Objectives Those with any psychiatric diagnosis have substantially greater rates of smoking and are less likely to quit smoking than those with no diagnosis. Using nationally representative data, we sought to provide estimates of smoking and longitudinal cessation rates by specific psychiatric diagnoses and mental health service use.
Design and participants Data were analysed from a two-wave cohort survey of a US nationally representative sample (non-institutionalised adults): the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC; 2001–2002, n=43 093; 2004–2005, n=34 653).
Main outcome measures We examined smoking rates (lifetime, past year and past year heavy) and cross-sectional quit rates among those with any lifetime or past year psychiatric diagnosis (DSM-IV). Importantly, we examined longitudinal quit rates and conducted analyses by gender and age categories.
Results Those with any current psychiatric diagnosis had 3.23 (95% CI 3.11 to 3.35) times greater odds of currently smoking than those with no diagnosis, and were 25% less likely to have quit by follow-up (95% CI 20% to 30%). Prevalence varied by specific diagnoses (32.4% to 66.7%) as did cessation rates (10.3% to 17.9%). Comorbid disorders were associated with higher proportions of heavy smoking. Treatment use was associated with greater prevalence of smoking and lower likelihood of cessation.
Conclusions Those with psychiatric diagnoses remained much more likely to smoke and less likely to quit, with rates varying by specific diagnosis. Our findings highlight the need to improve our ability to address smoking and psychiatric comorbidity both within and outside of healthcare settings. Such advancements will be vital to reducing mental illness-related disparities in smoking and continuing to decrease tobacco use globally.
- mental illness
Statistics from Altmetric.com
Current cigarette smokers are about half as likely to live to the age of 79 as individuals who never smoked.1 Those with psychiatric diagnoses are at increased risk of experiencing smoking-related morbidity and mortality due to exceptionally high rates of smoking in this subpopulation. Lasser et al2 found that 41.0% of those with a psychiatric diagnosis currently smoked, indicating nearly twofold greater prevalence than among those with no diagnosis (22.5%). Moreover, smokers with a diagnosis accounted for approximately 44.3% of cigarettes smoked in the USA. Cross-sectional cessation rates (ie, lifetime smokers who were no longer current smokers) were lower among those with a diagnosis than among those without (30.5% compared with 42.5%).
These estimates were based on data from 1990 to 1992 (National Comorbidity Survey (NCS)). Researchers have since examined differences in smoking based on mental illness using more current data.3 ,4 For example, using the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC) Wave 1 (2001–2002), Grant et al.4 found that, depending on specific diagnoses, those with psychiatric disorders were 2–16 times more likely to have nicotine dependence than those without these diagnoses. Lawrence and et al5 used data from the National Comorbidity Survey—Replication (NCS-R; 2001–2003) and the 2007 Australian Survey of Mental Health and Wellbeing to update and extend findings from Lasser et al2, and corroboratively found high rates of smoking among those with specific psychiatric disorders. Importantly, though, neither Grant et al4 nor Lawrence et al5 compared smoking cessation rates among those with psychiatric disorders. Using data from the National Survey on Drug Use and Health 2009–2011 surveys,3 the Centers for Disease Control and Prevention found that adults with mental illness had substantially higher rates of smoking (36.1% compared with 21.4% without mental illness) and lower rates of cessation. However, general mental illness (defined as non-specific psychological distress) was examined rather than specific DSM-IV psychiatric diagnoses. This is a highly relevant limitation, given important differences in smoking based on specific diagnoses.6 A limitation of both Lasser et al2 and the CDC report is that both investigations used cross-sectional data to examine cessation rates, rather than longitudinal data. Cross-sectional quit rates may be influenced by a number of historical factors, while longitudinal quit rates provide more accurate estimates of current differences.
The primary purpose of this study was to update and extend previous estimates of smoking and cessation among those with psychiatric diagnoses. The NESARC is the most recent longitudinal nationally representative survey with data on DSM psychiatric diagnoses and smoking cessation.7 ,8 The use of the NESARC has several advantages: significantly larger sample (n=43 093) than other national datasets with psychiatric diagnoses (eg, NCS, n=4411), standard measures of cigarette smoking and smoking cessation3 ,9 ,10 and longitudinal study design. The aims of the current investigation were to (1) estimate differences in smoking prevalence and quitting based on specific psychiatric diagnoses, (2) examine quit rates using longitudinal data, (3) study whether prevalence of heavy smoking increased with greater numbers of diagnoses, (4) examine differences in smoking among psychiatric diagnoses based on gender and age categories and (5) among those with psychiatric diagnoses, examine smoking rates and cessation by mental health treatment use.
The NESARC (wave 1: 2001–2002, n=43 093; wave 2: 2004–2005, n=34 653) is a survey of US civilian, non-institutionalised adults, administered with face-to-face, computer-assisted interviews in respondents' households. Self-identified African-Americans/Blacks, Hispanics and young adults were oversampled. The data were weighted to adjust for household and personal non-response, and to be representative of the US population (for a detailed account of the NESARC methodology, see refs.7 ,8). A subset of the original sample were contacted to participate in wave 2 (n=39 959; those who were not deceased, deported, mentally or physical impaired, or on active duty in the armed forces). The response rate for the second wave of data collection was 86.7%, and there was a mean of 36.6 months between interviews.
Axis I and axis II diagnoses were assessed using the Alcohol Use Disorder and Associated Disabilities Interview Schedule, DSM-IV version (AUDADIS-IV).11 ,12 The AUDADIS has demonstrated good-to-excellent reliability and validity in previous investigations.12 ,13 Lifetime diagnoses for axis I and axis II disorders included major depression, dysthymia, mania and hypomania; generalised anxiety, social phobia, agoraphobia, panic disorder and specific phobia; and alcohol abuse or dependence, drug abuse or dependence, and antisocial personality/conduct disorder. For lifetime psychotic disorder or episode, respondents were asked, “Did a doctor or other health professional ever tell you that you had schizophrenia or a psychotic illness or episode?” We separately examined past year diagnoses for these disorders, with the exception of antisocial personality/conduct disorder and psychotic disorder/episode. Past year diagnoses were defined as the presence of a lifetime diagnosis with active symptoms (enough to qualify for a continuing diagnosis) during the past year, as well as new diagnoses.
We defined cessation as long-term (at least 1 year) abstinence from all measured forms of tobacco (cigarettes, cigars, pipe, snuff and chewing tobacco).14 Using this definition, we generated measures of both cross-sectional quit rates (lifetime smokers, no current tobacco use at wave 1) and longitudinal quit rates (wave 1 smokers, no current tobacco use at wave 2). We defined heavy smoking as 24 or more cigarettes per day.2
Lifetime use of mental health services was assessed at wave 1 for the following psychiatric diagnoses: alcohol abuse/dependence, drug abuse/dependence, depression, dysthymia, mania, panic disorder, general anxiety, social phobia and specific phobia. For those with each of these lifetime diagnoses, respondents were asked if they ever had sought help through the following avenues: counsellor/therapist/doctor, emergency room, inpatient hospital and prescribed medications. We created a summary binary variable, coded 0 for not having sought any help and 1 for having sought any help.
We conducted analyses using Stata Statistical Software: Release 12,15 accounting for the NESARC survey design in all estimates. We first estimated prevalence and cessation rates for the following groups: (1) no diagnosis, (2) any lifetime diagnosis and (3) any past year diagnosis. We then calculated these estimates for each specific lifetime and current diagnosis. We examined the significance of all bivariate associations using Wald tests. We also calculated the prevalence of light–moderate smoking (0–23 cigarettes per day) and heavy smoking (≥24 cigarettes per day) based on number of diagnoses (0, 1, 2, 3 or 4+). We tested the significance of these differences using multinomial logistic regression. We first entered the variable for number of diagnoses as categorical (0 diagnoses as reference) in order to estimate prevalence for each group and test the significance of differences from those with 0 diagnoses. We then entered this variable as continuous to examine whether the likelihood of light/moderate or heavy smoking increased linearly with number of diagnoses. We used logistic regression to estimate associations between psychiatric diagnoses and lifetime smoking, current smoking and cross-sectional quit rates, adjusting for age, gender and education. These covariates were selected to account for associations between sociodemographic characteristics and both smoking and psychiatric diagnoses. We calculated relative risks for quitting by follow-up using generalised linear models, specifying a binomial distribution and a log link, and adjusting for the same sociodemographic covariates. We repeated our estimation of prevalence and bivariate associations for gender and age categories (see online supplementary tables S1–S8 and figures S1 and S2). In our final set of analyses, we selected for those who had current/lifetime mental illness, and examined associations between lifetime treatment use and smoking outcomes (prevalence and cessation) using the procedures outlined above. Regarding missing data, n=444 wave 1 respondents (1.0%) and n=62 wave 2 respondents (<1.0%) had unknown current smoking status. These respondents were not included in the analyses. There was no missing data for any of the diagnostic variables. For all analyses, we used a significance cut-off of 0.001 to account for multiple testing.
Prevalence of smoking across lifetime and current psychiatric diagnoses and sociodemographic subgroups are displayed in table 1.
Prevalence of smoking and cessation rates by any lifetime or past year diagnoses
Current smoking prevalence rates were 15.5% for those with no diagnosis compared with 33.4% for those with a lifetime diagnosis and 39.0% of those with a past year diagnosis (table 2). Those with psychiatric diagnoses were less likely to have quit at follow-up (18.4% and 17.7% for lifetime and current diagnosis compared with 22.3% for no diagnosis). These differences persisted after adjusting for age, gender and education (table 2). Those with a current diagnosis had 3.23 times greater odds of being a current smoker than those with no diagnosis (OR=3.23, 95% CI 3.11 to 3.35); and current smokers with a diagnosis at wave 1 were 25% less likely to stop using tobacco by wave 2 (RR=0.75, 95% CI 0.70 to 0.80). Cross-sectional quit rates were lower for those with a diagnosis (36.5% and 28.6% for lifetime and current diagnosis, respectively, compared with 48.3% for no diagnosis).
Prevalence of heavy smoking by count of psychiatric diagnoses
Compared with those with 0 diagnoses, those with multiple diagnoses had significantly greater likelihood of being a heavy smoker (figure 1; all differences p < 0.001). For example, the proportion of heavy smokers among those with 0 diagnoses was 3.7% compared with 16.1% for 4+ diagnoses. There was also a significant linear trend, whereby each additional diagnosis (from 1 to 4+) was associated with 67% greater odds of being a heavy smoker (p<0.001).
Prevalence of smoking among those with specific lifetime or past year psychiatric diagnoses
Smoking prevalence (both lifetime and current) was significantly higher for those with each lifetime disorder than for those with no diagnosis (p<0.001), and cross-sectionally and longitudinally assessed quit rates were significantly lower (p<0.001) (table 3). All past year diagnoses were associated with higher smoking prevalence and lower quit rates than those with no psychiatric disorders (p<0.001) (table 4).
Mental health treatment use
Treatment use was assessed for alcohol use disorder, drug use disorder, mood disorders and anxiety disorders. Among those with a lifetime diagnoses in any of these categories, 35.5% of respondents reported seeking help for their disorder. Among those who ever sought help, there was a lifetime smoking prevalence of 60.2% compared with 54.1% among those who had never sought help (OR=1.29, 95% CI 1.24 to 1.34). A similar pattern was found for current smoking, whereby those who sought help had a prevalence of 38.7% compared with 30.5% among those who ever sought help (OR=1.44, 95% CI 1.39 to 1.50). Lifetime smokers who ever sought help were less likely to have quit smoking by wave 1 (33.0% vs 38.9%; OR=0.85, 95% CI 0.82 to 0.87) or by wave 2 (16.3% vs 19.7%; RR=0.83, 95% CI 0.76 to 0.90).
Among those with a past year diagnosis of these select disorders, 42.7% reported ever seeking help for their disorder. Those who sought help were more likely to be lifetime smokers (61.1% vs 52.8%; OR=1.46, 95% CI 1.38 to 1.54) and current smokers (42.9% vs 35.2%; OR=1.38, 95% CI 1.30 to 1.46). Lifetime smokers with a past year diagnosis who ever sought help were slightly less likely to quit smoking by wave 1 (29.0% vs 27.6%; OR=0.95, 95% CI 0.91 to 1.00) or by wave 2 (15.1% vs 18.9%; RR=0.80, 95% CI 0.71 to 0.89).
Supplemental analyses: differences by gender and age categories
Supplemental analyses are reported in tables S1–S8 and figures 1 and 2 (online only). Gender differences varied by specific diagnosis. Men tended to have higher smoking prevalence than women; however, women with a lifetime alcohol use disorder or conduct/antisocial personality disorder were more likely to currently smoke than men with these corresponding diagnoses. Men with a past year diagnosis of agoraphobia, panic disorder and specific phobia were less likely to stop using all forms of tobacco by follow-up than women (p<0.001). Both men and women who sought help for their disorder were more likely to smoke and less likely to quit smoking than those who did not seek help, although these differences were slightly larger for men compared with women.
Regarding age differences, those in the youngest age group (18–29) typically had the highest rates of current smoking (compared with those in the 30–44 and 45+ age groups). This group also tended to be most likely to have quit at follow-up, with one exception: young adults with a current or lifetime diagnosis of social phobia were the least likely to have quit at follow-up among the three age categories (p<0.001). Results for treatment use analyses followed the same pattern as the general sample, with those who sought help having higher prevalence of smoking and lower quit rates than those who did not seek help, for all age groups.
In this US nationally representative sample, smokers with current psychiatric disorders have substantially higher prevalence of smoking than those with no diagnosis (39.0% vs 15.5%). Longitudinal quit rates indicated that those with psychiatric diagnoses had 25% lower likelihood of quitting by follow-up compared with those without a diagnosis. These differences in smoking prevalence and cessation rates between those with and without diagnoses were consistent across sociodemographic subgroups (eg, income, education and race/ethnicity) and remained significant after accounting for these sociodemographic variables as covariates. Prevalence was higher and quit rates were lower among those who had ever sought help for their disorder. Prevalence varied widely across specific disorders (23.4% to 66.7%), while there was somewhat less variation in quit rates (10.3% to 17.9%). Those with multiple lifetime diagnoses (40.1% of smokers) were more likely to smoke heavily than those with one or no diagnosis.
There was substantial overlap between psychiatric diagnoses in this study. This was evidenced by the high rates of psychiatric comorbidity reported in the results. Smoking among those with psychiatric comorbidity is an important issue, especially considering the particularly high rates of heavy smoking among those with multiple diagnoses and the paucity of research that addresses this topic. Regarding specific diagnoses, we were unable to make statistical comparisons (due to overlapping diagnoses); however, there were notable trends in the findings. Consistent with previous research,2 ,5 we found the highest prevalence of smoking among those with current substance use disorders. Interestingly, we found those with alcohol use disorders had the lowest cross-sectional quit rates relative to other diagnoses (consistent with Lasser et al,2), but had among the highest prospective quit rates. This likely reflects the age composition of those with alcohol use disorders, with younger adults more likely to have this diagnosis, and younger adults having the lowest cross-sectional quit rates and the highest longitudinal quit rates. This contradiction between cross-sectional and longitudinal quit rates highlights the methodological importance of examining cessation longitudinally, given the number of factors that can potentially influence commonly reported cross-sectional quit rates.
Cessation rates were generally lower among those with mood or anxiety disorder diagnoses. There was also a trend whereby those with disorders that are characterised by more consistent symptoms over time (eg, dysthymia, generalised anxiety) had lower cessation rates than those with disorders characterised by more episodic symptom profiles (eg, major depressive episode, panic disorder). This pattern of results was consistent with Lasser et al2 and may be indicative of more difficulty with stopping tobacco use among those with disorders characterised by unremitting symptomatology.
Large portions of those with psychiatric disorders reported they had never sought treatment for their disorder. This was particularly true for some specific sociodemographic subcategories. For example, young adults, despite being the age group with the highest prevalence of smoking, were the least likely to report having sought help for their disorders. This highlights the importance of studying and implementing public health interventions that reach smokers with psychiatric disorders outside of the healthcare system. As noted by multiple research groups,6 ,5 ,16 population-level interventions have not been the focus of tobacco control efforts among those with psychiatric diagnoses to date. Research and interventions have nearly exclusively focused on mental health treatment settings, resulting in a paucity of research on how population-level interventions may influence smoking rates among those with psychiatric diagnoses. Lawrence et al5 discussed a number of reasons that current population-level tobacco control interventions may be less effective for those with mental illness. For example, smoking bans and associated stigma may contribute to social isolation among those with psychiatric disorders. Policy related to pricing may place a disproportionate financial burden on those with psychiatric disorders and their families. Interventions that focus on the negative health effects of smoking may be less influential among those with psychiatric disorders, who may place less value on long-term health outcomes.
Concordantly, a substantial portion of those with psychiatric diagnoses reported seeking help for their disorders, supporting the continued investigation of integrating and improving cessation interventions in mental healthcare settings. These efforts are ongoing—many psychiatric hospitals have banned cigarette smoking and implemented smoking cessation programmes.17 There is promising evidence for effective cessation therapies designed for those with specific diagnoses,18 ,19 and ample evidence that smokers with psychiatric comorbidity are able to quit.20 Yet, despite these advances, multiple recent investigations and reviews have noted that non-treatment remains the norm,21–23 and smoking bans in psychiatric settings have been ineffective at generating lasting smoking cessation.17 Ziedonis et al6 outlined several recommendations for improving smoking cessation outcomes in mental healthcare settings, including (1) study of the interaction of psychiatric and smoking cessation treatments, (2) adequate samples and power in smoking cessation trials, (3) the adaptation of smoking cessation treatments to psychiatric populations and (4) integration of smoking cessation treatments within the current mental health treatment system.
The NESARC dataset was the most current and comprehensive dataset with which we could address the aims of this investigation. Still, there were limitations of this study to note. Although we were able to look at specific diagnoses and categorise by reports of treatment use, the NESARC data were not designed to distinguish between varying levels of mental illness severity. Among those who reported ever seeking treatment, smoking rates were higher and quit rates were lower. It is likely that treatment use was a proxy for symptoms severity. Additionally, there was no information on whether the respondents were currently in treatment or the extent of treatment success. Thus, we were unable to examine smoking outcomes by these more nuanced characterisations of those with psychiatric diagnoses. The estimates in the current study were based on data from 2000 to 2005, reflecting the most recent available data on smoking and cessation in the USA rather than the current US population. A limitation of this and other similar studies was that cigarette smoking and tobacco use measures were based on self-report; however, our broad definitions of smoking and tobacco use (any over the past year) would likely have reduced any recall bias. This study did not include diagnoses for all axis I and axis II psychiatric disorders known to be associated with smoking (ie, posttraumatic stress disorder, attention deficit hyperactivity disorder). However, misclassification of some individuals with disorders as having no diagnosis would have conservatively biased difference estimates. The study was based in the USA, and it is unclear how the findings may generalise to other parts of the world.
In conclusion, those with psychiatric diagnoses were substantially more likely to smoke cigarettes, and among those who smoked, were less likely to stop using tobacco compared with those with no disorders. This was particularly true for those with comorbid lifetime disorders, who made up nearly half of this nationally representative sample of smokers. Results varied by specific diagnoses, gender and age categories, suggesting the influence of treatment and policy may vary based on these subgroups as well. Continuing progress in reducing smoking in the USA will require advances in understanding the complexities of smoking among those with specific diagnoses and combinations of diagnoses, and the application of this knowledge to improving tobacco-control interventions and policies.
What this paper adds
Those with psychiatric diagnoses were more likely to smoke, smoked more heavily, and were less likely to quit smoking during the study period than those with no diagnoses.
Smoking increased with the number of co-morbid diagnoses.
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 figures
- Data supplement 2 - Online table 1
- Data supplement 3 - Online table 2
- Data supplement 4 - Online table 3
- Data supplement 5 - Online table 4
- Data supplement 6 - Online table 5
- Data supplement 7 - Online table 6
- Data supplement 8 - Online table 7
- Data supplement 9 - Online table 8
Correction notice This article has been corrected since it was published Online First. The wording in the abstract has been amended from ‘within and without healthcare’ to ‘within and outside of healthcare’.
Contributors All authors designed the study. PHS and SAM planned the analysis. PHS conducted study analysis and drafted the paper. All authors revised and edited the paper.
Competing interests All authors have completed the Unified Competing Interest form at http://www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare no support from any organisation for the submitted work, no financial relationships with any organisations that might have an interest in the submitted work in the previous 3 years and no other relationships or activities that could appear to have influenced the submitted the work.
Funding The project described was supported by grant number P50 DA033945 from the National Institute on Drug Abuse (NIDA), the Food and Drug Administration (FDA) and the Office of Research on Women's Health (ORWH), OD. This work was also supported by the National Institute of Mental Health (T32-MH014235, PI: Heping Zhang), and Women's Health Research at Yale. Study funders had no role in the collection, analysis or interpretation of data; in the writing of the report; or in the decision to submit the article for publication. All researchers were independent from funders.
Ethics approval This study consisted of secondary analyses of de-identified data, and was therefore exempt from formal ethics approval.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement All data are available to all authors. This is a publically available de-identified dataset.
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.