There are substantial socioeconomic gradients in rates of both current smoking and smoking cessation among adults, with people in more disadvantaged groups both more likely to be current smokers, and have lower quit rates. When considered in multivariate models, independent associations were found between a range of markers of disadvantage and smoking status. In terms of current smoking, our results on socioeconomic indicators are consistent with prior research, although psychosocial variables such as mental health status typically have not been included in this research previously [35–38].
Our findings do not support our original null hypothesis that there would be no independent association between mental health status and smoking after adjusting for a comprehensive set of markers of socioeconomic disadvantage. Indeed mental illness was the most prevalent risk factor with a strong independent association with current smoking and reduced likelihood of smoking cessation. This supports previous findings indicating the high burden of smoking associated with mental disorders such as depression and anxiety [4–6], which may be the most prevalent risk factor strongly associated with smoking behaviours. These results indicate that previous findings were not merely presenting a confounded relationship with socioeconomic status. While these data do not address the question of whether this association is causal and if so, in what direction, they do raise the question as to whether efforts to control tobacco use and to promote smoking cessation in this group would be more effective if the impacts of mental illness were explicitly incorporated into their design.
Similarly the reverse hypothesis, that there would be no independent association between socioeconomic status and smoking after adjusting for mental health status is not supported by these data. Even in those with no lifetime mental illness, gradients in both current smoking status and smoking cessation can be observed for several indicators of socioeconomic status. Living in rented accommodation remaining strongly associated with both current smoking and smoking cessation in the multivariate models, and being unemployed or ever having been in gaol, prison or a correctional facility strongly associated with current smoking.
There are several possible mechanisms that may explain an independent association between smoking and common mental disorders . Mental illness may be a factor in smoking initiation or addiction to nicotine. Smoking may be a risk factor for onset of mental illness. Alternatively both smoking and mental illness could be linked to common genetic, biological or environmental factors. There is some evidence to support each of these possible mechanisms . Depression and anxiety in teenagers have been found to be strong predictors of smoking experimentation and the transition to daily smoking [40–42]. Smoking has also been associated with the onset of psychiatric symptoms in teenagers [42, 43]. While it is possible that separate causal mechanisms may operate in both directions, other studies have identified common risk factors to both smoking and mental illness [44–47].
There are well-established biological mechanisms that help explain why smoking may be linked to mental illness. Nicotine is a psychostimulant that affects several neuroregulators that influence both mood and behaviour [48, 49]. Early reports suggested that nicotine cessation can also precipitate depressive symptoms, particularly in people with a history of major depression [50, 51]. Nicotine administration can relieve symptoms of both depression and anxiety [52–54]. Levels of cortisol, a component of the hypothalamic-pituitary-adrenal axis system that responds to stress, can be affected by nicotine [39, 55]. More recent studies have suggested that nicotine withdrawal symptoms can be quite similar to symptoms of anxiety and depression which may further reinforce smoking behaviours in people with these disorders [54, 56]. The knowledge that people with mental illness may perceive there to be some therapeutic benefit from nicotine led to the self-medication hypothesis, which suggests that people with mental illness may choose to smoke as it is a simple and readily accessible means to control symptoms of mental illness . However, levels of anxiety and depression in ex-smokers after the withdrawal period may be lower than immediately after smoking in current smokers suggesting that perceived benefit of smoking for people with depression and anxiety may merely reflect the similarity between nicotine withdrawal symptoms and symptoms of anxiety or depression [39, 58].
The finding that smoking rates are higher among disadvantaged groups is not new. However, to the best of our knowledge, the finding that strong gradients in smoking, and smoking cessation, are observed by mental illness status within both disadvantaged and non-disadvantaged groups has not previously been reported. We also found that smoking cessation rates are lower in socioeconomically disadvantaged groups and in those with common mental disorders, and that significantly lower cessation rates are observed in disadvantaged groups after controlling for mental illness. This is a more controversial finding.
Other reports have suggested that while smoking initiation rates may be higher in disadvantaged groups, smoking cessation rates are equal in all groups, although these reports tend to rely on coarse measures, and flawed interpretations of broad population measures. Two measures commonly cited to suggest that quit rates are equal across demographic groups are the proportion of population groups who are ex-smokers [19
], and the absolute percentage point reduction in smoking rates in various groups [19
]. For example, Australia’s National Preventative Health Taskforce reported that:
“Most disparities in smoking rates between socioeconomic groups in Australia result from differences in uptake rather than in cessation…around 30% of people can be classified as ex-smokers, regardless of the level of neighbourhood disadvantage.” [19, p. 12]
However, the proportion of ex-smokers is not a reliable estimate of cessation rate as the denominator is all persons, not smokers. Consider the following hypothetical example. Suppose two groups of 100 smokers are followed for a period of time, and 20 people quit smoking in each group. However group A comes from a relatively disadvantaged group with a total population of 200 people while the less disadvantaged group B has a total population of 400 people. In group A 20 of 200 people, or 10% are ex-smokers, while only 5% of group B are ex-smokers, despite both groups having the same smoking cessation rate. Indeed, if only 10 people had quit smoking in group A, both groups would have shown the same proportion of ex-smokers despite the smoking cessation rate being half that in group A compared with group B. As smoking rates are substantially higher in areas of high neighbourhood disadvantage, the data presented by Australia’s National Preventative Health Taskforce are consistent with our findings, as equivalent proportions of ex-smokers imply lower smoking cessation in groups with higher smoking initiation rates. However, these data have not previously been interpreted in this way. In a similar vein, the National Preventative Health Taskforce also reported that smoking rates had declined between 1989–90 and 2004–5 from 33% to 29% in areas in the highest quintile of socioeconomic disadvantage, and from 23% to 17% in areas in the lowest quintile of socioeconomic disadvantage . The argument that graphs of time trends in smoking rates in different groups are represented by approximately parallel trend lines has been advanced to suggest that cessation rates are equal in different groups [19, 59, 62]. However, again this does not account for the differences in smoking prevalence in the groups. If cessation rates were equal in different groups, trend lines in smoking prevalence, if linear, would be expected to converge to the same point, not be parallel. Considering again our hypothetical example, with the same smoking cessation rate in each group the smoking rate declines by 10 percentage points, from 50% to 40% in group A, and by 5 percentage points, from 25% to 20% in group B. If group A had only half the smoking cessation rate of group B, the decline in smoking rate would have been 5 percentage points from 50% to 45%.
The results of our study support the general principle of developing tobacco control strategies that specifically address the needs of disadvantaged groups with high smoking rates. There were substantial gradients in smoking rates observed across many of the indicators included in this study. These reflect not only the high relative burden of smoking associated with various forms of disadvantage, but the extent to which broad-based tobacco control efforts have had their greatest success among those who have fewer additional disadvantages in their lives. The strongly skewed sociodemographic of current smokers now has emerged from a much more homogeneous population of smokers thirty or forty years ago, when smoking was actually more common among the more affluent and well-educated [63–65].
For affluent adults with no history of mental illness, current smoking rates are now below 10% while rates are over 40% for those with 12-month mental illness and one of several indicators of socioeconomic disadvantage.
There are a number of programmes that have been developed that address smoking within mental health treatment settings, such as the Tobacco and Mental Illness project in South Australia , and Mental Health Tobacco Recovery in New Jersey . While these programmes include components to assist with the transition to community-based living, including the use of peer support workers , they focus on people with serious mental illness and generally recruit from specialist psychiatric services.
A number of strategies have been proposed as to how to address the high rate of smoking among disadvantaged groups and people with common mental illnesses [20, 57, 69]. Broadly these can be considered as either programmes that work with individuals or in small groups to support smoking cessation, and approaches that modify broad population health interventions to more specifically target disadvantaged groups. An intermediary approach is exemplified by the Tackling Tobacco programme initiated by the New South Wales Cancer Council. This programme is based on the likelihood that many people from disadvantaged backgrounds are in contact with health or social services for other reasons as a result of that disadvantage. This may then be a way of targeting this group. The programme seeks to integrate tobacco control into the services provided by a range of social and community sector organisations . As such it seeks to denormalise smoking in sectors working with disadvantaged people where smoking behaviours are commonplace, and to increase the impact of brief interventions and support for cessation beyond the health sector. So far this programme has been piloted and qualitatively reviewed [71, 72] and a randomised controlled trial is underway to investigate the effectiveness of this type of approach in achieving smoking cessation within disadvantaged groups . Preliminary data suggest that both the organisations and their clients are enthusiastic to have smoking cessation activities provided through these services [71, 72, 74].
Another possible approach to addressing smoking among disadvantaged groups is to adapt population health-based methods. The principal components of population health-based smoking cessation efforts, such as controlling supply, restricting all forms of promotion, increasing price, advertising health consequences and educating young people about them, denormalising or stigmatising smoking, and restricting use in public places, have generally not been tailored for people from specific demographic groups.
The design of population health interventions such as education and denormalisation can be considered within the framework of social marketing . Although targeted approaches, or market segmentation, are widely used in the marketing of commercial products and services, including historically in the promotion of cigarettes [76–78], this approach has not been strongly embraced in population health. Bloom and Novelli note that treating certain groups with special attention “is not consistent with the egalitarian and antidiscriminatory philosophies that pervade many social agencies” and that during program planning and implementation there is “a constant problem about whether to divide limited resources or simply take a general audience route” . This argument has also been made specifically in the area of tobacco control. Indeed, it has been suggested that money spent on targeted anti-smoking efforts is money wasted as it is money taken away from the most successful broad population-based approaches . Alternatively it has been argued that developing both targeted and broad population-based strategies would undermine both as no strategy would then be adequately funded to a level that could have impact . These arguments rest on the assumption that all population subgroups benefit equally from broad reach interventions, which is counter to the increasing burden of smoking concentrated in disadvantaged groups.
Of the various broad-based population-health intervention approaches to smoking cessation, most have greater impact in more advantaged groups . Only price increases through taxation have been suggested to be more effective in disadvantaged groups [82, 83]. However, the data on the effectiveness of taxation increases in reducing smoking in disadvantaged groups is equivocal, and others have questioned this finding [84, 85]. The question of whether any one aspect of tobacco control is more or less successful in disadvantaged groups misses the fact that the successes of tobacco control overall have come from the implementation of programs that employ multiple strategies.
While the preference for broad-based programmes may be based on egalitarian principles , in fact this focus on broad-reach interventions may promote inequality if these interventions are most successful amongst advantaged subgroups. Two common themes in health promotion in tobacco control are the long-term health consequences of smoking and establishing smoking as a stigmatised behaviour. The promotion of long-term health consequences may be less motivating of behaviour change in people whose ability to project in the longer term is limited by pressing life circumstances or stress or whose cognitive skills are impaired by psychological distress [86, 87]. Similarly the impact of the stigmatisation of smoking behaviours may be less motivating in people who are also stigmatised by other forms of disadvantage . Ceci and Papierno have argued that reliance on universal strategies will always result in the widening of gaps between advantaged and disadvantaged groups as the resources, skills and opportunities of advantaged groups act to increase their chances of utilising and benefitting from any universal strategy . Niederdeppe and colleagues point out that mass media campaigns, in particular, are often differentially effective in advantaged groups for multiple reasons: levels of exposure, levels of persuasiveness, opportunities to change, and access to supports all vary by level of disadvantage [90, 91].
Marsh and McKay noted that while price increases have played an important part in overall tobacco control, perversely some of the heaviest smokers are also among the poorest and pay a high financial price for their smoking. Yet their levels of addiction, financial literacy and overall decision making result in this strategy yielding lower results among the poorest smokers .
There is likely to be an important role for both types of approach in addressing the issue of smoking and disadvantage. Initiatives based around use of services can reach groups where broad-based strategies have little or no penetration, and these services are often the best way of reaching people with the most severe and multiple disadvantages. At the same time, not all disadvantaged smokers are in contact with services, and service-based programmes will only reach a proportion of the target audience. For example, the majority of people with mental disorders such as anxiety or depression are not in contact with services for these problems, although there is no difference in smoking or smoking cessation rates between those who do and do not use services [6, 21]. In responding to the association between markers of disadvantage and smoking, the strategies recommended by Australia’s National Preventative Health Taskforce have aimed to target the small proportion of people with very high levels of disadvantage, such as people who are currently homeless or in a correctional facility or long-term residents of psychiatric facilities . While there is no doubt that smoking rates are very high in these groups, our data show that the gradients in smoking status extend to a significantly larger proportion of the population. This has implications for tobacco control efforts. While it is possible to develop programs within institutional settings or that use street workers to target homeless people, targeted population-based strategies will also be required to reach the larger proportion of people living in the community who have more common mental disorders such as anxiety or depression, or who have a history of homelessness or contact with the justice system but aren’t currently homeless or in an institution. As these people are less likely to be in touch with services that can deliver programmes directly, strategies with population reach but which are targeted to the concerns and issues faced by people with other disadvantages will also be needed to address the substantially disproportionate burden of smoking that is associated with common mental disorders and socioeconomic disadvantage.
Socioeconomic and psychosocial gradients in smoking are a major contributor to socioeconomic gradients in major health outcomes such as life expectancy and quality of life. The majority of responses to smoking in disadvantaged groups to date have been individually-oriented treatments of tobacco dependence . The factors considered in this paper, common mental disorders and markers of socioeconomic disadvantage, are sufficiently prevalent to suggest policy making and population-based approaches as being a key part of the way forward. This could entail using the tools of population health but adapted to the specific characteristics of these groups, as many of these population groups are too big to reach effectively through individual treatment services.
Monitoring progress in future tobacco control activities, particularly those directed at disadvantaged groups, may require new measurement approaches. For instance, in Australia key indicators are derived from the National Drug Strategy Household Survey . This is a survey conducted by means of a self-complete questionnaire delivered to selected households. The 2010 survey obtained usable responses from around one-third of households selected in the original sample, and under-represented young adults, people who didn’t complete year 12 schooling, single person households, and households from low socioeconomic areas. It is quite likely that differences in methodology, and the low response rate and associated participation biases explain why the estimated proportion of current daily smokers derived from the National Drug Strategy Household Survey (15.1% in the most recent survey in 2010, 16.6% in 2007) is substantially lower than the estimate obtained from face to face surveys with higher response rates . The 2007–08 National Health Survey, conducted by the Australian Bureau of Statistics, estimated 18.9% of Australian adults are current daily smokers . This survey achieved a 91% response rate. Some people with severe forms of disadvantage, such as homeless people, people living in institutions such as correctional facilities or mental health facilities who are known to have very high rates of smoking, are not included in any household surveys.
Not all markers of socioeconomic disadvantage that have been associated with smoking status have been included in this analysis. Because of sample size considerations and in order to preserve the privacy of individual participants in the study, particularly low prevalence demographic indicators were not included on the unit record file released from the NSMHWB. Because of this we were unable to identify in this sample Aboriginal or Torres Strait Islander peoples, pregnant women or people who do not speak English. These groups represent a very small proportion of the NSMHWB sample. Additionally, to preserve the privacy of individual participants in the study, some continuous demographic measures, such as household income and area-level disadvantage, have been categorised on the unit record file released for the survey. Because of this, we were unable to assess whether there could be linear or non-linear associations between these measures of disadvantage and smoking outcomes, or whether the categorical cut-offs provided on the file are optimal for defining disadvantaged groups in respect to smoking behaviours.
As a population-based household interview survey, the NSMHWB was unable to assess low prevalence mental disorders such as schizophrenia or organic psychoses, nor did it include people living in institutional care. Smoking status in Australian adults with psychotic illness has recently been assessed in the second Australian national survey of people living with psychotic illness conducted in 2010 . This survey found that over two-thirds of adults with psychoses were current smokers, unchanged from the first national survey ten years previous [96, 97].
Cross-sectional studies, such as the NSMHWB, describe associations but cannot shed light upon causal pathways. The information collected in the survey pertains to current disadvantage or disadvantage in the past 12 months. We don’t have information on the long term accumulation of disadvantage or the intergenerational transfer of disadvantage.