Smoking Cessation Medication Prescribing for Smokers With and Without Mental Illness

People with mental illness have a higher prevalence of tobacco smoking and smoke more frequently, compared with people with no mental illness (1–4). Rates of smoking declined significantly during the 2000s among individuals without mental illness but not among those with mental illness (5, 6). Estimates show that less than 22% of individuals without mental illness smoke (1); however, the proportion is about 30% among those with any mental illness and over 60% among those with schizophrenia (1, 6, 7). Smokers with mental illness are also less likely to quit (4, 5). One study found a 25% lower likelihood of quitting by a 2-year follow-up for those with versus those without a mental illness diagnosis (2). High use and low quit rates among individuals with mental illness contribute to their being at significantly greater risk of dying from tobacco-linked diseases, such as chronic lower respiratory diseases, cardiovascular diseases, and cancers; tobacco-linked diseases account for about 50% of deaths among people with bipolar disorder, depression, or schizophrenia (8, 9).

Smoking cessation programs are most effective when they include pharmacotherapy and behavioral interventions (10–13), but only 4.7% of smokers receive cessation treatment that includes both (14). Evidence-based behavioral interventions include clinician counseling or cognitive therapy aimed at changing and managing behaviors, situations, and environmental cues associated with smoking; pharmacotherapy treatment includes three evidence-based medications approved by the U.S. Food and Drug Administration (FDA): varenicline, bupropion, and nicotine replacement therapy (NRT) (12, 15). Pharmacotherapy is safe and effective for smokers with mental illness, and evidence shows that these medications improve outcomes without worsening psychiatric symptoms (16–19). However, estimates indicate that only 25% of specialty mental health treatment facilities offer NRT and 21% offer non–nicotine cessation medications (20). To date, no studies have examined prescription patterns of cessation medications in a nationally representative sample of people with and without mental illness.

Smokers with mental illness may experience specific barriers to smoking cessation treatment, in part because of incorrect beliefs among health care providers that smokers with mental illness do not want to quit or would not benefit from quitting (21–23). People with mental illness commonly report that smoking alleviates symptoms of mental illness, such as poor concentration and stress (23–25); however, research shows that smoking is associated with worse long-term behavioral and general medical health outcomes for people with mental illness (26, 27). Nevertheless, providers have reported that they lack confidence to adequately address smoking cessation among patients with mental illness, and tensions exist between primary care providers and psychiatrists in deciding whose role it is to provide cessation treatment (22, 24, 28). To better understand how psychiatrists and general internists delineate care related to cessation treatment for people with mental illness, more evidence is needed.

This study had two objectives. The first objective was to examine trends in receipt of smoking cessation medications among smokers with and without mental illness in a nationally representative sample of the U.S. population. The second was to characterize general internist and psychiatrist attitudes and practices related to tobacco screening, smoking cessation treatment, and facilitators and barriers to delivering cessation treatment for patients with mental illness.

Methods

Quantitative

We analyzed Medical Expenditure Panel Survey (MEPS) data from 2005 to 2019 to examine receipt of smoking cessation medications among people with and without any mental illness, including serious mental illness. MEPS is a nationally representative longitudinal survey conducted by the National Center for Health Statistics covering the U.S. civilian noninstitutionalized population. It is the most complete data source on health insurance coverage and the cost and use of health care (29).

MEPS data files on demographic factors, medical conditions, prescribed medicines, inpatient and outpatient services, and office-based visits were merged across the 14 years of our study period (2005–2019). Final analytic files were constructed at the person-year level. Our sample was restricted to individuals ages 18 and older. People who smoke were identified by a binary variable of current smoking status for 2005–2017 and 2019. In 2018, MEPS removed this variable, so we created a comparable binary smoking variable by using a categorical variable on how often the patient smoked: smoking every day or some days was categorized as a current smoker. In 2017, MEPS included both variables, and sensitivity analysis showed less than a 5% difference between measures in identifying current smokers.

We constructed measures indicating receipt of one or more prescriptions for each FDA-approved smoking cessation medication—varenicline, bupropion, and NRT—and receipt of any cessation medication. Any mental illness was defined as an ICD-9 or ICD-10 diagnosis of mental illness (ICD-9 codes 295–301, 306–309, 311–314, and corresponding ICD-10 codes), any psychotropic drug use, a score of >12 on the Kessler Psychological Distress Scale (Kessler-6), or a Patient Health Questionnaire (PHQ) score of >2. The Kessler-6 is a six-item global measure of psychological distress including depression- and anxiety-related symptoms, and the PHQ is a nine-question instrument that screens for the presence and severity of depression. Serious mental illness was defined by ICD-9 or ICD-10 diagnosis codes for schizophrenia, bipolar disorder, or major depressive disorder (ICD-9 295–298) or a Kessler-6 score of >12 (30). The sample with any mental illness included those with serious mental illness.

We characterized the study sample by age (18–34, 35–54, 55–74, and ≥75), sex, race-ethnicity, and insurance status. The sample consisted of adult smokers between 2005 and 2019 and was divided between smokers without mental illness and with any mental illness, including serious mental illness. We measured the proportion of smokers who received varenicline, bupropion, or NRT and the proportion receiving any cessation medication. We used survey weights in the analysis to generate representative estimates. Data analysis was completed using Stata/IC, version 16.

Qualitative

To understand facilitators and barriers in delivering smoking cessation medications to patients with mental illness, we interviewed general internists and psychiatrists between October and November 2017. We used MedPanel, a health care provider research panel, to screen and recruit providers for our study (31). General internists were eligible if at least 30% of their work was clinical practice and their primary practice was either hospital outpatient, private practice, or a federally qualified health center. Psychiatrists were eligible if at least 30% of their work was clinical practice and either at least 10% of their patients had schizophrenia or bipolar disorder or their primary practice was a community mental health clinic. MedPanel screened based on region, and thus our sample was geographically diverse.

Interviews were conducted over the telephone by using a semistructured interview guide. On average, interviews lasted 16 minutes. Interviews were conducted by a graduate-level research team member (E.S.) trained in qualitative data collection. An oral consent process was completed before each interview. Interviews were audio-recorded and transcribed for analysis. Transcripts were validated by using the audio recordings, and any identifying personal information was removed from all transcripts.

The interview guide was created on the basis of a literature review and our research questions. All interviewees estimated the proportion of their patients with any mental illness and serious mental illness, defined as major depressive disorder, schizophrenia, or bipolar disorder. The guide had three domains: current cessation treatment practices, barriers and facilitators in the delivery of cessation treatment to individuals with mental illness, and differences in cessation treatment between individuals with and without mental illness (the third was for internists only, because psychiatrists do not treat patients without mental illness).

Transcripts were analyzed by using a hybrid inductive-deductive approach. The codebook development was informed by previous literature, our a priori knowledge, and summary memos created by the interviewer after each interview. The initial codebook was pilot tested and refined by two authors (E.S. and E.E.M.) through blind and independent double-coding until the organization of themes was consistent. The final codebook was applied to all interview transcripts. Coding and identification of themes was completed with NVivo 11. This research was reviewed and approved by the Johns Hopkins Bloomberg School of Public Health Institutional Review Board.

Results

Quantitative

Sample characteristics are presented in Table 1. The MEPS sample consisted of 55,662 smokers, 18,353 with evidence of any mental illness, 7,421 of whom had evidence of serious mental illness. Most were White (79.8%), male (55.5%), and between ages 35 and 54 (41.4%). Most smokers with mental illness and serious mental illness in the sample were female (56.0% and 57.1%, respectively), compared with 38.9% of smokers without mental illness. Among smokers without mental illness, 35.8% were privately insured, compared with 24.3% and 19.2% of smokers with any mental illness and with serious mental illness, respectively.

As shown in Figure 1, smoking cessation medications remained vastly underprescribed for all smokers in our sample. From 2005 to 2019, the proportion of all smokers—with any mental illness, with serious mental illness, and with no mental illness—prescribed any of the three cessation medications never exceeded 17%. The proportion receiving varenicline or NRT—frontline treatments for tobacco use disorder—never exceeded 3%. Across the 14-year study period, the proportion receiving varenicline peaked at 2.1%, in 2007, among those with no documented mental illness; 2.9%, in 2007, among those with any mental illness; and 2.4%, in 2019, among those with serious mental illness. The peak proportions for receipt of NRT were as follows: no documented mental illness, 0.4% in 2018; any mental illness, 1.1% in 2016; and serious mental illness, 1.6% in 2016. The peak proportions for receipt of bupropion were as follows: no documented mental illness, 1.2% in 2005; any mental illness, 8.4% in 2019; and serious mental illness, 16.7% in 2019.

Among smokers without mental illness, the proportion prescribed varenicline by a health care provider was 0.6% in 2006 (varenicline was not available for patients in 2005) and 0.6% in 2019. Among smokers with any mental illness, receipt of varenicline increased from 0.4% in 2006 to 1.8% in 2019. Receipt of varenicline among smokers with serious mental illness increased from 0.3% in 2006 to 2.4% in 2019. Among smokers without mental illness, receipt of NRT was less than 0.1% in 2005 and peaked in 2018 at just under 0.4%. Among smokers with any mental illness, receipt of NRT was 0.5% in 2005 and peaked at 1.1% in 2016. The proportion of smokers with serious mental illness prescribed NRT was 0.4% in 2005 and peaked in 2016 at 1.6%. Among smokers without mental illness, receipt of bupropion decreased from 1.2% in 2005 to 0.3% in 2019. For smokers with any mental illness, receipt of bupropion increased from 7.0% in 2005 to 8.4% in 2019. The largest increase in the prescription of smoking cessation medication for smokers with serious mental illness was found for bupropion, increasing from 8.0% in 2005 to 16.7% in 2019.

Qualitative

We interviewed 20 general internists and 20 psychiatrists; Table 2 summarizes provider characteristics. Most physicians were male (68%). Proportions by age group were as follows: 36–45, 13%; 46–55, 40%; ≥56, 48%. Half the interviewees (50%) worked in private practices. On average, general internists reported that less than one-third of their patient’s had any mental illness and just over one-tenth had serious mental illness. Psychiatrists reported that nearly two-thirds of their patients had serious mental illness (all had a mental illness). Most providers (85%) routinely screened patient’s smoking status. Most (95%) also reported offering smoking cessation medications, but only 58% personally offered counseling for smoking cessation. Ninety percent of psychiatrists reported that screening for smoking status was integrated into their practice’s electronic health records (EHRs), compared with 15% of general internists.

Qualitative themes were largely consistent across general internists and psychiatrists. Table 3 presents provider-reported facilitators and barriers to delivering smoking cessation treatment and differences in treatment approaches for patients with and without any mental illness. Three facilitators were most consistently noted: access to resources for patients, noted by 16 providers (eight internists and eight psychiatrists); the dual use of smoking cessation medications, mentioned by 12 providers (five internists and seven psychiatrists); and the importance of insurance coverage for cessation medication, noted by 11 providers (six internists and five psychiatrists). Three barriers most consistently cited were as follows: patients’ perception that tobacco is a helpful coping mechanism, noted by 28 providers (12 internists and 16 psychiatrists); patients’ unwillingness to engage in cessation practices, mentioned by 19 providers (11 internists, eight psychiatrists); and inability to use certain medications because of previous negative side effects for patients or interactions with psychotropic medications, noted by 14 providers (seven internists and seven psychiatrists).

Fifteen internists reported a perception that the rate of smoking among patients with mental illness is greater than among patients without mental illness. Fifteen internists also said that they approached smoking cessation treatment differently for patients with and without mental illness. This included tailoring conversations toward the patient’s mental health needs, possibly resulting in different medication and treatment options. Finally, 11 internists said that patients with mental illness often faced greater difficulty quitting, compared with those without mental illness, such as breaking away from the perception of smoking as “self-medicating.”

Discussion

Prescribing rates for all smoking cessation medications remained low among all smokers during our 14-year study period. From 2005 to 2019, nearly 98% of smokers, including those without mental illness, those with any mental illness, and those with serious mental illness, were not prescribed varenicline or NRT, which are first-line treatments for tobacco use disorder. Of the three medications studied—NRT, bupropion, and varenicline—bupropion was slightly more commonly prescribed, particularly among smokers with mental illness, but our data did not allow us to determine whether it was prescribed for smoking cessation, depression, weight loss, or a combination. The proportion of smokers with serious mental illness prescribed bupropion increased over the study period, a trend potentially driven by the 2009 PORT (Schizophrenia Patient Outcomes Research Team) guidelines recommending bupropion with or without NRT for smokers with serious mental illness (32).

In contrast to low rates of cessation pharmacotherapy evident in the national MEPS sample, general internists and psychiatrists in our qualitative convenience sample reported offering smoking cessation pharmacotherapy to patients with and without mental illness. The disparity between the MEPS results and practices reported in the qualitative interviews may be due to lack of representativeness of the qualitative sample—through underrepresentation of clinicians practicing in community settings that require documentation of tobacco use in EHRs (33). Additionally, the disparity may be due to patients’ lack of readiness to quit; their interest in quitting without pharmacotherapy; their ability to receive NRT without a prescription (e.g., over the counter or through quitlines), which would not be captured in the MEPS data (34); or providers’ underestimation of patients’ readiness to quit, which has been demonstrated in other studies (21, 22, 24, 25).

In 2009, the FDA added black box warnings to varenicline and bupropion to warn of risks of serious side effects—i.e., suicidal thoughts, depression, or hostility (35). However, these warnings were removed from both medications in 2016 for patients quitting smoking because of a review of large clinical trials showing that the mental health side effects of these medications were much lower than initially suspected (17, 36); the warning remains for bupropion for patients using it as an antidepressant drug. Studies examining trends in varenicline and bupropion use among individuals who were advised to quit smoking or who made quit attempts found trends broadly similar to those in the study reported here, with a decline in varenicline use at the time of the black box warning and fairly stable trends in bupropion prescribing (37, 38). However, our findings demonstrate an uptick in the prescription of bupropion for smokers with serious mental illness following the removal of black box warnings. Some providers interviewed reported that they were hesitant to prescribe varenicline for smokers with serious mental illness because of side effects that patients have reported or potential contraindications with medications a patient is taking, but no provider reported that these medications are clinically inappropriate for smokers with any mental illness.

Another major influence on the receipt of prescribed cessation medications is insurance coverage, which was emphasized by physicians interviewed in this study. More comprehensive insurance coverage of smoking cessation treatment is associated with increased cessation medication uptake and successful cessation rates (39–41). Coverage for tobacco cessation improved as part of the 2010 Affordable Care Act (ACA) (42, 43); however, our results do not suggest increases in the prescription of smoking cessation medications following ACA enactment. Evidence indicates that there are substantial barriers to more effective coverage, such as high cost-sharing and prior authorization requirements for treatment, which are widespread among commercial insurers, Medicare, and state Medicaid programs (44, 45). Further removing insurance barriers, such as copays, prior authorization, and limits on quit attempts, would likely improve access and outcomes related to smoking cessation treatment (39–41).

Prior research suggests that provider-level factors, such as beliefs about tobacco use and behavioral health and knowledge of clinical guidelines, have an important impact on cessation medication delivery (46, 47). Providers may be uncomfortable engaging patients they perceive are smoking as a coping mechanism or are not ready to quit (21, 24, 48), which was the most frequently mentioned barrier to cessation treatment noted by the general internists and psychiatrists in our study. Providers should offer pharmacotherapy to patients regardless of perceived quit readiness, because medication can reduce cravings and withdrawal, which may shift a patient’s willingness to quit (49). Additionally, providers may feel more comfortable relating consequences of smoking to medical issues that arise during a visit (e.g., wound healing and cancer screening). This strategy may allow clinicians to incorporate smoking cessation counseling into every visit, despite time constraints or not being a patient’s longitudinal provider (50).

At the system level, electronic decision support systems could incorporate more nuanced approaches for patients interested in quitting but who have not yet set a quit date, such as including pharmacotherapy options for patients prior to quitting or including a follow-up time frame. Such systems could also explicitly indicate that the known patient mental illness diagnoses are not contraindications to smoking cessation medications, a barrier noted by general internists and psychiatrists in our study, thereby leveraging real-time information from the EHR for provider education. In addition, availability of formulary information and costs for cessation medications may further address potential prescribing barriers.

These findings should be considered in light of several limitations. First, mental illness status was determined through medical records, which may be imprecise. Second, we were unable to determine whether bupropion was prescribed to treat depression, for weight loss, or for smoking cessation. Third, MEPS captures only the prescription of medications. NRT can be obtained without a prescription, and thus we likely did not capture the full use of NRT. Additionally, because MEPS cannot reliably measure receipt of behavioral counseling for smoking cessation, we focused on medications, which are a critical component of cessation treatment. Future work is needed to examine receipt of medication alongside behavioral therapy. Finally, qualitative results may be subject to social desirability bias, and providers may have reported that they prescribed more cessation medications than they did in actual practice.

Conclusions

Cessation pharmacotherapy for smokers remained vastly underprescribed across all groups. At least 83% of smokers with or without mental illness did not receive varenicline, NRT, or bupropion during the 14-year study period. Provider- and system-level strategies incorporating evidence-based smoking cessation treatment into standard workflows are needed to improve delivery of smoking cessation medication for people with and without mental illness.