Cigarette Smoking by Patients With Serious Mental Illness, 1999–2016: An Increasing Disparity

Cigarette smoking is the leading cause of preventable mortality in the United States (1). Smoking causes cancer, heart disease, stroke, lung diseases, diabetes, and chronic obstructive pulmonary disease (1). A higher proportion of persons with serious mental illnesses, particularly those with schizophrenia, smoke cigarettes, compared with the overall population (2,3). Smoking contributes to the premature mortality of persons with serious mental illness, a finding that has been well documented in many investigations (4–9).

In a previous study, we reported on cigarette smoking in a sample of 991 persons with schizophrenia or bipolar disorder and without a psychiatric disorder drawn from routine care settings in 1999–2011 who participated in our research studies; we found that 64% of individuals with schizophrenia, 44% with bipolar disorder, and 19% of those without a psychiatric disorder reported that they were current smokers (10). These group differences remained constant over the observation period, and there were no statistically significant time trends in smoking or cigarette consumption after adjustment for demographic covariates.

In the United States, intense public health campaigns directed toward the general population have resulted in steadily declining rates of smoking since the first Surgeon General’s report in 1964 (11–13). Efforts to address tobacco use by persons with serious mental illness have lagged significantly behind. Over the past five years, there have been more calls to action to promote smoking cessation in psychiatric populations (14–17). These initiatives are based on the stark data about increased morbidity and mortality among persons with serious mental illness that can be attributed in part to smoking. As part of this trend, more clinical trials have been performed demonstrating the safety and effectiveness of smoking cessation treatments for psychiatric patients (18–22), and the Centers for Medicare and Medicaid Services (CMS) now include as a quality measure the provision of smoking cessation treatment during a hospital stay and as part of the discharge plan (23). Data from national surveys collected five years ago or longer suggest that smoking is less likely to have declined among persons with more severe mental health problems compared with other segments of the population (24,25) and that the effect is stronger among older versus young adults (26). However, it is not known whether these trends have continued.

The purpose of this study was to expand our previous cohort with additional participants over the past five years, through 2016, in order to assess the temporal trends in smoking among persons with schizophrenia and bipolar disorder and among those without a psychiatric disorder, determine the clinical correlates of smoking status, and assess temporal trends in the quantity of cigarettes smoked per day.

Methods

Participants were individuals with a diagnosis of schizophrenia or bipolar disorder or without a diagnosed psychiatric disorder who were enrolled during the period January 1999 to November 2016 in the Stanley Research Program at Sheppard Pratt Health System in Baltimore for a study of the association between infection, immunity, and psychiatric disorders. All participants provided written informed consent after the study procedures were explained. The study was approved by the Sheppard Pratt Institutional Review Board.

The inclusion criterion for individuals with schizophrenia was a diagnosis of schizophrenia, schizophreniform disorder, or schizoaffective disorder. The inclusion criterion for individuals with bipolar disorder was a diagnosis of bipolar disorder, including bipolar I disorder, bipolar II disorder, or bipolar disorder not otherwise specified. The participants with psychiatric disorders were recruited from inpatient and day hospital programs of Sheppard Pratt and from affiliated psychiatric rehabilitation programs. The diagnosis of each of these participants was established by the research team, which included a board-certified psychiatrist, and was based on the Structured Clinical Interview for DSM-IV Axis I Disorders (27) and available medical records.

The inclusion criterion for the participants without a psychiatric disorder (control group) was the absence of a current or past psychiatric disorder as determined by screening with the DSM-IV Axis I Disorders, Nonpatient Edition (28). These persons were recruited from posted announcements at health facilities and universities in the same geographic area where the psychiatric participants were recruited.

Participants in all groups met the following additional criteria: age 18–65 (except for the control group, for which ages ranged from 20 to 60), proficient in English, absence of any lifetime history of intravenous substance abuse, absence of intellectual disability by history, absence of HIV infection, absence of a serious medical disorder that would affect cognitive functioning, and absence of a current primary diagnosis of alcohol or drug abuse or dependence per DSM-IV criteria. Participants were not selected on the basis of their smoking status.

As part of the background interview, each participant was asked, “Are you a current cigarette smoker?” And if yes, “How many packs per day do you smoke?” If the participant was on a smoke-free hospital unit, responses were based on his or her preadmission smoking. Each participant’s smoking status was assessed at one time point (study enrollment). All participants were also asked about demographic variables, including maternal education as a proxy for family socioeconomic status (29), and were assessed on a cognitive battery, the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS) (30). All participants with psychiatric disorders were evaluated on the Positive and Negative Syndrome Scale (31) and were categorized as to their history of alcohol or drug abuse or dependence (apart from caffeine). Medications received at the time of study enrollment for participants with psychiatric disorders were recorded.

Data Analyses

Demographic and clinical characteristics and smoking status were compared among study groups. Potential confounders were identified by determining which variables were significantly (p<.05) associated with both diagnostic group and smoking status in the entire sample and significantly associated with both diagnostic group and amount smoked among smokers. Analysis of variance was used for continuous variables, and Fisher’s exact tests were used for categorical variables.

Logistic regression was used to examine temporal trends in smoking prevalence and correlates of smoking. In all temporal-trends models, the variable “year” was the integer year of enrollment. Before logistic regression analyses were performed, mean imputation was used to fill in missing values for variables listed in Table 1. Because all demographic and clinical characteristics were highly associated with diagnostic group, the mean value within each diagnostic group was used; for categorical variables, the most common response was used. A logistic regression model was performed that included diagnostic group, year, and all variables identified as potential confounders. Next, interaction models were performed that included diagnosis × year interaction terms to determine whether the changes over time varied by diagnosis and an age × year interaction term to determine whether changes in smoking prevalence over time varied by participant age. Ordinal logistic regression was used to examine temporal trends in quantity of cigarettes smoked per day and to determine correlates of quantity of cigarettes smoked among participants who smoked. A modeling technique similar to that described above was used.

Results

Participant Characteristics

The sample consisted of 1,938 participants, including 991 participants who were in the previous study (10). The characteristics of the sample and smoking prevalence by diagnostic group are presented in Table 1. [Additional data about participants’ medications and treatment settings and the size of each diagnostic group by year are presented in an online supplement to this article.]

Trends Over Time and Correlates of Smoking

A logistic regression model was performed that consisted of schizophrenia diagnosis (yes-no), bipolar disorder diagnosis (yes-no), year of enrollment, age under 30 years (yes-no), gender, education (years), maternal education (years), and RBANS score. The variables age, gender, education, maternal education, and RBANS score were significantly (p<.05) associated with smoking (yes-no) in bivariate analyses and were included as covariates. The results of the logistic regression model are shown in Table 2. The effects of schizophrenia diagnosis (odds ratio [OR]=3.58, p<.001) and bipolar disorder diagnosis (OR=2.18, p<.001) were significant, indicating that smoking was more than three times as prevalent among participants with schizophrenia and more than twice as prevalent among participants with bipolar disorder than among participants in the control group, after the analysis took into account year of enrollment, age, and the other demographic covariates in the model.

The effect of enrollment year was also statistically significant (OR=.97, p=.017), indicating that smoking became less prevalent over time during this study, regardless of the participant’s age or diagnostic group. The effect of age under 30 was statistically significant (OR=.65, p<.001), indicating that participants who were under 30 were significantly less likely to be smokers than their counterparts over age 30, when the analysis adjusted for year of enrollment, psychiatric diagnosis, and demographic factors. Males were more likely than females to be smokers (OR=1.58, p<.001), and years of education was inversely associated with smoking (OR=.82, p<.001). The effects of maternal education and RBANS score were not statistically significant. Curves with locally weighted scatterplot smoothing of the percentage of smokers in each diagnostic group by year of enrollment, stratified by age group, are shown in Figure 1.

When an age <30 × year interaction term was added to the model, the interaction term did not achieve statistical significance (p=.051), and the main effects of age under 30 (p=.052) and year (p=.32) were no longer significant, indicating that the decrease in smoking prevalence over time did not vary substantially by age group.

When the interaction terms schizophrenia × year and bipolar disorder × year were added to the original model, the main effect of year and of the interaction term for schizophrenia × year were statistically significant (p=.004 and p=.024, respectively), but the interaction term for bipolar disorder × year was not (p=.11). When the ORs for the effect of year of enrollment on odds of smoking for each diagnostic group were calculated, the OR for the change in smoking prevalence over time showed a significant downward trend for the participants without a psychiatric disorder (OR=.92, 95% confidence interval [CI]=.86–.97. p=.004), and those for schizophrenia and for bipolar disorder were not statistically significant. This indicates that there was a significant decline during the study period in smoking prevalence for participants without a psychiatric disorder but not for participants with schizophrenia or bipolar disorder.

Quantity of Cigarettes Smoked and Trends Over Time

The sample was limited to smokers only (N=815), and an ordinal logistic regression model was performed that consisted of schizophrenia diagnosis (yes-no), bipolar disorder diagnosis (yes-no), year of enrollment, age under 30 years (yes-no), race, gender, education (years), maternal education (years), and RBANS score. The variables age, race, gender, education, maternal education, and RBANS score were significantly (p<.05) associated with the quantity of cigarettes smoked per day in bivariate analyses and were included as covariates in this model. The effects of both schizophrenia diagnosis (OR=3.70, CI=2.07–6.81, p<.001) and bipolar disorder diagnosis (OR=2.91, CI=1.65–5.26, p<.001) were statistically significant, indicating that compared with participants without a psychiatric disorder, those with schizophrenia were almost four times as likely and those with bipolar disorder patients were almost three times as likely to consume a higher quantity of cigarettes (in half-pack increments).

The effect of enrollment year was also statistically significant in this model (OR=.91, CI=.89–.94, p<.001), indicating that the quantity of cigarette consumption among all participants who were current smokers decreased over the course of the study, regardless of diagnostic group. The daily quantity of cigarette consumption was lower among participants who were under age 30 (OR=.62, CI=.44–.85, p=.003), independent of diagnosis or enrollment year. Cigarette consumption was higher among Caucasian smokers than among non-Caucasians (OR=2.59, CI=1.90–3.55, p<.001) and higher among males than among females (OR=1.35, CI=1.01–1.80, p=.04), and it was inversely associated with years of education (OR=.91, CI=.85–.97, p=.005). Maternal education and RBANS score were not significantly associated with quantity of cigarettes smoked. When schizophrenia × year and bipolar disorder × year interaction terms were added to the model, neither was statistically significant, indicating that the changes in cigarette consumption over time did not vary substantially by diagnostic group. Daily cigarette consumption by year of enrollment is shown graphically for the schizophrenia group in Figure 2.

Discussion

We found markedly elevated rates of smoking among individuals with schizophrenia and bipolar disorder over the past 18 years compared with participants with no psychiatric disorders. An average of 62% of those with schizophrenia and 37% with bipolar disorder identified themselves as cigarette smokers. By contrast, only 17% of the participants without a psychiatric disorder (control group) were current smokers. The prevalence of smoking in the sample as a whole decreased significantly over the study period, consistent with trends in the general U.S. population (1); however, the decrease was largely attributable to the control group. The prevalence of smoking in the two diagnostic subgroups in our study is generally consistent with reports in the literature, but data about smoking in psychiatric populations during the most recent five-year period are limited (2).

In our study, compared with the control group, participants with schizophrenia were more than three times as likely and those with bipolar disorder more than twice as likely to smoke, after adjustment for year of enrollment and demographic variables. The lack of significant time trends in the groups with psychiatric diagnoses is consistent with the results of our previous study and also with those of a study based on national data that covered a period through 2011 (10,25). The reasons for this lack of progress are multifactorial and may include limited treatment for smoking in psychiatric settings. This lack of progress has occurred against a background of calls to action by psychiatric leaders and organizations (14,16,32), research studies showing that treatments for smoking cessation are effective and well tolerated among persons with schizophrenia and bipolar disorder (18–22), and new CMS guidelines that provide incentives to hospitals for offering smoking cessation treatment to all patients (23). One can conclude that the myths about smoking and mental illness, articulated by Prochaska in 2011 (33), may remain prevalent. These include the misbelief that patients with mental illness need to smoke to cope with their symptoms, that quitting smoking leads to symptom exacerbation, that patients are not interested in quitting, and that they cannot quit—all of which have been found to lack supporting evidence.

The reasons that persons with schizophrenia and bipolar disorder are so much more likely to be smokers than those in the general population are not known with certainty, although a number of possible explanations have been offered in addition to the myths noted above. One suggested reason is that there may be an overlap in the genes associated with the risk for schizophrenia and with nicotine dependence; however, the genetic overlap is relatively small (34,35). In addition, randomized clinical trials have failed to show a consistent benefit in schizophrenia from medications that alter the function of nicotinic acid receptors (36,37). Another possible explanation is that patients derive special benefits from smoking in terms of ameliorating cognitive or psychiatric symptoms; however, the evidence here is also lacking (38,39). Another explanation, recently proposed, is that smoking is itself a risk factor for schizophrenia; the evidence to support this theory is incomplete (40,41). Social factors have also been cited as explanatory, including that smoking provides an opportunity for social interaction, that psychiatric settings tolerate smoking, and that patients with serious mental illness may be shielded from some of the stigma and public health messages about smoking (42). Many of these reasons may also account for low rates of quitting.

One encouraging finding is that smokers in our sample showed a statistically significant decline in the amount of cigarettes consumed over the 18-year study period. As an example of this trend, in 1999, the first year of data collection, 39% of participants with schizophrenia who smoked reported smoking more than a pack a day, but by 2016 that percentage was only 4%. It is not clear why the quantity of cigarettes consumed by smokers has declined, but the decline may be related to increased taxation on cigarettes and to state and local laws enacted during the study period that restrict smoking in public places. However, although smoking fewer cigarettes confers some harm reduction, only quitting altogether offers substantial health benefits (43).

Another encouraging finding is that persons under 30 in our sample were significantly less likely to be cigarette smokers than their older counterparts. This finding is consistent with another recent investigation of persons with serious mental illness that found that younger age was associated with less cigarette dependence (26). The other demographic and clinical correlates of smoking that we found are generally consistent with previous studies of persons with mental illness and the general population: lower education, male gender (1,44), a history of substance abuse or dependence (45), and Caucasian race (46); however, the correlation between Caucasian race and smoking was not found for the U.S. overall population (1).

The study had some limitations. Data through time were represented by different samples at each year, rather than by following the same participants over the study period. In addition, we did not collect information about participants’ smoking history, and thus we do not know the percentage of persons in each group who were previous smokers and who successfully quit. We also had only limited follow-up data about the smoking status of persons after they were assessed at the single time point. The small sample size by year in each group should also be noted. In addition, we did not collect data systematically about other nicotine products, such as e-cigarettes. Our sampling frame remained fairly constant over the study years, but research participants and those from intensive treatment programs tend not to be fully representative of patients overall. The schizophrenia group was disproportionately male compared with the other groups, and this may have inflated the prevalence of smoking in this group. Also, our enrollment of patients with schizophrenia began earlier than for the other groups, which may have affected group comparisons. Strengths of the study included the relatively large number of well-characterized participants studied over an 18-year period from the same set of treatment settings, and the unbiased collection of data about smoking.

Conclusions

The prevalence of smoking has remained alarmingly high among individuals with schizophrenia and bipolar disorder in our studies, and the disparity with participants without psychiatric disorders and with the general population is increasing. Additional measures are urgently needed to address this major public health problem.