Article Text

Download PDFPDF

Trends in educational inequalities in smoking and physical activity in Canada: 1974–2005
  1. P Smith1,2,
  2. J Frank2,3,4,
  3. C Mustard1,2
  1. 1
    Institute For Work & Health, Toronto, Canada
  2. 2
    Dalla Lana School of Public Health, University of Toronto, Canada
  3. 3
    Scottish Collaboration for Public Health Research & Policy, MRC Human Genetics Unit, Western General Hospital, Edinburgh, UK
  4. 4
    Chair in Public Health Research and Policy, University of Edinburgh, Edinburgh, UK
  1. Dr P Smith, Institute for Work & Health, 481 University Ave, Suite 800, Toronto, ON, Canada M5G 2E9; psmith{at}iwh.on.ca

Abstract

Background: In Canada levels of smoking have decreased and levels of physical activity have increased over the last 20 years. However, little research has examined if educational inequalities in either of these important health determinants have changed.

Methods: A secondary analysis of Canadian population-based surveys from 1974 through to 2005 was conducted. The prevalence of both smoking and physical activity across educational groups for both men and women, as well as relative and absolute measures of inequality was estimated.

Results: Differences in both smoking and physical activity across educational groups in all surveys examined between 1974 and 2005 were found, with lower educational groups more likely to be heavy smokers and inactive in each survey. Both relative and absolute educational inequalities in smoking widened between 1974 and 2005 (relative concentration index (RCI) for smoking 10 plus cigarettes per day changed from −7.9 to −26.9 among men; and from −4.8 to −27.4 among women), with inequalities in physical activity narrowing between 1981 and 1996, then widening between 1996 and 2005 (RCI for inactivity −4.34 to −6.75 among men; −3.57 to −5.54 among women). In general, results among men and women did not differ substantially.

Conclusions: It is unlikely that the widening educational inequalities in smoking and physical activity documented here reflect lower knowledge of the consequences of smoking and physical inactivity among lower educated groups. The results suggest more work needs to be done in both designing population health approaches that focus on equity and the creation of supportive environments that provide equal opportunities for behaviour change for all educational groups in Canada.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Smoking and physical inactivity are important determinants of (ill) health.1 Specifically in relation to socioeconomic health inequalities, both are risk factors for cardiovascular disease, which continues to be both the leading cause of mortality in Canada and the leading cause of socioeconomic inequalities in potential years of life lost before age 75.2 3 Reports using Canadian data suggest that in general levels of smoking have decreased, and levels of physical activity have increased over the last 20 years.1 While comparable data are not available in Canada, reports from other developed countries suggest that changes in health risk behaviours are thought to explain at least half (or more) of the declining incidence of cardiovascular disease mortality.48

However, monitoring general trends in health behaviours at the population level does not let us know whether particular groups are benefiting more, or less, from population level health behaviour interventions or trends. Previous Canadian analyses reporting that rates of smoking and inactivity are higher in population groups with the fewest resources.9 10 Further, it is likely that those with the most resources will adapt to new situations, and derive more benefits from population level interventions, possibly leading to greater inequalities in health behaviour patterns across socioeconomic groups.11 12 In short, effective population health strategies will require efforts beyond individualised messages to recognising the broader social determinants of behaviour change.13

Surprisingly little research in Canada has examined temporal trends in health behaviours across socioeconomic groups. Sheilds and Tjepkema14 recently reported that differences in the prevalence of obesity, using objectively measured height and weight, had reduced across household income quartiles over the period 1986–1992 to 2004, due to increases in the prevalence of obesity among higher income groups, in particular among men. However, a consistent inverse relationship between level of education and obesity still remains for both men and women.15

The objective of this article is to examine trends in socioeconomic inequalities in smoking and physical activity using Canadian population-based surveys from 1974 through to 2005.

METHODS

This study used secondary data from various population-based Canadian Health Surveys. These included the 1974 Smoking Habits of Canadians, the 1981 Canadian Fitness Survey, the 1990 Health Promotion Survey, the 1996 National Population Health Survey and the 2000 and 2005 Canadian Community Health Surveys. Each of these surveys used a multi-staged, stratified design to collect information. In each case respondents who were inmates of institutions, full-time members of the armed forces, and residents of the Yukon and Northwest Territories, and Indian Reserves were excluded (approximately 3% of the Canadian population). For each survey information was collected directly from respondents using either in-person or telephone interviews, or a combination of both. These databases are described in more detail below.

The 1974 Smoking Habits of Canadians Survey (SHC), conducted by Health and Welfare Canada (now Health Canada), was designed to examine trends in the smoking behaviour of the Canadian population 15 years of age and over. The survey was administered as an annual supplement to the Labour Force Survey (LFS). Response rates to the LFS generally consist of about 95% of eligible households.16 The 1981 Canadian Fitness Survey (CFS) was designed to assess the physical recreation habits, physical fitness and health status of a representative sample of the Canadian population aged 7–69 years of age. The CFS had a household response rate of 88%, with a selected person response rate of 76%.17 The 1990 Health Promotion Survey (HPS) was conducted by Health Canada to assist in the planning of programmes to encourage Canadians to adopt and maintain healthy lifestyles. The household response rate was 87.1%, with a selected person response rate of 96.7%.18 The 1996 National Population Health Survey (NPHS) and the 2000 and 2005 Canadian Community Health Surveys (CCHSs) were all conducted by Statistics Canada. The household response rate to the NPHS was 94.3%, with the selected person response rate being 98.7%.19 The household response rate to the CCHS’s were 91.4% in 2000 and 84.9% in 2005, with the selected person response rates being 91.9% in 2000 and 92.9% in 2005.20 21

MAIN OUTCOMES

Smoking status

Smoking status was taken from the 1974 SHC, 1990 HPS, 1996 NPHS and 2000 and 2005 CCHS surveys. Respondents were asked if they ever or currently smoked cigarettes, and the number of cigarettes they currently smoked. From these questions we grouped respondents into five categories: heavy smokers (more than 10 cigarettes per day); light smokers (1–10 cigarettes per day); occasional smokers (those smoking less than 1 cigarette per day); former smokers (those who have ever smoked, but do not currently smoke); and never smokers. There was a slight change in the wording of the questions assessing smoking status across surveys. In the surveys up to and including the 1996 NPHS, respondents were asked if they had ever smoked (yes or no). In the 2000 and 2005 CCHs the wording of this question changed to asking respondents if they had ever smoked a whole cigarette (yes or no).

Leisure time physical activity

The 1981 CFS, 1996 NPHS and the 2000 and 2005 CCHS surveys asked respondents to report on their physical activity levels. The 1981 CFS asked respondents to report over the previous month, and previous 12 months, while the 1996 NPHS and 2000 and 2005 CCHS’s asked respondents to report physical activity levels over the previous 3 months. In each survey, reports of duration and frequency of numerous different activities are combined to calculate the average daily energy expenditure in the previous 12 months (kcal/kg/day). Given changes in physical activity trends over time there were slight differences in the activities reported (eg, snowboarding was not reported in the 1981 CFS or the 1996 NPHS); however, in general activities reported in each survey cycle could be grouped into 21 different activity groups. Consistent metabolic equivalent (MET) values for each activity were assigned to calculate energy expenditure. Further details of activities and MET values are presented in Appendix A (available online). Energy expenditure was grouped into the following three categories: inactive (less than 1.5 kcal/kg/day); moderately active (1.5 to 2.9 kcal/kg/day); and active (greater than or equal to 3.0 kcal/kg/day).

MAIN INDEPENDENT VARIABLE

Highest level of education completed

Highest level of education was chosen to represent one aspect of each individual’s socioeconomic status. Education level was chosen over occupational position or household income—the two other most commonly used measures of socioeconomic position—because unlike occupation, it is applicable to both labour force and non-labour force participants; and unlike household income, it is accurately reported without a high percentage of missing responses. In addition, education level is the only one of these measures that has been consistently reported across each survey. Educational attainment was grouped into four levels: less than secondary education; secondary education completed; post-secondary diploma or certificate completed; and university education (bachelors or higher) completed.

ANALYSIS

In each survey, the sample was restricted to respondents aged between 25 and 64 years of age, with valid responses for age, education and the outcome of interest. To account for differences in age across gender and educational groups, estimates within each education by year group were adjusted using direct standardisation, with the age distribution from the 1996 Canadian Census taken as the standard reference population. All analyses were stratified by gender.

Educational inequalities were measured in multiple ways. The first was the simple prevalence ratio, which is the ratio of the prevalence in each educational group relative to the prevalence in the highest educational group. However, given the changing distribution of education in Canada over the time period examined we also incorporated inequality measures that include adjustment for the changing size of each educational group. These were the relative concentration index and the absolute concentration index (see Refs 22 and 23 for more details on these measures; formulas are provided in Appendix B (available online)). Both the relative and absolute concentration indexes measure the extent to which particular behaviours are concentrated among particular social groups. In the case of this paper, when the health behaviour is undesirable (eg, smoking), a negative concentration index indicates that the behaviour occurs more often among lower socioeconomic groups. Similarly, when the health behaviour is desirable (eg, being physically active) a positive concentration index indicates that the behaviour is less common among lower socioeconomic groups.22 23 The relative concentration index (RCI) has a mathematical relationship with another measure of inequality—the relative index of inequality (RII).22 We have included results using the RII in Appendix C (available online) for comparison. For all surveys, with the exception of the 1974 SHC survey, estimates were weighted to account for the probability of selection and non-response. In addition variance around each prevalence estimate for the 1996 NPHS and the 2000 and 2005 CCHS’s have been adjusted for the complex sampling design for these surveys using 500 bootstrap replicate weights in line with guidelines from Statistics Canada.24 No details on the cluster design of surveys previous to 1996 were available; hence the variances around these estimates have not been adjusted. Variance estimates around relative concentration and absolute concentration indices were estimated by re-estimating each measure 1000 times, as described elsewhere.23 25

RESULTS

Table 1 presents the age-adjusted rates of smoking by educational groups from 1974 to 2005 for both men and women. Although there were reductions in both the absolute and relative percentage of respondents who smoked more than 10 cigarettes a day across all educational groups, these reductions differed to a large extent among both men and women, with absolute reductions among men being larger than reductions among women (14–21% among men vs 5–16% among women). The proportions of each educational group reporting never smoking decreased among women, but remained relatively steady (with the exception of the lowest educational group) among men. The percentage of former smokers (as a percentage of ever smokers) increased over time across all educational groups with the greatest absolute increases taking place in higher educational groups in particular among women (41.8% reduction).

Table 1 Age-adjusted percentage and 95% CIs for heavy smoking, never smoker and former smokers by education

Table 2 presents the trends in activity levels between 1981 and 2005 by educational groups. There were moderate reductions in inactivity observed between 1981 and 2005 in all educational groups, with these absolute reductions being larger among women than men in general (3–8% among men vs 8–11% among women). Between 1996 and 2005 these decreases were much larger across higher educational groups than the lowest educational group across both men and women. The percentage of respondents classified as active increased between 1981 and 2005, with these increases relatively similar across educational groups and among men and women.

Table 2 Age adjusted percentage and 95% CIs for respondents in different levels of physical activity by education

Tables 3 and 4 present trends in relative and absolute measures of inequality for both smoking and physical activity respectively. Socioeconomic inequalities in heavy smoking, as measured by both relative and absolute indices, increased between 1974 and 2005 for both men and women (relative concentration index (RCI) −7.93 to −26.90 among men; RCI −4.81 to −27.44 among women). Absolute inequalities in never smoking and former smoking also widened, with relative inequalities in never smoking also increasing (RCI 8.01 to 12.30 among men; RCI −2.84 to 7.87 among women). These trends were more pronounced in women than in men, due to lower inequalities among women in 1974.

Table 3 Relative and absolute risk changes in educational inequalities in smoking and physical activity
Table 4 Relative and absolute risk changes in educational inequalities in physical activity

Trends in physical activity (table 4) were more complicated. In general, inequalities in activity and inactivity narrowed between 1981 and 1996 as measured by both absolute and relative indices. However, between 1996 and 2005 it appears that inequalities widened, as measured by both relative and absolute indices for both men and women (RCI for inactivity −4.34 to −6.75 among men; −3.57 to −5.54 among women).

DISCUSSION

The objective of this paper was to examine trends in socioeconomic inequalities in smoking and physical activity using Canadian population-based surveys from 1974 through to 2005. We found differences in both smoking and physical activity across educational groups in all surveys examined between 1974 and 2005, with lower educational groups more likely to be heavy smokers and inactive in each survey. In general, these results were consistent for both men and women. Both relative and absolute educational inequalities in smoking increased between 1974 and 2005, with inequalities in physical inactivity narrowing between 1981 and 1996, then expanding between 1996 and 2005.

However, the results of this paper should be interpreted in light of the following limitations. There have been changes in the administration in the surveys used in this paper over time. While the questions used to assess levels of smoking and physical activity have remained relatively consistent, we have limited information on survey sampling design (other than they were multi-staged and clustered) in the surveys previous to the 1996 NPHS. Trends in physical activity should be interpreted given the limitations in the self-reported physical activity measure used in each of these population-based surveys. The measure used in each survey is based on only subjective recall of leisure-time activity over a specified time period, and therefore may not cover other types of health enhancing activity as equally or completely (eg, riding a bicycle to work).26 27 There were also small changes in the types or names of activities included in self-reported physical activity estimates. However, these changes were in activities that have a relatively low contribution to total population energy expenditure (eg, in-line skating, fishing, snowboarding) and may have been captured in the option in each survey to list other activities. Limitations aside, our study utilises large, representative Canadian samples to provide a profile of changes in educational inequalities in health behaviours over a long time period. In addition, our measures of absolute and relative inequalities in these behaviours have been automatically adjusted for the changing distribution of education in Canada.

Similar to a previous Danish study we have found that reductions over time in heavy smoking were greater among more highly educated respondents.28 This finding has also been replicated in other European countries, with the exception of Turin in Italy, over similar time periods.29 30 We also found educational inequalities in physical activity decreased between 1981 and 1996, but then reversed between 1996 and 2005.

Of particular interest is the effect that the increasing educational inequalities in smoking and physical activity documented in this paper may have on future inequalities in cardiovascular disease in Canada. Little research has examined trends in socioeconomic inequalities in coronary heart disease in Canada. The research done has documented a slight reduction in the relative level of inequalities in coronary heart disease mortality in Canada across quintiles of neighbourhood income (an ecological rather than household or individual measure of socioeconomic position) between 1971 and 1996, with the sharpest reductions occurring before 1986. This is in contrast to studies from other European countries and the USA which have reported widening relative socioeconomic inequalities in morbidity31 32 and mortality3335 from coronary heart disease between the early 1970s to the late 1990s. More comprehensive surveillance of socioeconomic inequalities in coronary heart disease—both from incidence and mortality—using multiple measures of socioeconomic position (eg, income, education and occupation) are required to better understand the distribution of, and trends in, health inequalities in Canada.

It is unlikely that the widening educational inequalities in smoking and physical activity documented here reflect lower knowledge of the consequences of smoking and physical inactivity among lower educated groups. Data from the 1996 NPHS demonstrates that only 4% of Canadians aged 12 and over believe there are no health risks to smoking, and although this belief is lower among higher educational groups, still only 6% of respondents aged 12 and over with less than high school education believe smoking provides no health risks.1 Rather, the trends in this article highlight some of the challenges faced by health promotion strategies at the population level. As suggested by others, population health strategies must be designed with a focus on equity, as individualised strategies and messages that do not recognise the broader social determinants of behaviour change will ultimately, albeit unintentionally, result in greater inequalities in health behaviours between socioeconomic groups.11 13 For example, a recent study using longitudinal Canadian data suggests that self-reported control at work and level of neighbourhood stress are both important factors linking lower levels of education to lower levels of physical activity.36

The widening gaps in health behaviours across socioeconomic groups necessitate creative action from all levels of government. Previous research has suggested that restrictions on tobacco advertising, smoking in public places, subsidised smoking cessation and increased cigarette prices may be beneficial in reducing socioeconomic inequalities in smoking,37 although others have suggested that only prices are effective in reducing inequalities.38 Previous Canadian research suggests that increases in the prices of cigarettes were the only policy level intervention associated with higher rates of non-smoking and reduced smoking among both men and women.39

The period from 1994 through 2007 has seen increases in the prices of cigarettes in all Canadian provinces. Between 1996 and 2005 we observed a levelling (and in some cases a reduction) in inequalities associated with never smoking and smoking cessation among men and women (see table 3) – although inequalities in heavy smoking continued to increase over this time period. This may reflect the need for more comprehensive policy approaches to not only increase cigarette prices, but to effectively restrict the sale of contraband cigarettes. For example in 2005–6, 14% of all cigarettes smoked by current smokers in Ontario (Canada’s largest province) were bought tax-free on First Nations reserves.40

Complete restrictions on smoking in public areas were only enacted after 2005 (our last year of available data) in all Canadian provinces, with subsidised smoking cessation programmes only recently being introduced in some Canadian provinces. Future research will have to examine the effects that these policies have on smoking inequalities.

Given the continued increases in smoking inequalities, other policy options worth considering might include consideration of economic incentives such as the provision of insurance benefits for smoking cessation with co-payments graduated by income level. Provincial health insurance plans could also establish ‘bonus payments’ graduated by income level, to individuals who have sustained tobacco abstinence. Less research has examined policy interventions for physical activity. Options for supporting regular physical activity using economic incentives or by investing in the physical and social environments in neighbourhoods (eg, the building a parks and recreational spaces) which respond to the circumstances of different socioeconomic groups would also be appropriate to evaluate. The results presented here suggest more work needs to be done to create supportive environments that provide equal opportunities for behaviour change for all educational groups in Canada.

What is already known on this subject

  • People with lower levels of education are more likely to smoke and be physically inactive.

  • Most countries have documented widening of socioeconomic inequalities in levels of smoking.

  • No Canadian research has examined trends in socioeconomic inequalities in either smoking of physical activity.

What this study adds

  • Both relative and absolute educational inequalities in smoking increased between 1974 and 2005 in Canada.

  • Inequalities in physical activity were more varied with a narrowing between 1981 and 1996, followed by a widening between 1996 and 2005.

  • More work needs to be done in both designing population health approaches that focus on equity and the creation of supportive environments that provide equal opportunities for behaviour change for all educational groups in Canada.

Acknowledgments

While completing this work John Frank was the Scientific Director of the Institute for Population and Public Health at the Canadian Institutes of Health Research.

REFERENCES

Supplementary materials

Footnotes

  • Competing interests: None.

  • Funding: Partial funding for this work was provided by the Canadian Heart Health Strategy. Data used in this paper were provided through Statistics Canada’s Research Data Centers and the Data Liberation Initiative at the University of Toronto, Canada. PS is supported by a New Investigator Award with the Canadian Institutes of Health Research.

  • Ethics approval: Approval for the secondary data analyses was obtained through the University of Toronto, Health Sciences I Ethics committee.

  • ▸ Additional appendices are published online only at http://jech.bmj.com/content/vol63/issue4