Smoking and mental illness: results from population surveys in Australia and the United States

Data sources

Measures

Weighted estimates and standard errors

Survey weights were applied to calculate estimates of totals and proportions. These weights have been calculated to adjust for potential non-response. For the NCS-R, the supplied weights had been normalised to sum to the sample size, whereas the weights supplied with the SMHWB sum to the population size. We re-weighted the NCS-R data by multiplying the normalised weights by a constant factor so that the weights summed to the population size. This adjustment to the weights has no effect on estimates or standard errors for proportions. Standard errors and confidence intervals for the NCS-R and the NHIS were calculated adjusting for the complex nature of the sample design using expansion in Taylor series [54]. Standard errors and confidence intervals for the SMHWB were calculated using the jack-knife method of replicate weighting [54].

Modelling the relationship between smoking and psychological distress

The NHIS assessed the level of non-specific psychological distress using the K6 scale, which generates a score between 0 and 24. We used logistic regression modelling to assess the relationship between psychological distress score, and probability of being a smoker. As logistic regression models the log of the odds ratio, we did not enter the K6 score as a linear variable in the model as that would imply an exponential relationship between K6 score and probability of smoking. With no theoretical basis to assume a particular shape for the relationship, we used the generalised additive models (GAM) framework to fit a non-parametric spline curve to describe the relationship [55]. Similarly we used GAMs when using regression to model the average number of cigarettes smoked per day by level of psychological distress. All of the statistical analyses were undertaken using SAS software, Version 9.1 [56].

Mental illness and smoking in Australian adults

From the 2007 SMHWB, the ABS estimated that 20% of the adult population had a mental illness in the 12 months prior to the survey (95% CI: 18.9%–21.1%). Of those adults with a mental illness, 1 156 600 were current smokers – a smoking prevalence of 36.2% (95% CI: 32.9%–39.6%), almost double the 18.8% smoking prevalence among adults with no mental illness (95% CI: 17.2%–20.4%). In total, 3 566 800 adults were current smokers in 2007, and people with mental illness represented 32.4% (95% CI: 29.5%–35.3%) of current smokers. In contrast, only 15.6% (95% CI: 29.5%–35.3%) of people who have never smoked had a mental illness in the 12 months prior to the survey.

Mental illness and smoking in non-institutionalised adults in the US

Basing mental illness diagnosis on ICD-10, and using as similar as possible definitions of disorders as was applied in the Australian study, we estimated that 19.7% of the US adult civilian non-institutionalised population had a mental illness in the 12 months prior to the survey, conducted in 2001–2003 (95% CI: 18.9%–20.6%). Among those adults with a mental illness, 40.1% were current smokers (95% CI: 37.6%–42.7%) which was almost double the 21.3% smoking prevalence in adults with no 12-month mental illness (95% CI: 20.1%–22.5%). People with mental illness represented 31.7% of current smokers (95% CI: 29.5%–33.8%), or 16.1 million people out of an estimated total 51.0 million adult smokers in the US. In contrast, only 15.1% of people who had never smoked had a mental illness in the 12 months prior to the survey (95% CI: 13.8%–16.4%).

Age, mental illness and smoking

In both the US and Australia, smokers aged 16–24 years, who are likely to have started smoking most recently, had the highest rates of mental illness (Figures 1 and 2). In Australia, 37.2% of male smokers aged 16–24 years (95% CI: 28.0%–46.4%) and 58.7% of female smokers in this age group (95% CI: 48.7%–68.5%) had a 12-month mental disorder.

Age, mental illness, comorbid substance use and smoking

Use of services

Of the 1 156 600 Australian smokers who had a 12-month mental disorder, as estimated in the SMHWB, 740 000 did not access any health services for their mental health problems in the 12 months prior to the survey (64.0%, 95% CI: 55.4%–72.0%). The smoking rate was not substantively different between those people with 12-month mental disorder who did or did not access services. Of people with 12-month mental disorder who did not use any services, 35.4% were current smokers (95% CI: 31.2%–39.6%), while 37.7% of those with 12-month mental disorder who did use one or more services in the past 12 months were current smokers (95% CI: 32.9%–42.5%). Thus people with 12-month mental disorder who did not have contact with services in the past 12 months represent 20.7% of all current smokers in Australia (95% CI: 18.2%–23.3%).

Non-specific psychological distress and smoking

Figure 3 shows the relationship between level of psychological distress (measured using the Kessler 6 scale) and smoking rates in the US adult population. Although males have a higher rate of current smoking than females, the shape of the association is the same for both sexes. For both sexes, both the linear and non-linear components of the fitted curves were statistically significant (p < 0.01) indicating both that the increase in smoking rate with increasing level of non-specific psychological distress is significantly greater than zero, and that the deviation of the curve from the linear fit for scores above 13 is statistically significant. On the Kessler 6 scale, scores above 13 are generally considered to represent serious psychological distress likely to require intervention from mental health services.

By age group, large increases in smoking rates with increasing level of non-specific psychological distress are seen for younger people, with the proportion of smokers increasing from 18% to 50% among people aged 18–44 years, and increasing from 17% to 42% for adults aged 45–64 years. Smaller changes were observed in adults aged 65 and over with the smoking rate peaking at 14% (Figure 4). For all three age groups both the linear and non-linear components of the regression curves were statistically significant (p < 0.01).

The average number of cigarettes smoked per day by current smokers increased with level of psychological distress, from 12 per day among smokers with no psychological distress, to 19 per day among smokers with serious psychological distress (Figure 5). Both the linear and non-linear components of the regression curve were statistically significant (p < 0.01) indicating that the increase in number of cigarettes smoked by level of psychological distress is unlikely to be attributable to chance alone.

Implications for tobacco control

The main focus in responding to the high prevalence of mental illness among current smokers has been on interventions aimed at mental health treatment settings, particularly hospital and clinic based settings [66, 67]. There is no doubt that the high rate of smoking among patients of mental health services contributes to the excess physical morbidity and mortality consistently observed in these patients, and it is likely that smoking complicates or exacerbates mental illness and its treatment. Efforts to reduce smoking in patients of mental health services are clearly important. However, service-based smoking cessation programmes will be ineffective for the large proportion of smokers with common mental disorders who are not in contact with these services.

Approaches to smoking cessation outside the clinical setting have largely ignored the issue of mental illness. For example, the 2000 US Surgeon General report on reducing tobacco use does not mention this group [68], while the recent Australian National Preventative Health Taskforce report on tobacco control only mentions mental illness outside of mental health services to suggest that people with common mental illnesses are as likely as anyone else to benefit from general population-based smoking cessation strategies [69]. The principal components of population health-based smoking cessation efforts, such as reducing availability, restricting all forms of promotion, increasing price, advertising health consequences and educating young people about them, and restricting use in public places, have generally not been tailored for people with mental illness, and may be less relevant or less effective for people dealing with mental illness. For instance, increases in price may be less successful in motivating quit attempts in people with mental illness. Some of the heaviest smokers have the least ability to pay for their tobacco use [70]. The principle of increasing price to reduce smoking rates is predicated on smokers being rational agents able to maximise utility of available funds. Mental disorders, particularly depression and anxiety disorders can affect people's decision making. It is possible that continuing to increase the price of cigarettes may have a negative effect on mentally ill smokers, penalising both them and their families [37]. Similarly, it has been suggested that bans on smoking in public places may increase the level of social isolation felt by people with mental illness who smoke [38]. Denormalising smoking may be less influential on people who already feel marginalised from mainstream society for other reasons. Without programmes that ensure people with mental illness have as much chance of successfully quitting smoking as anyone else, the success of these types of public policies may further the level of inequality between people with and without mental illness.

Similarly, health promotional messages have concentrated on communicating the longer term negative health consequences of smoking. While this knowledge has prompted some people to quit smoking and prevented others from starting smoking, these messages may have less impact on people suffering mental illness or dealing with stressful and difficult life circumstances. Mental illnesses such as depression or anxiety disorders are associated with increased risk-behaviour, possibly motivated by the lower value placed on a long life by people whose sense of fulfilment in life is compromised [38, 71, 72]. The ability to focus on longer-term goals and consequences can be affected by the pressures of more immediate and challenging circumstances. There is little evidence in the literature of attempts to develop smoking cessation messages to target issues that are more relevant to people with mental illness.

Population health-based smoking cessation promotional activities have tended to focus on messages for the whole population or segmenting the population by demographic characteristics, such as age and sex. It has been suggested that this focus may reflect the way surveillance data are collected [73], but it is in contrast to the way that cigarettes historically have been marketed in the US and Australia, and the way they continue to be marketed in countries with less restrictions on cigarette marketing and promotion. Analysis of tobacco industry marketing research and market segmentation studies suggests the tobacco industry, in addition to developing products and campaigns targeting youth or women, for example, also targeted market segments defined by psychosocial characteristics, such as personality type. Products and campaigns were developed to address concerns such as reducing anxiety, stress or nervousness, improving mood, increasing social confidence, reducing irritability, or increasing concentration [11, 73–76]. Academic studies investigating motivations for taking up and for continuing smoking have also commonly found that psychological factors are important. Coping with stress, and controlling emotions are two factors commonly extracted from these types of studies [77–79]. Smokers are also more likely to deal with problems themselves rather than seek help [77]. As these psychosocial factors cut across demographic groups, demographically targeted smoking cessation campaigns may miss opportunities to develop more effective targeted strategies.

Tobacco control groups have tended to focus on treating smoking as a single issue, arguing that not only are sub-groups such as people with mental illness equally likely to benefit from population-based access control and promotion of long-term health consequences, but that developing targeted campaigns would weaken the effectiveness of the overall campaign [80]. Few would argue with the gains that the universal targeting of smoking cessation has made in the past decades in achieving reductions in the prevalence of smoking. And yet, the evidence here also suggests that substantial inequalities in benefit are accruing in the sub-population of individuals with mental illness and that strategies in addition to universal approaches are needed to address these vulnerable populations.

The universal approach may also miss the context in which young people initiate smoking or partake in some other risk-taking behaviour. A decline in smoking rates would be less meaningful if it is achieved through young people substituting some other harmful substance or behaviour to cope with their difficulties. Perhaps when population health goals are formulated, such as the Healthy People 2010 objectives [81], a broader context should be considered, such as reducing smoking related harm without increasing harm through other risk behaviours. As mentally ill smokers may be more susceptible to addictions in general [82], focussing narrowly on the smoking behaviour only without considering other risk behaviours may compromise the overall goal of smoking cessation – improved health outcomes.

Monitoring smoking in people with mental disorders in routine surveillance systems

The CIDI, which was used for assessing mental health conditions in both the NCS-R and the SMHWB, is time consuming to administer and impractical to incorporate in omnibus surveys of the type that are routinely used for monitoring smoking rates. The K6 measure of non-specific psychological distress has been included in the US National Health Interview Survey since 1997 [52], and has more recently been incorporated into the US Behavioral Risk Factor Surveillance System [83] and the Australian National Health Survey [84]. Although it is far less comprehensive than the CIDI, the strong association between smoking rates and level of psychological distress suggest it could be used to track progress in reducing smoking rates among the mentally ill. As the measure is already incorporated in monitoring surveys, we hope that it can be incorporated into routine surveillance of smoking rates. While some collections have employed a binary categorisation of psychological distress, the significant gradient in smoking rates and numbers of cigarettes consumed across low, mild, moderate, and severe levels of psychological distress suggest at least four or five categories should be employed to adequately monitor this relationship. The K6 scale is most closely correlated with anxiety and affective disorders and less relevant to measuring impulse control or substance use disorders. Its use in routine surveillance should be supplemented by other measures in these areas.

Currently there is no long-term time series on smoking rates in people with and without mental illness. Although the NCS and the SMHWB have both been run twice, the changes in instrumentation and the shift from ICD-9/DSM-III-R to ICD-10/DSM-IV limit the comparability of these figures over time. It is hoped that the inclusion of short mental health measures such as the K6 in routine health surveillance programmes will lead to the accumulation of time series data in this area.

Study limitations

All three surveys are based on self-report for both mental health status and smoking status. Diagnosis via the CIDI correlates well with diagnosis via structured clinical interview administered by a psychiatrist, but is not the same thing as a validated psychiatric interview. The NCS-R and SMHWB were both voluntary surveys, and response rates may be an issue. It is possible that people with mental illness were under-represented in the surveys due to non-response problems. The surveys targeted the non-institutionalised population, and will thus not include the high rates of mental health problems within institutional settings such as hospitals, hostels, and prisons.