Health Reports
Socioeconomic disparities in life and health expectancy among the household population in Canada

by Tracey Bushnik, Michael Tjepkema and Laurent Martel

Release date: January 15, 2020

DOI: https://www.doi.org/10.25318/82-003-x202000100001-eng

Life expectancy (LE) and health expectancy have increased throughout much of the world,Note 1 including Canada.Note 2Note 3 However, these gains in years lived and years lived in good health are not distributed equally across all population groups. Disparities exist, particularly according to socioeconomic position. Understanding the magnitude, distribution and shift over time in these disparities is increasingly relevant for policy development and planning to advance health equity.Note 1Note 4

It has been reported in many countries—including the United States, Norway, Denmark and Belgium—that people with less education or lower income are disadvantaged in terms of life and health expectancies, and that this disadvantage has persisted or increased over time.Note 5Note 6Note 7Note 8Note 9 Past and current findings suggest that such disparities also exist in Canada.Note 4Note 10Note 11 However, differences in methodologies and data sources have made it difficult to ascertain how, if at all, these disparities have changed with time.

This study uses the 1996 and 2011 Canadian Census Health and Environment Cohorts (CanCHECs), with a five-year mortality follow-up, to estimate the LE of the household population. It also incorporates information from two national health surveys to estimate health-adjusted life expectancy (HALE).

The objectives of this study are to examine LE, HALE and disparities in LE and HALE in the 1996 and 2011 cohorts at ages 25 and 65 for men and women, according to highest level of educational attainment and household income quintile; to examine these disparities according to the combination of education and income in the 2011 cohort; and to examine how education- and income-related disparities in LE and HALE changed over time.

Methods

Data sources

1996 and 2011 CanCHECs

The 1996 and 2011 CanCHECs are population-based linked datasets that follow the non-institutional population at the time of the census for different health outcomes such as mortality, cancer and hospitalizations.Note 12Note 13 In brief, records from census years 1996 and 2011 were linked to mortality data using Statistics Canada’s Social Data Linkage Environment (SDLE). The records from 1996 included mandatory long-form census respondents only,Note 14 aged 19 or older, from about one in five non-institutional households including collectives. The records from 2011 included voluntary National Household Survey (NHS) respondents,Note 15 from about one in three households in private dwellings, and no age restriction. The SDLE helps create linked population data files for social analysis through linkage to the Derived Record Depository (DRD), a dynamic relational database that contains only basic personal identifiers. For this analysis, records were included for individuals who were aged 25 or older on Census Day and who were living in private households. This resulted in an analytical sample of 3,203,700 in the 1996 cohort and 4,526,300 in the 2011 cohort (count rounded to the nearest 100).

Mortality

Mortality data were based on the Canadian Vital Statistics Death Database, which was linked to the DRD. The linkage rate of deaths to the DRD exceeded 99% for 1996 and 2011.

The National Population Health Survey and the Canadian Community Health Survey

Estimates for the Health Utilities Index Mark 3 (HUI3) are derived from responses to the 1994/1995 National Population Health Survey (NPHS) and the 2009 and 2010 Canadian Community Health Survey (CCHS). Information about both surveys is available at www.statcan.gc.ca. The target population of the NPHS household component was residents of private households in the provinces, excluding residents of Indian reserves, Canadian Armed Forces bases, and some remote areas in Ontario and Quebec. The selected household and selected person response rates were 88.7% and 96.1%, respectively. The target population of the CCHS was the household population aged 12 or older in the provinces and territories, with similar exclusions as the NPHS (representing less than 3% of the CCHS target population). The combined household and selected person response rate for the 2009/2010 CCHS was 72.3%.

This study uses data from respondents with a valid HUI3. In general, the non-response rate for HUI3 was less than 1% in either survey year, resulting in an analytical sample size of 15,989 from the NPHS and 121,606 from the CCHS.

Measures

Health Utilities Index Mark 3

HUI3 measures eight attributes of self-reported health status: vision, hearing, speech, ambulation, dexterity, emotion, cognition and pain.Note 16 A respondent’s attribute levels—from normal to highly impaired—are summarized by a weighted scoring function into a single value that represents their overall health state. This value can range from -0.36 (state worse than death; death is represented by 0) to 1.00 (best possible health state).

Highest level of educational attainment

This was the highest certificate, diploma and degree of the individual collected by the census and NHS and the NPHS and CCHS. It was grouped into four separate categories: less than secondary graduation (E1), secondary graduation or trades certificate (E2), postsecondary certificate or diploma excluding university degree (E3), and university degree or equivalent (E4). The proportion of men and women in each category is presented in Appendix Table A.

Income quintiles

Income was self-reported in the NPHS and CCHS, and in the 1996 Census. In the 2011 NHS, 73% of respondents gave permission for income information available from their tax data to be used.Note 17 Weighted quintiles—Q1 (lowest), Q2, Q3, Q4 and Q5 (highest)—were derived from total annual household pre-tax income from all sources, adjusted for household size.Note 18 In the NPHS and CCHS, these were tabulated within each census metropolitan area (CMA) or provincial residual. In the census and the NHS, the weighted quintiles were derived within each CMA, census agglomeration or provincial residual.

Statistical analysis

Life expectancy

The number of deaths by sex, age group and socioeconomic measure (education and income in 1996 and 2011, composite in 2011 only) during a five-year follow-up period were tabulated for each CanCHEC. The number of people who were alive during the follow-up periods (i.e., the at-risk population) by sex, age and socioeconomic measure was also tabulated. Person-years-at-risk were calculated based on census date and date of death or end of follow-up. Since most individuals who were alive during the follow-up period did not remain at the same age for an entire follow-up year, a year-at-risk was partitioned between two ages, and potentially two age groups. For example, someone who turned from age 49 to 50 exactly halfway through the follow-up year contributed 0.5 person-years-at-risk to the 45-to-49 age group, and 0.5 person-years-at-risk to the 50-to-54 age group. A five-year follow-up period was chosen to ensure enough deaths to provide reliable estimates and to minimize mortality overlap in follow-up periods across the different CanCHEC years.

Life expectancy (LE) is the number of years a person at a given age would be expected to live if the mortality rates observed during a specific period persisted throughout their remaining life. For this study, abridged period life tables (based on five-year age groups starting at age 25 and ending at age 90 or older) were calculated according to the Chiang method,Note 19 using deaths and person-years-at-risk from each CanCHEC for men and women according to education category and income quintile, with an additional table based on a composite measure of education and income by sex for the 2011 CanCHEC only. Because of data constraints, a table for the composite measure was not estimated for the 1996 CanCHEC. The cohort weight was applied to ensure that the LE estimates were representative of the target population (people aged 25 years or older in private households on Census Day), and the bootstrap replicate weights were used to estimate appropriate standard errors and 95% confidence intervals.Note 20

Health-adjusted life expectancy

To estimate HALE, mean HUI3 scores by sex and age group according to education category and income quintile were tabulated using the 1994/1995 NPHS and the 2009 and 2010 CCHS, and according to the composite measure of education and income using the 2009 and 2010 CCHS only. The age groups were 25 to 44, 45 to 54, 55 to 64, 65 to 79, and 80 or older. These age groups maximized sample size and were age groups within which mean HUI3 scores—assessed for the full population—remained relatively stable. Survey weights were applied so that the mean HUI3 estimates were representative of the health status of the underlying target populations, and bootstrap weights were applied so that the standard errors were estimated taking into account each survey’s complex design.Note 20 Appendix Table A presents estimates for age groups 25 to 44 and 65 to 79 by education and income categories. A difference of 0.03 or greater in mean HUI3 is considered clinically important.Note 21

HALE was estimated for each cohort using a modified version of the Sullivan method.Note 22 The life expectancy information from each set of CanCHEC-based abridged period life tables was weighted by the number of life-years lived at a particular age x using the mean HUI3 for that age, sex and socioeconomic measure. HALE was obtained by then dividing the sum of the adjusted life-years beyond age x by the number of survivors at that age.Note 23 The HALE variance was estimated using the method proposed by Mathers,Note 24 which takes into account stochastic fluctuations in the observed death probabilities and the mean global HUI3 scores. The ratio of HALE to LE (HALE/LE) was multiplied by 100 and expressed as a percentage. The HALE/LE variance was estimated taking into account the variance of the mortality rates, using the Jagger et al.Note 25 approach (Appendix Equation 1).

Testing of equality

The equality of two estimates of LE, HALE or HALE/LE across groups (disparities) or over time (2011 versus 1996) was conservatively tested by the following Z-scoreNote 25 (Appendix Equation 2).

Results

Education and income

In the 2011 cohort, 17% of men and women had less than a secondary graduation (E1), compared with 23% and 24%, respectively, with a university degree (E4) (Appendix Table 1). In the 1996 cohort, 32% of men and 33% of women were at E1, compared with 16% and 13%, respectively, at E4. In both cohorts, a larger proportion of men were in the highest income quintile (Q5) than in the lowest income quintile (Q1). Among women, 21% were in Q1 and 19% were in Q5 in the 1996 cohort, while 20% were in both Q1 and Q5 in the 2011 cohort. In the 2011 cohort only, 5% of men and 6% of women were in the lowest combined education and income category (E1, Q1), whereas 9% of men and women were in the highest (E4, Q5) (data not shown).

Education-related disparities in LE and HALE

LE25 and HALE25 of men and women increased monotonically from E1 to E4 (Table 1). The disparity in LE25 between E1 and E4 was significantly larger for men (7.8 years) than for women (6.7 years). The disparity in HALE25 was larger than for LE25, but was similar for men (11.3 years) and women (10.6 years). The disparity in HALE25/LE25 was also similar for both sexes. Respectively, men and women in E1 could expect to spend 81% and 79% of their remaining years in good health, compared with 89% and 87% for those in E4. Education-related disparities in LE65, HALE65 and HALE65/LE65 were similar for both sexes, but smaller than at age 25.

Income-related disparities in LE and HALE

There was a positive gradient in LE25, HALE25 and HALE25/LE25 moving from the lowest to the highest income quintile (Table 1). The disparity in LE25 and HALE25 between Q1 and Q5 was significantly larger for men (7.7 and 12.2 years) compared with women (5.4 and 10.1 years), and the disparity in HALE25/LE25 was 10 percentage points for men and 9 percentage points for women. Income-related disparities in LE65 and HALE65 were larger for men than for women, but smaller than at age 25.

Education-within-income-related disparities in LE and HALE

For the most part, the gradient in LE25 and HALE25 for men and women persisted by level of education within and across income quintiles (Figure 1). Among women with a university degree (E4), however, LE varied little by income. Those in the lowest combined socioeconomic category (E1, Q1) had the greatest LE25 and HALE25 disadvantage, which was larger than disadvantages in either E1 or Q1. The LE25 of men and women in E1, Q1 was 13.0 and 9.7 years lower, respectively, than the LE25 of men and women in E4, Q5. The HALE25 disadvantage was even greater: 19.9 fewer HALE25 years for men and 16.2 fewer HALE25 years for women. There was also a gradient in HALE25/LE25 across combined socioeconomic categories. Those in E1, Q1 could expect to spend about 75% of their total life expectancy in good health, compared with 89% to 91% among those in E4, Q5. Gradients in LE65 and HALE65 across combined categories were evident but attenuated, and there was less of an education gradient in HALE65/LE65 within income categories (data not shown).

LE and disparities in LE over time

The gradient in LE25 across education and income categories in 2011 was also present in 1996 (Table 2). Although LE25 increased significantly between 1996 and 2011 for all groups—with a larger increase for men than for women—the gradient in 2011 was steeper for both sexes because of greater gains in LE25 among those with more education or a higher income. This steeper gradient resulted in a significant increase in the disparity between E1 and E4 and between Q1 and Q5 for men and women over the period. The greater relative increase in LE25 over time in favour of E4 versus E1 and in favour of Q5 versus Q1 was higher among women than men. However, in absolute terms, men gained more years of LE25 and had larger disparities in LE25 than women. LE65 also increased over the period for all groups, as did the disparity between E1 and E4 for men and between Q1 and Q5 for both sexes (Appendix Table B).

HALE and disparities in HALE over time

Like LE25, the gradient in HALE25 in 2011 was also evident in 1996, but was steeper in 2011 because of greater gains in HALE25 over time among people with more education or a higher income (Table 3). The gains in favour of E4 versus E1 were higher among men than women, whereas the gains in favour of Q5 versus Q1 were higher among women than men. As a result, there was a significant increase in the disparity between E1 and E4 over the period for men (2.7 years, p=0.003), but not for women (1.5 years, p=0.149), whereas the disparity between Q1 and Q5 significantly increased for women (2.6 years, p=0.030), but not for men (1.1 years, p=0.202). Like LE25, men generally had larger disparities in HALE25 in absolute terms than women. There was less of a gradient to the increase over time in HALE65 across education and income categories, resulting in the disparities in HALE65 remaining relatively unchanged (Appendix Table B).

Discussion

This study found that disparities in LE and HALE still exist in Canada. People with higher levels of education or a higher income have longer life expectancies and are expected to spend a greater portion of those years in good health compared with those with less education or with a lower income. A distinct stepwise gradient in LE and HALE also exists by level of education within and across income quintiles. There is evidence that disparities are wider than they were 15 years ago, but not necessarily to the same extent for both sexes or at different ages.

The pathways through which socioeconomic position can affect health outcomes are multi-factorial and complex.Note 26 Education and income are frequently used as indicators of socioeconomic position in health disparities researchNote 27 and, though related, are not considered interchangeable.Note 28 Education is widely thought to increase health knowledge and literacy, which in turn can promote the adoption of healthier lifestyles and facilitate access to appropriate health care.Note 26Note 29 Higher income allows access to better-quality material resources—such as food and shelter—and better, easier or faster access to services, which can have a direct (e.g., health services) or indirect (e.g., education) effect on health.Note 27 That this study found gradients in LE, HALE and HALE/LE by education or income is consistent with other studiesNote 29Note 30Note 31 and speaks to the well-recognized role of social stratification in determining health outcomes.Note 32 Moreover, the stepwise gradient in LE and HALE by education level within income strata underscores how multiple aspects of social disadvantage can intersect in their association with health outcomes.Note 29 This is emphasized by the finding that people in the lowest combined socioeconomic categories were at a greater LE and HALE disadvantage than those in either a low education category or low income quintile.

Many studies have examined education-related disparities because of the availability and appeal of education as an indicator of socioeconomic status.Note 11Note 26 Although differences in data sources, methodologies and definitions limit the direct comparability of these studies, it is possible to compare overall patterns and trends. This study found significant disparities in LE at ages 25 and 65 for both sexes between the lowest and highest education categories, and greater gaps for men than for women. These findings are consistent with what has been reported in many Organisation for Economic Co-operation and Development countries.Note 11Note 29Note 30 The widening education gap in LE between 1996 and 2011 for both sexes that was reported in this study has also been reported elsewhere.Note 5Note 9Note 33 This widening has been partly attributed to the significant decline in the population size of the lowest education category, a category that is thought to be increasingly composed of individuals with characteristics that compound the risk of ill health and death.Note 34 This study found that the health status of those in the lowest education category declined between 1996 and 2011. However, it has been noted that compositional change cannot fully account for the worsening LE of those with the lowest education, particularly among women.Note 33

This study’s finding of significant and widening income-related disparities in LE has also been reported by others,Note 6Note 35Note 36 despite significant heterogeneity in the way income has been defined (e.g., career earnings versus tax data, linked at the individual level or area-based). Alcohol and smoking have been identified as contributing substantially to income differences in LE.Note 37Note 38 Negative health behaviours such as these might also help explain why people in the lowest combined education and income category in this study were at the greatest LE disadvantage.

Health expectancy goes beyond LE by estimating the number of years a population may expect to live in good health.Note 1 Disparities in HALE—this study’s measure of health expectancy—and in HALE/LE were significantly larger at age 25 across education and income for both sexes than disparities in LE. In other words, people with the highest education and income were not only living longer than those with a lower income or less education, but were also spending a greater share of those years in good functional health. This is consistent with other published work, regardless of the measure used to estimate health expectancy,Note 9Note 39Note 40 and suggests that disparities may be more pervasive with respect to quality rather than quantity of remaining life.Note 31 However, the disparity in HALE65 was considerably smaller than the disparity in HALE25. This may reflect the “age as leveller” theory, which suggests that earlier gaps in healthy life will narrow in advanced age.Note 41

Disparities in HALE25 widened between 1996 and 2011 for men across education categories and for women across income quintiles. The former partly reflects the fact that the functional health of men in the lowest education category declined over time, while it remained relatively unchanged for men in the highest education category. The reverse occurred for women. The functional health of women in the lowest income quintile remained stable, whereas it increased for the women in the highest income quintile. These findings suggest there may be differences by sex in how the association between different components of social disadvantage and health outcomes may be evolving over time.

This study has many strengths. The CanCHECs are large, nationally representative cohorts that were created using a consistent methodology and allow for a robust examination of change over time. Two important social determinants of health—education and income—measured at the individual level for people aged 25 or older were examined. Education has the advantage of having a low risk of “reverse causality” with health,Note 11 while household income is a strong indicator of material living standards.Note 32 Rarely have both determinants at the individual level been included in the same study of disparities in LE or HALE. HALE was estimated using HUI3, which is a continuous scale. This makes it less sensitive to measurement error than dichotomous estimates of health status.

Several limitations should be acknowledged. The analysis did not account for other population characteristics (e.g., ethnicity, marital status) that may have changed over time within education or income categories and may have, in turn, contributed to the observed disparities. The results pertain solely to the household population since data constraints prevented the institutional population from being included in this study. Excluding the institutional population has been shown to significantly increase population estimates of HALE and HALE/LE.Note 2 Changes over time in question wording, collection mode and response rates for each census and health surveyNote 14Note 15Note 42Note 43 should be kept in mind when interpreting the results of this study. Specifically, revisions introduced in the 2006 Census to correct the underreporting of high school completion that had occurred previously,Note 44 coupled with a potential increase over time in the homogeneity of people in the lowest education category, may have complicated the interpretation of trends. Additionally, 2006 was the first census year in which respondents were given the option to allow linkage to their tax records rather than self-report their income data. This reduced the clustering around “round” dollar amounts, such as $30,000, which then increased the variability in the income distribution compared with previous censuses.Note 45 However, the fact that this study derived income quintiles for each cohort separately helped circumvent the potential impact of distributional change in income groups over time.Note 46 Although the cohort weights were designed to help mitigate bias associated with data linkage, unknown bias might exist if people missing from the cohorts differed systematically from those who were included.

Conclusion

Education- and income-related disparities in life and health expectancy persist and may be wider than they were 15 years ago among the household population in Canada. These findings underscore the importance of ongoing data development for routine monitoring of trends in mortality and morbidity, which in turn can inform policy development and planning to advance health equity.

Acknowledgements

The authors gratefully acknowledge the help of Philippe Finès, who provided the syntax to estimate life expectancy and health-adjusted life expectancy based on the Canadian Census Health and Environment Cohorts.

Appendix

Equation 1
Calculation of variance of the ratio of HALE to LE (HALE/LE)

Var ( HALE LE )= HAL E 2 L E 2 *[ Var( HALE ) HAL E 2 Var( notHALE )Var( LE )Var( HALE ) HALE*LE + Var( LE ) L E 2  ] MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaaeaaaaaaaaa8 qacaWGwbGaamyyaiaadkhacaGGGcWaaeWaa8aabaWdbmaalaaapaqa a8qacaWGibGaamyqaiaadYeacaWGfbaapaqaa8qacaWGmbGaamyraa aaaiaawIcacaGLPaaacqGH9aqpdaWcaaWdaeaapeGaamisaiaadgea caWGmbGaamyra8aadaahaaWcbeqaa8qacaaIYaaaaaGcpaqaa8qaca WGmbGaamyra8aadaahaaWcbeqaa8qacaaIYaaaaaaakiaacQcadaWa daWdaeaapeWaaSaaa8aabaWdbiaadAfacaWGHbGaamOCamaabmaapa qaa8qacaWGibGaamyqaiaadYeacaWGfbaacaGLOaGaayzkaaaapaqa a8qacaWGibGaamyqaiaadYeacaWGfbWdamaaCaaaleqabaWdbiaaik daaaaaaOGaeyOeI0YaaSaaa8aabaWdbiaadAfacaWGHbGaamOCamaa bmaapaqaa8qacaWGUbGaam4BaiaadshacaWGibGaamyqaiaadYeaca WGfbaacaGLOaGaayzkaaGaeyOeI0IaamOvaiaadggacaWGYbWaaeWa a8aabaWdbiaadYeacaWGfbaacaGLOaGaayzkaaGaeyOeI0IaamOvai aadggacaWGYbWaaeWaa8aabaWdbiaadIeacaWGbbGaamitaiaadwea aiaawIcacaGLPaaaa8aabaWdbiaadIeacaWGbbGaamitaiaadweaca GGQaGaamitaiaadweaaaGaey4kaSYaaSaaa8aabaWdbiaadAfacaWG HbGaamOCamaabmaapaqaa8qacaWGmbGaamyraaGaayjkaiaawMcaaa WdaeaapeGaamitaiaadweapaWaaWbaaSqabeaapeGaaGOmaiaaccka aaaaaaGccaGLBbGaayzxaaaaaa@834C@

where Var = variance and notHALE = the difference between LE and HALE.

Equation 2
Testing the equality of two estimates of LE, HALE or the ratio of HALE to LE (HALE/LE)

Zscore= (HA)L E 1 ( HA )L E 2 ( S 2 ( HA )L E 1 + S 2 ( HA )L E 2 ) MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaaeaaaaaaaaa8 qacaWGAbGaeyOeI0Iaam4CaiaadogacaWGVbGaamOCaiaadwgacqGH 9aqpdaWcaaWdaeaapeGaaiikaiaadIeacaWGbbGaaiykaiaadYeaca WGfbWdamaaBaaaleaapeGaaGymaaWdaeqaaOWdbiabgkHiTmaabmaa paqaa8qacaWGibGaamyqaaGaayjkaiaawMcaaiaadYeacaWGfbWdam aaBaaaleaapeGaaGOmaaWdaeqaaaGcbaWdbmaakaaapaqaa8qacaGG OaGaam4ua8aadaahaaWcbeqaa8qacaaIYaaaaOWaaeWaa8aabaWdbi aadIeacaWGbbaacaGLOaGaayzkaaGaamitaiaadweapaWaaSbaaSqa a8qacaaIXaaapaqabaGcpeGaey4kaSIaam4ua8aadaahaaWcbeqaa8 qacaaIYaaaaOWaaeWaa8aabaWdbiaadIeacaWGbbaacaGLOaGaayzk aaGaamitaiaadweapaWaaSbaaSqaa8qacaaIYaaapaqabaaapeqaba GccaGGPaaaaaaa@5CBE@

where (HA)LE = (health-adjusted) life expectancy, or HALE/LE, and S2(HA)LE = variance of the (health-adjusted) life expectancy, or HALE/LE.

References
Report a problem on this page

Is something not working? Is there information outdated? Can't find what you're looking for?

Please contact us and let us know how we can help you.

Privacy notice

Date modified: