and the diagnosis of chronic pulmonary and ischemic heart diseases controls for much of the effect of smoking. However, it is likely that some residual confounding of the income–mortality relation remains.
has argued that the most important factors pertaining to income-associated differences in mortality are the broader determinants of health, including education, employment, and physical and social environments. In the Hamilton–Burlington urban airshed, we found an association between income and estimated air pollution exposure in the vicinity of the study population's residences. People living in lower-income neighbourhoods tended to have higher mean levels of pollution exposure than those living in higher-income neighbourhoods. Those living in neighbourhoods with above-median levels of particulate and sulfur dioxide air pollution had higher mortality rates than those living in neighbourhoods with below-median levels of such pollution.
Our study population, derived from the database of a lung function testing laboratory, was enriched with subjects with CPD. It has been suggested that people with COPD are more susceptible to the effects of acute increases in levels of air pollution.17
In our population, the relative risk of death for those with CPD was 1.4 in the Cox regression analysis; however, the mortality risk associated with residence in neighbourhoods with long-term exposure to above-median levels of air pollution was independent of underlying CPD status.
Although our patients were not a randomly selected sample of the population, we believe that our results can be generalized. The variability of mortality rates by socioeconomic status has been observed elsewhere in Canada.18
The variation of pollution levels by neighbourhood income in Hamilton was independent of the identities of people in our study population. Neither of the American air pollution cohort studies3,4
used estimates of income level and instead used education level as the measure of socioeconomic status. In both of the American studies, increased mortality risk was observed only in the subset of the population without a postsecondary education (41.1% of those in the American Cancer Society study and 66% in the 6-city study).19
Explanations for this finding of apparent immunity to the adverse effects of pollution among the better educated have been discussed.20,21
An important longitudinal study of a nationally representative sample of 3617 Americans reported that differences in mortality by education level were explained in full by the strong association between education and income.22
We thus suspect that the educational differential in the American studies reflects a misclassification introduced by assigning the same exposure level to all subjects. Less educated people probably lived in poorer neighbourhoods and experienced higher exposures than the better educated people. Their higher mortality rates are consistent with our findings in Hamilton.
Our study had a number of limitations. Household income was estimated from census data, so some misclassification is likely. Information was unavailable about change of residence during the observation period. The levels of pollution exposure were estimated by interpolation from a network of fixed sampling stations; this measurement error would tend to bias the regression slope toward null. In addition, we used mean levels of total suspended particulates and sulfur dioxide measured during the 3 years when there were maximum numbers of sampling stations and applied the levels to the entire 8-year observation period. There was little change in pollution levels in this airshed during the late 1990s;8,23,24,25,26
thus, it is unlikely that substantial error was introduced. Other pollutants (e.g., carbon monoxide, nitrogen dioxide and ozone) are also likely to be associated with differences in mortality,27
but data were unavailable to perform intra-urban interpolations for these pollutants.
In conclusion, mortality rates varied by neighbourhood of residence in our study cohort. At least part of this variation is likely related to differences in biologic risk factors that were not controlled for. Two of the broader determinants of health — income and air pollution levels — were important correlates of mortality in this population.