Elsevier

Social Science & Medicine

Volume 59, Issue 7, October 2004, Pages 1435-1447
Social Science & Medicine

Socioeconomic determinants of mortality in two Canadian provinces: Multilevel modelling and neighborhood context

https://doi.org/10.1016/j.socscimed.2004.01.024Get rights and content

Abstract

The influence of individual and contextual socioeconomic variables on mortality is compared in two Canadian provinces, Manitoba and Nova Scotia. Although differing substantially in size, ethnic mix, and history, both provinces provide greater access to health and social services as well as fewer income inequalities than the United States. A total of 8032 Manitoba respondents (followed from 1996–97 to 2002) and 2116 Nova Scotia respondents (followed from 1990 to 1999) were linked to the appropriate Canadian census as a source of neighborhood characteristics.

Data were analyzed using individual- and multi-level logistic regression. Well-educated and higher income individuals were less likely to die during follow-up. No significant direct effect was found between neighborhood socioeconomic characteristics and mortality. However, both provinces showed an increased importance of individual income vis-à-vis mortality in advantaged neighborhoods relative to disadvantaged neighborhoods. Additional Manitoba analyses showed a “healthy mover” effect among respondents changing place of residence, regardless of whether they moved to more advantaged or more disadvantaged neighborhoods. The findings are discussed in the context of differences in health and health care among Canada, the United States, and other OECD countries.

Introduction

The relationship between individual socioeconomic characteristics and health status has been well documented; higher socioeconomic status is generally associated with better health (Smith, Bartley, & Blane, 1990; Kaplan, 1996). This association holds true in Canada, despite national health insurance and smaller income inequalities than in the United States (Veugelers, Yip, & Kephart, 2001; Mustard, Derksen, Berthelot, Wolfson, & Roos, 1997; Wolfson & Murphy (2000), Wolfson & Murphy (1998)).

However, studies examining neighborhood-level variables and their effect on individual health status have produced diverse results (McMichael, 1999). Many believe community context independently affects the health of all residents (Yen & Kaplan, 1999; Pampalon, Duncan, Subramanian, & Jones, 1999; Mitchell, Gleave, Bartley, Wiggins, & Joshi, 2000; Malmstrom, Johannson, & Sundquist, 2001; Kolegard, Diderichsen, Reuterwall, & Hallqvist, 2002). Others have reported neighborhood-level effects to be due entirely to compositional effects, to an aggregation of the relationships between individual socioeconomic context and individual health status (Sloggett & Joshi, 1994; Duncan, Jones, & Moon, 1995; Bobak, Pikhart, Rose, Hertzman, & Marmot, 2000; Sturm & Gresenz, 2002). Several papers have shown contextual effects to vary by gender (Kolegard et al., 2002), age (Waitzman & Smith, 1998), and neighborhood type (Jessop, 1992; Haynes & Gale, 1999).

Shouls, Congdon, and Curtis (1996) and Yen and Kaplan (1999) noted greater morbidity among individuals in deprived areas (of the United Kingdom and the United States) even after taking individual characteristics into account. In Nova Scotia, Canada Veugelers et al. (2001) found no direct association with neighborhood-level factors. Contextual factors were suggested to be of less importance in Canada than in the United States because of the higher levels of social services provided in Canada. Despite differing results regarding the direct effect of neighborhood characteristics on mortality, Shouls et al. (1996), Yen and Kaplan (1999), and Veugelers et al. (2001) reported a cross-level effect of neighborhood status on the relationship between individual income and mortality.

This paper replicates and extends the Nova Scotia work of Veugelers et al. (2001) in Manitoba, another Canadian province. Using two sites alleviates risks of Type 1 error (inferring a relationship when one does not actually exist) due to conducting a number of statistical tests (implicit in multilevel modelling) with moderate statistical power. Comparing the effects of socioeconomic determinants on mortality in each province using surveys differing in administration, in time period covered, and in length of follow-up increases generalizibility. The two provinces’ substantial differences in size, ethnic mix, and history also add to the generalizability of the results.

Manitoba data permitted assessing the role of two additional variables: urban/rural residence and residential mobility. Glass and Balfour (2003) have pointed out that little research has compared the income and health associations found in urban areas to those in suburban and rural communities. Urban neighborhoods show more residential segregation by income than rural neighborhoods (Wilkins, Berthelot, & Ng, 2002); therefore, differences in health among income groups in urban areas might be expected to be greater than those in rural areas.

Residential mobility information allows for comparison of mortality among movers and non-movers. Individuals “exposed” to a neighborhood for a longer period of time may show more area-specific health effects than those recently moving to that area (Robert, 1999; Lynch, Kaplan, & Shema, 1997). Selective migration of the healthy out of disadvantaged neighborhoods may leave behind “unhealthy survivors” (predisposed toward early mortality) (Macintyre & Ellaway, 2003).

This research helps determine both the generalizability of neighborhood effects and the importance of the urban/rural and mobility variables. Manitoba and Nova Scotia provide similar universal access to basic health and social services and are representative of a distinctly Canadian approach. Such work facilitates the meaningful comparison between Canadian work on socioeconomic determinants of health and that in other countries.

Section snippets

Individual characteristics

Interviews gathered information on lifestyle and other individual characteristics from non-institutionalized residents aged 18–75 years (18 years or older and less than 75 years of age). Individual characteristics were taken from the 1990 Nova Scotia Nutrition Survey and from Manitoba respondents to the 1996–97 National Population Health Survey. Individual variables included age, gender, smoking status, diabetes, body mass index (BMI), household income, and education. Essentially complete

Individual and neighborhood characteristics

Table 1 shows individual and neighborhood characteristics and their age- and gender-adjusted ORs for respondents’ mortality. The age- and gender-adjusted mortality risk was significantly greater among smokers and diabetics in both provinces; BMI of less than 20 was also associated with higher mortality in Nova Scotia (Table 1). Mortality ORs adjusted for age, gender, smoking status, diabetic status, and BMI are presented in Table 2.

In Manitoba, individual household income was significantly

Discussion

This research has combined a multilevel approach and a comparative, longitudinal design to examine the role of individual and neighborhood socioeconomic characteristics. Well-educated, higher earning respondents live longer, even after controlling for age, gender, smoking status, BMI, and diabetes. In both provinces, after controlling for individual-level variables, no neighborhood-level socioeconomic characteristics were found to have a statistically significant direct effect on individual

Acknowledgements

This work was funded by the Canadian Population Health Initiative. Ms. Magoon benefitted by a studentship from the Western Regional Training Centre, supported by the Canadian Health Services Research Foundation. A Canadian Institutes of Health Research Career Award to Dr. Veugelers provided additional support. The results and conclusions are those of the authors, and no official endorsement by Manitoba Health or the Nova Scotia Department of Health was intended or should be implied. We are

References (54)

  • R.B Deber et al.

    Canadian health expendituresWhere do we really stand internationally?

    Canadian Medical Association Journal

    (1999)
  • A.V Diez Roux

    The examination of neighborhood effects on healthConceptual and methodological issues related to the presence of multiple levels of organization

  • Duchesne, D., Nault, F., Gilmour, H., & Wilkins, R. (1999). Vital statistics compendium 1996. Ottawa, Ont.: Statistics...
  • C.C Duncan et al.

    Psychiatric morbidityA multilevel approach to regional variations in the UK

    Journal of Epidemiology and Community Health

    (1995)
  • R.G Evans

    Interpreting and addressing inequalities in health: From Black to Acheson to Blair to…?

    (2002)
  • K.R Gabriel et al.

    Simultaneous confidence intervals for product-type interaction contrasts

    Journal of the Royal Statistical Society, Series B

    (1973)
  • T.A Glass et al.

    Neighborhoods, aging, and functional limitations

  • K.M Gorey et al.

    An international comparison of cancer survivalMetropolitan Toronto, Ont.. and Honolulu, Hawaii

    American Journal of Public Health

    (2000)
  • K.M Gorey et al.

    An international comparison of cancer survivalRelatively poor areas of Toronto, Ont., and three US metropolitan areas

    Journal of Public Health Medicine

    (2000)
  • S Gupta et al.

    Delivering equitable careComparing preventive services in Manitoba, Canada

    American Journal of Public Health,

    (2003)
  • E.B Hook

    Letter to the Editor: Re: Neighborhood social environment and risk of deathMultilevel evidence from the Alameda county study

    American Journal of Epidemiology

    (2000)
  • E.G Jessop

    Individual mortality and neighbourhood deprivation in a non-metropolitan area

    Journal of Epidemiology and Community Health

    (1992)
  • G.A Kaplan

    People and placesContrasting perspectives on the association between social class and health

    International Journal of Health Services

    (1996)
  • S.M Kolegard et al.

    Socioeconomic context in area of living and risk of myocardial infarctionResults from Stockholm heart epidemiology program (SHEEP)

    Journal of Epidemiology and Community Health

    (2002)
  • L.K Korn et al.

    Epidemiologic studies utilizing surveysAccounting for sampling design

    American Journal of Public Health

    (1991)
  • B.G Link et al.

    Social epidemiology and the fundamental cause conceptOn the structuring of effective cancer screens by socioeconomic status

    Milbank Quarterly

    (1998)
  • Cited by (84)

    • Education system stratification and health complaints among school-aged children

      2019, Social Science and Medicine
      Citation Excerpt :

      Increased social homogeneity through stratification therefore means that socially disadvantaged pupils to a larger extent compare themselves with peers of a similar social status, making social status competition from more advantaged peers less salient (cf. Dupriez et al., 2008). While research on social homogeneity and health inequalities is not unequivocal (Goodman et al., 2003), several studies suggest that social (or ethnic) homogeneity have a protective effect for the health, including somatic health, of disadvantaged groups (Wolinsky et al., 2012; also Roos et al., 2004), probably because of psychosocial health benefits such as less risk of stigmatization and poor self-concept when the minority status becomes less salient (Goosby and Walsemann, 2012). In sum, we can expect that stratification can shape the extent of health inequalities by altering the social distribution of various health determinants related to the school learning and social environment (cf. CSDH, 2008).

    • Inequality, neighbourhoods and welfare of the poor

      2016, Journal of Development Economics
    • Effects of individual and neighborhood socioeconomic status on the risk of all-cause mortality in chronic obstructive pulmonary disease: A nationwide population-based cohort study, 2002-2013

      2016, Respiratory Medicine
      Citation Excerpt :

      Despite the analysis showed that there was not an interaction between SES and lifestyle factors, we could not conclude whether smoking and the other lifestyle factors were confounder or not effect modifier because our health behaviors information could not be representative our country. The second mechanism is that socio-cultural norms regarding healthy lifestyles are more easily cultivated in advantaged neighborhoods than in disadvantaged neighborhoods [18]. The third possible explanation involves psychosocial factors.

    View all citing articles on Scopus
    View full text