The influence of community well-being on mortality among Registered First Nations people
by Lisa N. Oliver, Chris Penney and Paul A. Peters
Aboriginal groups in Canada experience higher rates of mortality and morbidity than do non-Aboriginal people.Note 1Note 2Note 3Note 4 A number of studies have compared the health status of Aboriginal and non-Aboriginal populations,Note 1Note 5 but few have explored the broader determinants of health and mortality among Aboriginal populations.Note 6
Lower socioeconomic status, measured by variables such as income, education and employment, is associated with higher levels of mortality.Note 7Note 8 High mortality rates among the First Nations population can be partially explained by individual socioeconomic characteristics, but differences persist when these factors are taken into account.Note 1Note 9
A growing body of research has examined the impact of community characteristics on health.Note 10 Residing in a low-income neighbourhood or community has been associated with poorer health outcomes and increased mortality.Note 11Note 12Note 13 However, studies specifically examining the impact of community characteristics on health outcomes among Aboriginal populations in Canada are rare. One analysis based on the Aboriginal Children’s Survey found that off-reserve First Nations people in neighbourhoods with a higher percentage of renters tended to have poorer mental health even when the influence of individual and family socioeconomic factors was controlled.Note 14 Several studies have linked overcrowding, isolation, income, and community control in First Nations communities with outcomes such as tuberculosis, hepatitis A, and hospitalizations.Note 15Note 16Note 17 However, these studies did not examine the impact of individual and community characteristics simultaneously.
The Community Well-Being index (CWB) contains education, labour force, income, and housing components.Note 18 This study uses the 1991-to-2006 Census Mortality and Cancer Cohort to assess the influence of community factors as measured by the CWB and individual characteristics on the mortality of individuals who identified as Registered First Nations people or Indian band members.
Data and methods
The data are from the 1991-to-2006 Census Mortality and Cancer Cohort, a 15% sample (n = 2,735,150) of the non-institutional population aged 25 or older.Note 7 The cohort was tracked for mortality from June 4, 1991 through December 31, 2006.
Respondents to the 1991 Census long-form questionnaire were eligible to be linked to the Canadian Mortality Database using a probabilistic linkage, as names were not captured on the census database. Eligible census respondents were linked to a 1990 and 1991 nominal tax-filer data list that contained common variables (date of birth, postal code, spousal date of birth) and encrypted names. The census and encrypted names were linked to the Canadian Mortality Database. Because encrypted names were required from the tax-filer data, only tax-filers could be followed for mortality. The probabilistic linkage meant that all groups were not equally likely to be linked. Specifically, people reporting Aboriginal ancestry were less likely to be linked than were those with non-Aboriginal ancestry.Note 19
Everyone aged 25 or older enumerated by the 1991 Census long-form was eligible to be part of the cohort. In most of Canada, the long-form questionnaire was administered to one in five households, but it was administered on a 100% basis to residents of several remote and northern communities, Indian reserves and settlements, and non-institutional collective dwellings. However, 78 Indian reserves and settlements, with a population of approximately 38,000, were not enumerated or incompletely enumerated, and thus, could not be included in the cohort.
Data quality reports found that 3.4% of Canadian residents were missed by the 1991 Census. Compared with the enumerated population, these people were more likely to be young, mobile, low-income, homeless, or of Aboriginal ancestry.Note 20
The present analysis was based on 57,300 cohort members who, on the 1991 Census, reported that they were members of an Indian band or a Registered Indian as defined by the Indian Act. Throughout the remainder of this study, they are referred to as Registered First Nations people. Cohort members who reported their ethnic origin (ancestry) as North American Indian, but did not report membership in an Indian band or being a Registered Indian, were not included.
Socioeconomic characteristics of individuals were derived from the 1991 Census: marital status, labour force status, income adequacy, educational attainment, and housing quality. Three marital status categories were specified: legally married/common-law, divorced/separated/widowed, and single. Three categories of labour force status were used: employed, unemployed, and not in the labour force. The measure of income adequacy was based on quintiles of economic family income; to account for regional differences in income, quintiles were constructed within each Census Metropolitan Area (CMA)/Census Agglomeration (CA) and in all other areas.
Highest level of educational attainment was grouped into four categories: less than secondary school graduation, secondary school graduation, some postsecondary, and university degree.
Housing quality was assessed by a dichotomous variable indicating if respondents lived in a dwelling that needed major repairs.
To track residential mobility, individual census records were linked to postal code data obtained from tax files submitted between 1990 and 2007. For 76.7% of cohort members, complete postal code information was available from the tax-filer data. Among most (99.8%) of those with incomplete information, postal code could be imputed from previous tax-filer years; information for the remaining 0.2% was imputed from census data.Note 19 A change in postal code over the 16-year follow-up signalled a residential move. Three categories were created: non-movers; short movers (within a Census Subdivision or Census Division); and long movers (across Census Divisions or provinces). This does not cover all moves. Because some Indian Reserves/Settlements and small municipalities have a single postal code, moves within these areas would be missed. As well, multiple moves within a single year, or if the tax-filer used a previous address (not the actual address), could not be detected.
The Community Well-Being index (CWB measures the social and economic well-being of communities based on data from the Census of Canada.Note 18 The CWB is calculated at the Census Subdivision (CSD) level. The CWB consists of seven indicators in four equally weighted components: education, income, labour force activity, and housing. These indicators are also important determinants of health.Note 21 The seven CWB indicators are combined into an index between 0 and 100.
For the population aged 20 or older, the CWB education component identifies the percentage whose highest level of attainment is at least secondary graduation, and for the population aged 25 or older, the percentage with at least a university degree. The labour force component consists of labour force participation and percentage of the labour force who are employed or unemployed. The income component is income per capita, calculated as total community income divided by total population. The housing component is comprised of an indicator of crowding (percentage of population in dwellings with more than one person per room) and an indicator of dwelling quality (percentage in dwellings that need major repairs). Information about the CWB and its application to the First Nations population is available elsewhere.Note 18Note 21Note 22Note 23Note 24 The CWB is analysed as a dichotomous variable using the average for First Nations communities in 1991 as a cut-point.
Metropolitan Influence Zone (MIZ) was used to assess the remoteness of a municipality. MIZ assigns a category to municipalities outside a CMA/CA based on the percentage of the employed labour force who commute to work in the CMA/CA. The four categories are: strong, moderate, weak or no MIZ. People living in a moderate or strong MIZ were combined into a single category because preliminary models indicated that mortality did not differ significantly between them.
Age-standardized mortality rates
Age-standardized mortality rates (ASMRs) were calculated per 100,000 person-years at risk by sex for individual and community variables, using the direct method. The population for age-standardization was the age distribution, in five-year age groups, of cohort members who self-identified as Registered Indians under the Indian Act, were members of an Indian band/First Nation, or had Aboriginal ancestry. Variances derived with the Spiegelman method were used to calculate confidence intervals.Note 25
Cox proportional hazards models
For in-scope cohort members, person-days of follow-up were calculated from baseline (June 4, 1991) to the date of death, emigration, or end of the study (December 31, 2006). Cox proportional hazards survival models were used to examine the influence of community factors and individual characteristics on mortality among Registered First Nations cohort members.Note 26
Three sets of models were run. The first set included only the CWB components; the second, only individual characteristics. The purpose of these models was to examine the effects of community and individual characteristics separately.
The third set of models consisted of both the CWB components and individual characteristics in order to determine the extent to which the effects of CWB indicators are attenuated with the addition of individual characteristics.
All models were stratified by sex and five-year age group. Ties were handled using the Efron method.Note 27
First Nations cohort members
At baseline (June 4, 1991), the average age of Registered First Nations cohort members was 40.3 years for men and 39.7 years for women (Table 1). By the end of the follow-up period (December 31, 2006), 16% of the men and 12% of the women had died. Equal percentages of men and women were married (69% and 68%, respectively); men were less likely than women to be divorced or widowed (10% versus 16%), but more likely to be single (22% versus 16%). About 5% of Registered First Nations cohort members were in the highest income quintile, and around 40% were in the lowest. A majority (60% of men and 56% of women) were not secondary school graduates. Half (50%) of men were employed, and 21% were unemployed; the corresponding figures for women were 40% and 11%. A higher percentage of women than men were not in the labour force (49% versus 29%). About three-quarters of cohort members lived in crowded dwellings, and two-thirds were in dwellings that needed major repairs.
By the end of follow-up, 30% of Registered First Nations cohort members had moved within a Census Division, and 20% had moved from one Census Division or province/territory to another.
Age-standardized mortality rates
In 1991, the ASMR for cohort members in communities below the First Nations CWB average was 1,056.6 per 100,000, compared with 912.1 for those in communities above the First Nations average (Table 2). The rate ratio was 1.16, indicating that the mortality rate was 16% higher among those in communities with a below-average CWB score.
Examination of the CWB components separately shows that living in communities with a below-average education score was associated with a 9% higher mortality rate overall. Differences were not significant for male cohort members, but female cohort members in communities with a below-average CWB education score had an 11% higher mortality rate.
The mortality rate for people in communities with below-average income components was 24% higher than for those in communities with above-average income components. Living in a community with a below-average labour force participation component was associated with a 17% higher mortality rate. These differences by income and labour force components prevailed for both sexes.
Mortality rates of residents of communities with above- or below-average housing components did not differ significantly.
Cox proportional hazards models
A series of Cox proportional hazards models demonstrates how the CWB components and the socioeconomic characteristics of individuals were associated with the hazard of death among Registered First Nations cohort members over the 1991-to-2006 period.
Model 1 contains only the CWB components (Table 3). Living in a community with a below-average overall CWB index score increased the hazard of death by 14% for men, and by 17% for women. A below-average community score on the CWB education component raised women’s hazard of death by 11%, but no significant effects emerged for men. Residing in a community with a below-average CWB income score was associated with a 27% increase in the hazard of death for men and a 24% increase for women. For both sexes, a below-average CWB labour force score raised the hazard of death by approximately 17%. The CWB housing component was not associated with mortality for either sex, and was, therefore, excluded from subsequent models.
Model 2 contains only the individual-level characteristics, most of which were significant in the expected direction (Table 4). Compared with being married, being single or divorced/widowed was associated with an increased hazard of death for both sexes. Education was also significant—those with less than secondary school graduation had an increased hazard of death, compared with those reporting higher attainment. Similarly, relative to being employed, being unemployed or not in the labour force increased the hazard of death for both sexes. Being in the lowest income adequacy quintile, rather than in the three highest, raised the hazard of death for both sexes. Also, being in the second-lowest income quintile was significantly associated with an increased hazard of death for women, but not for men. Living in a dwelling that needed major repairs was related to an increased hazard of death for women, but not for men.
Compared with not moving, a short move during the past 16 years was related to a lower hazard of death for both sexes; long moves were not associated with a higher or lower hazard of death. Living in weak MIZ rather than in a CMA or a moderate or strong MIZ was not associated with an increased hazard of death. However, residence in an area that was not a MIZ was associated with a reduced risk of mortality among men, but not women.
Model 3 (Table 5) contains the CWB values and the full set of individual characteristics. Model 3A shows that the overall CWB score was no longer significantly associated with mortality when individual socioeconomic characteristics were considered. However, the picture differs for specific CWB components.
Model 3B shows that living in a community with a below-average CWB income score was associated with an increased hazard of death for men (20%) and women (15%), even when accounting for individual socioeconomic characteristics. By contrast, with the addition of individual socioeconomic characteristics, the CWB education component was no longer significant for women (Model 3C). According to model 3D, residence in a community with a below-average CWB labour force score was associated with an increase (7%) in the hazard of death for men, but not for women.
This study examines how community socioeconomic indicators and individual characteristics were associated with mortality among a large sample of Registered First Nations people.
Mortality was 16% higher among residents of communities with a CWB index score below the average for First Nations communities. For both sexes, living in communities with below-average CWB income and labour force scores was associated with increased mortality. As well, for women, living in communities with below-average CWB education scores was associated with increased mortality. However, these results do not consider the influence of individual characteristics
Consistent with previous work,Note 28Note 29 individual characteristics—income adequacy, educational attainment, and employment—were also associated with mortality among First Nations cohort members. A 2009 study showed that adjusting for socioeconomic factors reduced the mortality hazard ratios between Registered Indians and non-Aboriginal people.Note 5 As well, a 2010 analysis of self-reported survey data found that adjusting for income and education reduced, but did not eliminate, the health disparity between Aboriginal and non-Aboriginal respondents.Note 9 In the present study, when individual characteristics were included in the analysis, the overall CWB index score was no longer significant. Similar results have been reported using survey data.Note 30Note 31
In the calculation of ASMRs and in the survival models, the CWB housing component was not associated with mortality. According to the model that contained only individual characteristics, “dwelling in need of repairs” increased the hazard of death for women; the “crowding” variable was not included, as preliminary models showed few significant results. Using the same cohort, Tjepkema et al.Note 32 reported that adjusting for housing conditions (crowding, home ownership, major repairs) at the individual level did not reduce the hazard ratio for premature death for Status Indians living on reserve.
Registered First Nations cohort members who moved a short distance during the 16-year follow-up period had a reduced hazard of death. Motivations for moving are not known, but people who move within a CD may be doing so for reasons associated with lower mortality (for example, employment). Almost two-thirds (63%) of Registered First Nations cohort members who were non-movers had not completed secondary school, compared with 52% of those who moved a short distance. Because individual socioeconomic characteristics are available only at baseline, it is not possible to determine if moving coincided with changes in characteristics such as income, education, or employment. However, an international longitudinal study reported that working-age rural-to-urban migrants experienced lower mortality than did non-migrants.Note 33
Strengths and limitations
A strength of this study is the use of a large census-linked dataset containing more than 50,000 Registered First Nations respondents. Linkage to tax data over the 16-year follow-up allowed tracking of residential mobility.
However, the findings should be considered in the light of several limitations. The study included only individuals enumerated by the 1991 Census long-form—78 Indian reserves with a total population of about 38,000 were either not enumerated or incompletely enumerated, and therefore, were not part of the cohort. As well, 3.4% of Canadian residents were missed by the census and were more likely to be of Aboriginal ancestry.Note 20 Although census long-form over-sampling of residents of Indian reserves and remote/northern communities means these areas were over-represented in the cohort, those reporting Aboriginal ancestry were the least likely to be linked and included.Note 19
Except for the residential mobility variable, individual characteristics were assessed only at baseline by the census long-form. Consequently, information is lacking about subsequent changes in factors such as income, education, or employment, which may influence mortality.
The CWB index has several limitations. CWB scores were assessed only at baseline, and some cohort members may have moved to communities with higher or lower scores. Furthermore, communities themselves may have experienced an increase or decrease in CWB over the study period, which was not taken into account in this analysis. In fact, from 1991 to 2001, CWB scores of First Nations communities increased at a faster rate than did those in other communities.Note 18 In addition, the CWB index does not cover traditional economic pursuits that do not directly result in monetary income or paid employment.Note 18 This study did not capture all community-level factors, such as material resources (libraries, schools, health care services), social support and community control, that may influence health and mortality.Note 34Note 35
Another limitation is that this analysis was unable to use multilevel modelling techniques that take account of the structured nature of the data for individuals nested in communities. Therefore, the variation in mortality attributed to the community level could not be determined.Note 36 Instead, community-level variables were added at the individual level. Nonetheless, similar techniques have been used with this dataset in provincial analyses.Note 1
This study demonstrates that Registered First Nations members of the 1991-to-2006 Census Mortality and Cancer Cohort who lived in communities with a CWB score below the average for First Nations communities had relatively high ASMRs. The relationship persisted for the CWB income and labour force components even when individual socioeconomic characteristics were taken into account.
- Date modified: