A Pragmatic Cluster-Randomized Trial of Provider Education and Community Health Worker Support for Tobacco Cessation

Persistently elevated tobacco use disorder rates among persons with serious mental illness (1) contribute significantly to premature mortality in this population (2, 3). Quitting smoking is associated with a reduced risk for tobacco-related disease and premature death and with improved quality of life (4, 5). Although both safety and efficacy of first-line pharmacotherapeutic smoking cessation aids—varenicline and combination nicotine replacement therapy (NRT) (i.e., combining short-acting NRT [nicotine gum or lozenge] and long-acting NRT [transdermal nicotine patch])—have been established (6–13), these aids remain underutilized among adults with serious mental illness (14, 15), and behavioral cessation treatments delivered without pharmacotherapy yield unacceptably low tobacco abstinence rates in this population (6, 8). Few strategies to improve delivery of first-line smoking cessation pharmacotherapy in community mental health settings have been tested (16–19).

Barriers to provision of smoking cessation medication to people with serious mental illness who smoke include a lack of knowledge among clinicians regarding the safety of smoking cessation medications for people with serious mental illness, clinicians’ self-assessed perception of inadequate training in tobacco use disorder treatments (20), and the assumption that people with serious mental illness have low motivation to quit smoking and a low likelihood of success (21–23). Educational outreach is a form of continuing medical education that can improve provision of evidence-based clinical care (24) through correction of provider knowledge gaps and biases (25). Such outreach has been reported to increase smoking cessation pharmacotherapy prescriptions (26), tobacco quitline referrals (27), and smoking cessation rates (28) when delivered in primary care settings. Early evidence suggests that educational outreach through academic detailing with decision support and data-driven feedback increases use of smoking cessation medications and tobacco abstinence in community mental health settings (19).

Community health worker (CHW) interventions can improve adherence to evidence-based treatments (29), health behavior change, and medical outcomes for low-income patients (30) by providing health education, helping to address adverse social determinants of health and other barriers to care, and assisting with care navigation and patient follow-up (31). To our knowledge, no studies have assessed the impact of CHW support on health behaviors among individuals with serious mental illness.

Here, we tested the effect of provider education (PE) in the form of clinician-delivered educational outreach to primary care and affiliated psychiatric providers alone and in combination with CHW support for smoking cessation over a 2-year period in a population of adult outpatients with serious mental illness who were receiving psychiatric rehabilitation services and who did not necessarily want to quit smoking. The study design included cluster randomization of clinicians for receipt of educational outreach. The primary outcome was biochemically verified, 7-day point prevalence tobacco abstinence at the year 2 assessment. Our hypothesis was that the combined PE and CHW intervention would increase abstinence rates over PE alone and over usual care.

Methods

We invited all adults who smoked tobacco and were eligible for Department of Mental Health (DMH) psychiatric rehabilitation services for serious mental illness through the two behavioral health organizations serving the largest number of people in Greater Boston to participate in this pragmatic randomized controlled trial. Participants who agreed were interviewed annually at baseline and in years 1 and 2 of the intervention. The study was approved by Massachusetts General Hospital and DMH institutional review boards. Participants received written information about the study before the first survey, and participation in the survey was considered evidence of consent, with written informed consent requirements waived for annual surveys.

Study staff contacted all adult outpatients who were eligible for DMH-contracted psychiatric rehabilitation services from Bay Cove Human Services, Inc. and Vinfen Corp. and who at their annual evaluation reported smoking in the past 30 days. Desire to change smoking behavior was not required for enrollment.

Of 1,496 potentially eligible participants, 276 (18%) declined to participate, 55 (4%) were not outpatients and were thus ineligible, and 1,165 (78%) agreed to participate and provided information on their primary care clinic site. Of the 1,165 who agreed to participate, 155 (13%) were ineligible because they received primary care at a clinic serving fewer than three participants; thus, 1,010 participants were enrolled in the trial and randomly assigned to a treatment arm (a flow diagram is included in an online supplement to this article). Enrolled participants who received primary care at a clinic treating too few participants to be eligible for PE were enrolled into a separate cohort and randomly assigned to CHW versus usual care to gather preliminary data on the effect of CHW alone (see table in online supplement).

Primary care clinics serving three or more enrolled participants were cluster randomized to one of two conditions: PE provided to clinical staff or no PE (usual care). Randomization was done by staff otherwise not involved in the study who used a random sequence that was computer generated at the Massachusetts General Hospital Biostatistics Center in a 2:1 ratio. Clinics that shared staff were yoked to minimize contamination, resulting in a 2.5:1 clinic ratio of PE to usual care. A nested participant-level randomization to CHW or no CHW in a 1:1 ratio was conducted by random computer-generated sequence among participants who received care in clinics assigned to PE. Consent was waived for providers; attendance at educational sessions was considered evidence of consent and was incentivized with free continuing medical education credits.

PE Intervention

Doctoral-level clinical study staff (A.E.E., C.C., S.R., A.N.T., K.S.) received 2 days of training in PE delivery (32). From December 2017 to February 2020, they then delivered educational outreach to primary care physicians (PCPs) and any psychiatrists, physician assistants, and prescribing nurses in the primary care clinical settings. PE aimed to increase use of first-line smoking cessation pharmacotherapy by identifying and overcoming clinician knowledge gaps and biases (21, 33) regarding risk-to-benefit ratio of prescribing first-line, evidence-based smoking cessation medication, particularly varenicline and a combination of short- and long-acting NRTs, to people with serious mental illness who smoke, emphasizing substantial benefits and limited risks of smoking cessation pharmacotherapy, risks of not using smoking cessation medication, and low abstinence rates with behavioral smoking cessation treatment alone. PE also drew attention to the removal of safety warnings for varenicline and bupropion by regulatory agencies and highlighted recent practice guidelines (34) recommending that smoking cessation medication be offered to all smokers (35, 36), using an “opt-out” approach (37) (see figure in the online supplement). Both PE and CHW interventions encouraged use of varenicline because of its superior efficacy (7, 8) and relative underuse in this population (14, 15) and also encouraged behavioral support for abstinence.

CHW

Assigned participants provided written informed consent before receipt of CHW support. Participants who refused CHW support remained in the study, because evaluation of CHW support used intent-to-treat analyses. Twelve bachelor’s-level staff without previous clinical training or experience in working with people with serious mental illness or in smoking cessation were hired by partnering psychiatric rehabilitation agencies. The CHW role was multifaceted, with a strong supervisory structure (38, 39). CHWs received 2 weeks of training that included core competencies for CHW certification (40), agency-required training in working with people with serious mental illness, Tobacco Treatment Specialist Core Training (41), and review of PE materials. CHWs received 1 hour of weekly clinical group supervision and individual supervision as needed (39). Participants met with CHWs on a mutually determined schedule from November 2017 to January 2020 in their homes or neighborhoods (42, 43). Participants assigned to CHW support had access to community-based smoking cessation group counseling sessions co-led by a CHW and clinically trained study staff (C.C., G.N.P., S.P., B.D.).

CHWs were introduced to participants by a member of their psychiatric rehabilitation team. CHWs aimed to build trust, support identification of smoking cessation as a goal, and encourage engagement in evidence-based smoking cessation pharmacotherapy. CHWs provided education on safety and efficacy of smoking cessation medications for people with serious mental illness (10, 11), encouraged all participants to try smoking cessation medication (35), provided transportation and accompaniment to physician visits and to group counseling sessions on smoking cessation, assisted with communication with PCPs and community support staff, and often helped with unmet health-related social needs (see table in the online supplement for details on CHW approaches and activities).

The primary outcome was 7-day point prevalence abstinence at 24 months, defined as self-report of no smoking in the past week and expired air carbon monoxide (CO) of ≤5 parts per million. Secondary outcomes included change in quantity, frequency, and type or types of tobacco products used; prior-year receipt of physician advice to quit smoking and one or more prescriptions for NRT, varenicline, or bupropion; and quantity and frequency of behavioral and pharmacotherapeutic smoking cessation treatment used. Severity of nicotine dependence was assessed with the Heaviness of Smoking Index questionnaire (44). Participants were asked to report any smoking-related illnesses, and a single-item self-report of overall health, the SF-1 (45), was included. Expired CO concentration was measured with a Bedfont Smokerlyzer II. Outcome assessments were conducted in participants’ communities at baseline and years 1 and 2 of the intervention by assessors blind to intervention allocation.

The primary analysis compared biochemically verified year 2 abstinence rates in the PE+CHW, PE, and usual care arms in a pairwise manner by using logistic regression with a random effect for clinic. Participants lost to follow-up before the year 1 assessment did not provide adequate information for the multiple imputation and were excluded from the primary analysis. For participants who completed baseline and year 1 surveys, year 2 missing data were imputed by using chained equations (R package “mice,” version 3.10.0) (46). Data were imputed 10 times, and estimates from the logistic regression model were pooled by using Rubin’s method (47). Data used for imputing year 2 abstinence status were year 1 smoking status, age, sex, race (Black, White, or other for races with fewer study participants), ethnicity (Hispanic or not Hispanic), baseline housing status (independent or supervised), SF-1 score, self-report of smoking within 5 minutes of awakening, and pharmacotherapeutic smoking cessation treatment prescribed (yes or no) or used (yes or no) in the year before baseline.

In prespecified mediation analyses, we assessed associations between intervention assignment, smoking cessation pharmacotherapy use (defined as any varenicline use, single or combination NRT use, or bupropion use), and year 2 abstinence via path analyses based on logistic regression, using the R package “mediation” (version 4.5.0) (48). In these analyses, we sought to determine the influence of each intervention mediated via its impact on medication use, the unmediated impact of the interventions on abstinence, and the impact of medication use on abstinence, operationalized as self-report of taking at least one dose of the medication. Separate analyses assessed the impact of taking any smoking cessation medication and of taking varenicline. Estimates of the indirect effect of the interventions were computed per each imputation run and again pooled by using Rubin’s method (47).

Post hoc analyses explored the effects of CHW assignment on taking at least one dose of smoking cessation medication and its effectiveness, operationalized as year 2 abstinence. We used linear probability models to estimate abstinence rates by CHW status, medication use, and their interaction, adjusting for PE and including a clinic-varying random intercept. In this analysis, we excluded 223 participants missing both year 1 and year 2 surveys; because of small cells, we also excluded 21 participants who took bupropion.

Results

Participant characteristics are summarized in Table 1. Participants reported moderately severe nicotine dependence, as measured by the Heaviness of Smoking Index. In the overall sample of 1,010 enrolled participants, about 40% smoked tobacco products other than commercial cigarettes; 33% smoked little cigars, which are often chosen because of their lower cost compared with cigarettes due to differing excise taxation rules for cigars versus cigarette products (49). Very few participants (1%) used e-cigarettes. In total, 65% reported receiving prior-year physician advice to quit smoking; 33% reported receiving a smoking cessation medication, primarily a nicotine patch, and 68% reported having a smoking-related medical illness.

Clinics were cluster randomized to receive PE (55 clinics serving 677 participants) or usual care (22 clinics serving 333 participants). The 677 participants whose clinic was assigned to PE were randomly assigned to be offered CHW support (PE+CHW, N=336) or not (PE only, N=341) (see figure in the online supplement). Of the 1,010 participants, 608 had observed data and 179 had imputed data for year 2 outcomes. Of the 1,010 participants, 223 (22%) were lost to follow-up before the year 1 assessment, of whom 56 were lost because of severe illness or death (32 died, and 24 were too severely ill to be interviewed safely), 72 were lost after a rule change made them ineligible for ongoing psychiatric rehabilitation services, and 95 were lost because they refused to participate further or could not be contacted. Retention rates did not significantly differ by intervention assignment.

PE was delivered in person to 50 clinics and via written materials to three clinics. No PE was attempted at one clinic that closed before PE could be provided or at another clinic that had fewer than three participants at the time the intervention began (see table in the online supplement). Of 336 participants assigned to CHW support, 220 (65%) consented to receive it. These participants had a median of 16.5 CHW contacts (interquartile range=39.5) over a mean±SD of 14.2±9.4 months. The mean length of visits was 35.1±17.3 minutes, and 5,298 of 6,192 visits (85.6%) were in person. Of participants who received CHW support, 100 (45%) attended a smoking cessation group; the mean number of sessions attended was 29.1±26.1. No unanticipated harms were attributable to study interventions (see table in the online supplement).

In the primary analysis, year 2 abstinence rates were higher among participants assigned to PE+CHW than among those assigned to usual care (12% [N=32 of 259] vs. 5% [N=13 of 264]; adjusted odds ratio [AOR] with multiple imputation=2.40, p=0.013). Year 2 abstinence rates were also higher among those assigned to PE+CHW than among those assigned to PE alone (12% [N=32 of 259] vs. 7% [N=19 of 264]; AOR=1.84, p=0.036). Abstinence rates among those receiving primary care at a clinic assigned to PE alone were not significantly different from abstinence rates among those receiving usual care (Tables 2 and 3). Alternative methods for handling missing data yielded similar results, whether a completer analysis was conducted or whether the analysis considered all participants nonabstinent who were missing data at the year 2 assessment (see tables in the online supplement).

The effect of PE+CHW on abstinence was mediated by greater varenicline use (OR=2.77, p<0.001), which nearly doubled the odds of abstinence (OR=1.97, p=0.012) (Figure 1B). A significant direct effect of the PE+CHW intervention on abstinence was also noted and was independent of varenicline use (OR=2.15, p=0.022). Although the point estimates were positive for the effect of the PE+CHW intervention on use of any smoking cessation medication and the effect of any smoking cessation medication use on abstinence, these effects were not statistically significant (Figure 1A). PE alone had no significant effect on smoking cessation medication use, varenicline use, or abstinence, although varenicline use had a significant effect on abstinence among those assigned to PE (see figure in the online supplement).

To further explore CHW effects on medication effectiveness, we modeled abstinence rates by randomization to a CHW or not (Table 4). In this exploratory analysis, among participants who received no smoking cessation medication, CHW assignment was not significantly associated with higher abstinence rates in year 2, compared with those not assigned to a CHW. Among participants who received smoking cessation medication, those assigned to receive CHW support had approximately threefold higher year 2 abstinence rates, compared with those not assigned to receive CHW support. Among those who received varenicline, the number needed to treat (with offer of CHW support) for an additional participant to quit smoking was 4.6.

Discussion

As we hypothesized, the results show that PE combined with CHW support was superior to PE alone and to usual care for smoking cessation among adults with serious mental illness who smoked tobacco. Year 2 abstinence rates among participants served at clinics assigned to receive PE who were offered CHW support were more than double the rates among those receiving usual care. Importantly, these results were observed among participants who were enrolled without regard for their reported readiness to quit smoking. No evidence was observed for an effect of PE alone.

The effect of PE and CHW support on abstinence was mediated in part by greater use of varenicline. Participants assigned to PE clinics and CHW support had nearly threefold greater odds of varenicline use, and varenicline use in the context of PE+CHW nearly doubled abstinence rates. Consistent with previous findings, the effect of varenicline on abstinence was stronger than the effects of other smoking cessation pharmacotherapies (7). Of note, those who received varenicline, NRT, or both in the context of PE+CHW support had approximately threefold higher abstinence rates than participants who received these medications without CHW support. This exploratory finding suggests that the PE+CHW intervention not only increased use of effective smoking cessation medications but also increased medication efficacy, likely through mechanisms such as promotion of medication adherence, smoking cessation behavioral group attendance, and repeated efforts to quit while on medication.

Both the CHW and the PE interventions encouraged treatment with varenicline, combined short- and long-acting NRT, and in some cases varenicline together with combination NRT for all participants, regardless of readiness to make a quit attempt right away, extending previous reports of efficacy of an opt-out approach to smoking cessation treatment (50, 51) for this population with serious mental illness, functional impairment, and significant socioeconomic barriers to health care. Overall, 78% of eligible participants enrolled, suggesting broad generalizability, and 65% of those offered CHW support consented to receive it, underscoring the acceptability of CHW support in this population. The intent-to-treat approach employed in this study estimated the efficacy of the intervention at the population level and likely underestimated the abstinence rates if the study design had been restricted to enroll or randomly assign only those who consented to receive CHW support.

The CHW role in this trial was high touch and interactive and emphasized building trusting relationships and gaining knowledge of participants’ home environments early in the intervention, so that CHWs would be positioned to individually tailor behavioral support for smoking cessation. The results exemplify successful application of a physician extender role, maximizing the impact of participants’ interactions with physicians (scarce specialty human resources) through supportive work of individuals with a lower level of training. Proactive outreach via telephone has increased smoking cessation medication use in community mental health populations (16, 52), and future studies should evaluate the effect of combining CHW support with telephone and digital interactions that could increase the reach of CHW support at scale.

We did not find a significant effect of PE without CHW support, even though the trial was conducted when important information about medication safety, efficacy, regulations, and clinical guidelines was newly available (10, 11, 35). Although we reached 91% of assigned clinics with in-person, clinician-delivered educational outreach and 96% of assigned clinics with either in-person contact or printed materials, we often did so in group meetings, which did not give providers an opportunity to express individual biases or knowledge gaps that may have affected treatment provision and did not give detailers an opportunity to address them. The evidence-based model for academic detailing (25) has increased abstinence rates over baseline rates among adults with serious mental illness when delivered in community mental health clinics with decision support and feedback (19). This trial, however, had low uptake of one-on-one meetings between physicians and detailers (33), primarily because of physician time constraints, and we thus draw conclusions on the basis of PE delivered primarily in group settings, which is scalable.

The study had some limitations. First, as expected, substantial loss to follow-up occurred. However, no apparent differential dropout occurred across randomization arms. Seventy-two participants who were lost to follow-up after randomization with insufficient data for imputation were lost because of changes in eligibility criteria for psychiatric rehabilitation services that were independent of the randomization and thus did not bias the study. Second, without a CHW-only arm in the randomized sample, we could not determine whether a CHW intervention would be effective without PE. The intervention was tested in urban settings among individuals with functional impairment related to serious mental illness, making them eligible for psychiatric rehabilitation services. Results may not generalize to individuals with less severe functional impairment or to other health systems. Finally, although CHWs were hired into a psychiatric rehabilitation job category in the widespread psychiatric rehabilitation model of care, implementation at scale will require funding to enable training and caseloads small enough to allow meaningful longitudinal interaction in the community to foster cardiovascular risk reduction and other health goals. Implementation studies are needed in which CHW and tobacco treatment specialist training are provided to psychiatric rehabilitation workers who are part of existing community-based psychiatric rehabilitation care teams.

Conclusions

In this pragmatic trial, a community-based CHW intervention combined with education of providers improved smoking cessation rates in a historically difficult-to-reach population of smokers with serious mental illness who did not necessarily want to quit smoking. Educational outreach to providers alone did not significantly increase abstinence. CHWs provided services that are beyond the scope of primary care and psychiatric providers, and receipt of CHW support increased both the use and the efficacy of first-line, evidence-based smoking cessation medication, particularly varenicline, among individuals with serious mental illness. Future work is needed to determine the benefit of CHW support alone.