Article Text
PDF
Abstract
Background During the 1990s, inequalities in smoking prevalence by socioeconomic status (SES) have widened in Europe. Since then, many tobacco control policies have been implemented. Yet, European overviews of recent trends in smoking inequalities are lacking. This paper aims to provide an overview of long-term trends of socioeconomic inequalities in smoking cessation in Europe.
Methods We used data for 11 countries taken from Eurobarometer surveys from 1987 to 1995 and 2002–2012, with a total study sample of 63 737 respondents. We performed multilevel logistic regression to model associations of the quit ratio (proportion former smokers of ever smokers) with SES, measured by education and occupation separately, with adjustments for age, sex and time.
Results We found a significant, positive association for education and occupation with the quit ratio. The strength of the association decreased slightly from 1987 to 1995 and increased again from 2002 to 2012. Inequalities increased between the two periods in most countries and decreased in only one country. While in 1987–1995, the quit ratio increased among all SES groups and most strongly among the low SES group, in 2002–2012 it increased only among the high-education group (OR=1.38, 95% CI 1.02 to 1.87), and non-manual occupation group (OR=1.59, 95% CI 1.19 to 2.12).
Conclusions Socioeconomic inequalities in smoking cessation rates have strongly increased since the 1990s and during the 2000s. This suggests that the tobacco control policies implemented during the 2000s have not been able to counter the trend in increasing inequalities.
- SMOKING
- SOCIAL INEQUALITIES
- TIME-SERIES
- SOCIAL CLASS
Statistics from Altmetric.com
Introduction
In the European Union (EU), the average prevalence of daily smoking around 2006 was 25.1%, ranging from 45.2% in Bulgaria to 19.7% in Portugal.1 In the EU, an estimated 700 000 people die prematurely from smoking every year and an additional 13 million people each year suffer from smoking-related diseases.2 Furthermore, smoking in Europe is not equally distributed among low and high socioeconomic status (SES) groups; the odds of smoking for low SES are around 1.6 when compared with high SES.3
Recent reviews of tobacco control policies have shown that although many have been successful in the overall population, most were more effective among high-SES groups, except the raising of tobacco taxes.4 ,5 Therefore, while these policies might increase smoking cessation rates among all SES groups, most of them could exacerbate existing inequalities between these groups.
Most international studies that investigated SES inequalities in smoking in Europe are cross-sectional;3 ,6–8 only one study assessed trends in these inequalities.9 Giskes et al9 compared repeated national surveys from nine European countries, spanning the period from 1985 to 2000, and concluded that inequalities increased during the 1990s in most Western European countries. These data are by now quite outdated, especially since the 2000s saw the introduction of new tobacco control policies10 along with broader societal changes.
This begs the question whether trends in inequalities in smoking have changed. A small number of recent trend studies, only within single countries, mostly reported increasing inequalities.11–13 However, comparing these results is often difficult due to the different methods used. Additionally, these studies tend to focus on smoking prevalence, which is influenced by initiation and cessation, making it hard to untangle the effects of processes related to youth and adulthood, respectively.
This lack of data can be addressed by using the Eurobarometer surveys. These surveys have been held at irregular intervals since the 1980s and provide comparable smoking estimates across the EU. There were five surveys between 2002 and 2012, enough to provide data on recent trends in smoking cessation rates. The earlier surveys also allow us to make a direct comparison from 1987 to 1995.
This paper focuses on smoking cessation, so as to focus on changes occurring among adult populations rather than the youth. It aims to provide an overview of trends in smoking cessation rates in Europe over a quarter of a century, with a focus on socioeconomic inequalities. This includes comparisons of periods (roughly the 1990s and the 2000s) and comparisons of countries. As there was a long-term pattern from 1985 to 2000, we expect to see this pattern of increasing inequalities to continue up to 2012.
Our specific goals:
To study associations between indicators of SES and smoking cessation in different time periods, and how these change over time—in other words, to assess trends in inequalities.
To study trends in smoking cessation rates and how these patterns differ by SES. An increase in inequalities can be caused by one group improving and the other getting worse, or both groups improving, but one more than the other. Therefore, we will also assess inequalities in trends.
To compare these results between countries so as to identify possible variations from the general pattern.
Methods
Population
We used data of 11 EU member states from 12 Eurobarometer surveys taken at irregular intervals from 1987 to 2012.14 The Eurobarometer surveys are conducted on behalf of the European Commission since 1973. They aim to monitor public opinions and a wide range of topics including health in the member states, with the aim to inform decision-making. Each wave of the survey consists of around 1000 respondents per country, except for 1989 when the sample per country was 2000. For Austria, Finland and Sweden, no data were available before 1995.
The Eurobarometer samples were independently selected at random for each new survey, through multistage probability sampling. For the sampling, the population distribution across urban and rural areas, from each administrative region, were taken into account to ensure national representativeness of the samples.
The initial study sample contained 139 279 respondents. We excluded 63 318 never smokers, along with 3704 respondents under 20 years of age and respondents with missing data on age (13), sex (12), education (1186) or smoking status (7309). This resulted in a final study sample of 63 737 respondents.
Variables
Our main outcome variable, the quit ratio, was composed of current and former smokers. Never smokers were excluded from the analyses, because they were not ‘at risk’ for smoking cessation. The quit ratio measures former smokers as a proportion of ever smokers (current smokers plus former smokers).
We used two measures of SES: education and occupation. Education was measured as the age when finished or left full-time education (low≤15 years, middle=16–19 years, high≥20 years). As we had excluded respondents aged under 20 years, all those who were still studying would be 20 years or older when they finish their studies. Therefore, they were included in the high-education group. Occupational class was combined into four groups: manual (farmers, fishermen, manual workers), non-manual (office employees, white collar workers, professionals, management), self-employed (business owners, craftsmen) and other/not gainfully employed (students, military service, housewives, unemployed, retired). The self-employed group was very small; therefore, the results for this group are not shown even though the group was included in the model. Individual-level covariates that were included were sex (male/female), and age (continuous when adjusting for confounding, and in four categories when stratifying, 20–34, 35–49, 50–64, 65 years or older).
Statistical analysis
To study trends in inequalities, we used multilevel logistic regression models15 to investigate associations for education and occupation separately, with the quit ratio as the outcome. Survey year and country were used as the higher levels in the three-level structure with a random intercept. In addition, we adjusted for age, sex and time (in years) as fixed effects. These models were also stratified by year, using only country as a higher level, and stratified by country, using only year as a higher level. There was a large gap between surveys from 1995 to 2002. Therefore, we have chosen to analyse the periods 1987–1995 and 2002–2012 separately.
To answer the second goal, on inequalities in trends, we used the same multilevel structure with the quit ratio as the outcome, but with time (in 10 years) as the determinant to model absolute trends over time. Interaction variables with time were tested for education, occupation, sex and age in separate models, adjusting for the other variables. Subsequently we modelled the association between time and quit ratio stratified by education, occupation, sex and age. The results stratified by education and occupation for the per study period are also shown per country. All statistical analyses were performed using R (V.2.13.1).
Results
Table 1 shows the characteristics of the sample per year; the mean age was 46; 44.8% were women. The overall quit ratio was 0.428, indicating that about 43% of ever smokers in the EU had quit smoking; this increased from 0.358 in 1987 to 0.506 in 2012. The educational groups were on average about the same size, but the low-education group decreased in size over the years while the high-education group increased. For occupation, there were 17.6% in the manual group, 31.3% in the non-manual group, only 6.1% in the self-employed group and 45.1% in the other or not employed group. Descriptive information on the study population is also available per country in the online supplementary table S1.
Descriptive information of the study population, per year
In table 2, the associations between indicators of SES and the quit ratio are shown. There is a significant positive association of education and occupation with smoking cessation rates, in both time periods. The association with the quit ratio is significant for middle compared with low education in 1987–1995 (OR 1.16, 95% CI 1.10 to 1.23) and in 2002–2012 (OR 1.13, 95% CI 1.06 to 1.21). It is also significant for high compared with low education in 1987–1995 (OR 1.44, 95% CI 1.35 to 1.54) and in 2002–2012 (OR 1.77, 95% CI 1.64 to 1.91). The odds of quitting smoking were higher among those with a non-manual compared with manual occupation in 1987–1995 (OR 1.38, 95% CI 1.29 to 1.48) and in 2002–2012 (OR 1.54, 95% CI 1.42 to 1.67). Thus, the difference between high and low education was significantly larger in 2002–2012 as compared with 1987–1995. For occupation, a similar trend was observed, but the CIs overlapped.
Associations between education, occupation and covariates with the quit ratio
The associations of education and occupation with the quit ratio per year are shown in table 3. The gap between low and high-education groups slightly decreased from a high point in 1987 (OR 1.94, 95% CI 1.55 to 2.43) towards a low point starting around 1995 (OR 1.51, 95% CI 1.29 to 1.77), and increased again from 2002 (OR 1.44, 95% CI 1.22 to 1.70) to 2012 (OR 2.01, 95% CI 1.69 to 2.38). The gap between low and middle education was larger in 1987 (OR 1.46, 95% CI 1.21 to 1.77) than in 2012 (OR 1.18, 95% CI 1.02 to 1.38), but there was no clear pattern of change. The pattern of inequality between manual and non-manual occupation slightly increased over time from 1987 (OR 1.57, 95%CI 1.24 to 2.00) to 2012 (OR 1.73, 95% CI 1.44 to 2.09). For the other occupation groups, there were some significant positive associations, but no clear changes over time.
Associations between education and occupation with the quit ratio per year
The associations between SES and the quit ratio are displayed per country in table 4. In 2002–2012, significant inequalities favouring the higher SES groups, by education and occupation, were present in nine and seven countries, respectively. Educational inequalities increased between 1987–1995 and 2002–2012 in most countries, most notably in Ireland and the UK; they decreased only in the Netherlands. Occupational inequalities also increased in several countries and remained stable in the others, except for the Netherlands, where occupational inequalities were no longer significant in 2002–2012.
Associations between education and occupation with the quit ratio, per country
Table 5 shows absolute trends in the quit ratio per SES group and per period. The OR of 1.42 for the total population in 1987–1995 can be interpreted as a 42% increase in the quit ratio over a period of 10 years. In 1987–1995, the quit ratio increased significantly among all SES groups, but more strongly among the low-education group (OR 1.51, 95% CI 1.23 to 1.84) than the high-education group (OR 1.29, 95% CI 1.03 to 1.63). Conversely, in 2002–2012, there was a significant increase in the quit ratio among the high-education group (OR 1.38, 95% CI 1.02 to 1.87), but not for the low-education and middle education group. In 2002–2012, there was also a significant increase in the quit ratio among the non-manual occupation group (OR 1.59, 95% CI 1.19 to 2.12), but not among the other occupation groups. The tests for interaction of time with education (p=0.195) and occupation (0.781) show that differences in trends between SES groups were not significant for 1987–1995. However, SES differences in trends were significant for 2002–2012, for education (p=0.005) and occupation (p=0.008).
Trends in the quit ratio per period, per sociodemographic group
Differences in the quit ratio between countries, for both periods, are shown in the online supplementary table S2. In 1987–1995, quit ratios in most countries increased most strongly among the low-educated and manual occupation groups. This was significant for measures in the Netherlands and Spain, and for one measure in Ireland and Portugal. Conversely, in 2002–2012, quit ratios increased most strongly among the high-educated and non-manual occupation groups, except in Ireland and Portugal. The increase in quit ratios among the non-manual occupation groups was largest in Finland, Germany, Sweden and the UK, while in the Netherlands, quit ratios decreased significantly among the low educated.
Discussion
Our results show that from 2000 to 2012 there have been large inequalities in smoking cessation rates in the EU, by education and occupation. During the period 1987–1995, inequalities were relatively small and in many countries cessation rates were higher among the low SES groups. In the period 2002–2012, however, cessation rates did not significantly increase among most low SES groups, while they did among most high SES groups, leading to an increase in inequalities.
Our results suggest that the widening of inequalities, observed up to around 2000 in earlier cross-national studies,9 has continued into the 2000s. This was also seen in the USA, where between 1984 and 2004, despite dramatic cigarette price increases following the Master Settlement Agreement in 1998, inequalities in smoking increased.16 However, where we found decreasing inequalities between 1987 and 1995, a similar European study found increasing inequalities from 1985 up to 2000.9 Differences in study period, selection of countries, and data sources may perhaps explain this discrepancy. Moreover, that study looked at smoking prevalence, whereas we studied cessation. Elsewhere, increasing inequalities in prevalence were shown to be driven by increasing inequalities in initiation rather than cessation.17
Limitations
Doubts have been raised on the validity of smoking prevalence estimates obtained from the Eurobarometer surveys, when compared with those from national surveys.1 For Austria, Italy, Sweden and the UK, discrepancies over five percentage points were found, but the mean EU smoking prevalence from both data sources was less than one percentage point apart.1 We compared the 2012 Eurobarometer results with two recent studies.6 ,8 The smoking prevalence for most countries was similar across surveys: less than one percentage point apart for men and around three for women. As an additional test for the consistency between different data sources, we calculated the correlation, across countries, between prevalence rates as estimated from the Eurobarometer and those estimated from the two other sources. We found correlations of 0.621 and 0.655, respectively, both slightly higher among men. There were some larger discrepancies (around 7 percentage points) for Ireland, the Netherlands, Portugal and Spain. Although the estimates for the latter countries should be interpreted with caution, in general, the overall smoking prevalence rates as obtained from the Eurobarometer are comparable to those from the national sources.
As the Eurobarometer only questioned respondents on current/former/never smoking, we are unable to present a more sophisticated measure of smoking cessation than the quit ratio. Although it is an often used measure,13 ,18 it does have some limitations. Most notably, it provides no information on when the respondent quit smoking. Therefore, we cannot determine changes in cessation rates within a given year. However, it does provide us with a general estimate of smoking cessation across the population and over time.
The measure of education used in the Eurobarometer is the age, in years, at which the respondent completed or left full-time education. This measure has the benefit of being comparable over countries with different educational systems. However, it is not directly comparable to educational level or the years of schooling as dropping out of full-time education would result in a lower educational attainment than completing it. By using a proxy measure of educational level, we would have misclassified some people in comparison to their completed education. This might have led to an underestimation of the magnitude of educational inequalities. Similar problems of misclassification might occur with occupational status, although it is hard to predict how this might have influenced trends.
Interpretation of results
The between-country differences in the trends in inequalities can perhaps be partly explained by the tobacco epidemic model.19 This model explains the smoking prevalence shifting across populations in four phases. Inequalities in the Southern European countries during the 1990s were insignificant or at least smaller than in Western and Northern Europe. During the 2000s, the inequalities in the Southern European countries increased, except in Portugal, but were still less pronounced than inequalities in Western and Northern Europe. In agreement with the tobacco epidemic model, studies from Southern Europe found inequalities to be small or even absent.3 ,9 ,20 However, this alone cannot explain why inequalities increased more during the 2000s than during the 1990s. For instance, the increase in neoliberal globalisation policies has been known to contribute to rising health inequalities worldwide.21 An additional explanation might be the hardening hypothesis, which states that a growing proportion of current smokers, mostly of low SES, are unwilling or unable to quit. However, the evidence to support this theory is mixed.22 ,23
The emergence of new tobacco control policies during the 2000s could also be related to the increasing socioeconomic inequalities in this period, but this is hard to determine. Recent review studies have shown that different types of tobacco control policies can have a different effect on inequalities; this is also known as the equity impact.24–26 No consistent evidence was found on the equity impact of mass–media campaigns; the equity impact of advertising bans was mostly neutral. Population-level cessation support policies mostly had a positive equity impact, that is, resulting in decreased inequalities.24 However, all positive results are from the UK, where these services are specifically targeted at low SES groups. Their higher reach among low SES groups compensates for the lower success in this group and thus, achieves a positive equity impact.24 Individual-level cessation support tends to have a negative equity impact, that is, they increase inequalities.26
The strongest evidence for negative equity impact was found for smoke-free policies—especially voluntary, regional or partial laws.24 These types of smoke-free policies have been implemented since the harms of second-hand smoking became clear.27 The first national, comprehensive smoke-free workplace policies were introduced in Europe in 2004,28 and followed by smoke-free bars and restaurants between 2008 and 2011.29 Around half of compulsory, national and comprehensive smoke-free policies had a neutral equity impact and a further third had a negative equity impact.24 The smoke-free workplace policies will probably have had more impact on our results, as the smoke-free bar and restaurant policies were only introduced near the end of our study period.
Contrary to these negative equity impacts, price increases (through taxes) are often mentioned as having the most positive equity impact.24 ,25 However, there is also some evidence that raising taxes might actually have a negative equity impact.16 ,24 Tax increases have been implemented throughout the entire study period, but most countries tend to increase taxes gradually; only in the late 2000s did some strong tax price increases take place. The gradual nature of the tax increases, combined with the industry's pricing strategies, have led to only a gradual rise in the real price of tobacco in most of the countries in our study.30 An additional effect of tax increases is that smokers, especially from low SES groups, often switch to cheaper brands,31 or cut down rather than quit.32 These strategies have a strong potential to negate the impact of tax increases.
Results for specific countries could shed some light on the general role of policies. For instance, in Ireland and the UK, inequalities in 2002–2012 were much larger than in 1987–1995, and also consistently larger than in other countries. Ireland and the UK had introduced stronger tobacco control policies than most other countries after 2003.33 It is possible that some of these policies might have contributed to the rise in inequalities, as we found that absolute quit rates increased in the high SES groups, but not in the low SES groups. In the Netherlands, the only country for which we find decreasing inequalities, tobacco control policies have been found to be similarly effective for low and high SES groups.34
Conclusions
Socioeconomic inequalities in smoking were present during the 2000s in all but one of the studied countries and they were larger than during the 1990s. During the 2000s, cessation rates among the low SES groups did not increase significantly, as did the rates during the 1990s. The 2000s have also seen the introduction of new tobacco control policies in Europe. Although these policies are generally credited with bringing down the overall smoking prevalence, they have not succeeded in decreasing the inequalities in smoking, possibly even increasing them. The finding that socioeconomic inequalities continued to increase during a period of strong improvements in tobacco control policy supports the plea,25 for future tobacco control policies, to be more specifically targeted at low SES groups.
What is already known on this subject
During the 1980s and 1990s, socioeconomic inequalities in smoking prevalence have widened across Europe. Since 2000, in Europe, many tobacco control policies have been implemented that could have influenced trends in inequalities in smoking prevalence or smoking cessation. There is a lack of evidence on trends in smoking inequalities in the 2000s.
What this study adds
There has been a rise in inequalities in smoking cessation in the 2000s that seems to be accelerating, as compared with the trends of inequalities in smoking prevalence from the 1980s to the 1990s. This suggests that recent tobacco control policies have failed to prevent the widening of inequalities and future policies need to be more explicitly focused on addressing inequalities.
Acknowledgments
The authors used data from Eurobarometer surveys 27, 29, 32, 34.1, 38.0, 41.0, 43.0, 58.2, 64.1, 66.2, 72.3 and 77.1 made available from the GESIS data archive, Cologne. They would like to thank Wim Busschers for advice on and assistance with statistical analyses. Furthermore, they also thank all participants of the SILNE project for feedback on an early draft of the paper. This paper is deliverable within the SILNE Project ‘Tackling socioeconomic inequalities in smoking: Learning from natural experiments by time trend analyses and cross-national comparisons’.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online table 1
- Data supplement 2 - Online table 2
Footnotes
Contributors JRB and AEK conceived and designed the study. JRB prepared, analysed and interpreted the data and led the writing; and is the guarantor. MCW, KS and AEK interpreted the data and provided critical revisions. All authors have read and approved the final version of this paper.
Funding This study is part of the project ‘Tackling socioeconomic inequalities in smoking (SILNE)’, which is funded by the European Commission, Directorate-General for Research and Innovation, under the FP7-Health-2011 program, with grant agreement number 278273.
Competing interests None.
Provenance and peer review Not commissioned; externally peer-reviewed.
Data sharing statement All data are previously published and freely available.