Smokers identified from the electronic health record (EHR) were randomly allocated at a ratio of roughly 1:1 to intervention or control status. The number of individuals who consented served as the denominator for our analyses.
Primary outcome was intention to treat; sensitivity analysis used multiple imputation of both baseline characteristics and the outcome variable. OR indicates odds ratio.
Odds ratios (95% CIs) compared with controls are displayed for selected subgroups. There were no significant interactions between subgroups and intervention status.
Haas JS, Linder JA, Park ER, Gonzalez I, Rigotti NA, Klinger EV, Kontos EZ, Zaslavsky AM, Brawarsky P, Marinacci LX, St Hubert S, Fleegler EW, Williams DR. Proactive Tobacco Cessation Outreach to Smokers of Low Socioeconomic StatusA Randomized Clinical Trial. JAMA Intern Med. 2015;175(2):218–226. doi:10.1001/jamainternmed.2014.6674
Widening socioeconomic disparities in mortality in the United States are largely explained by slower declines in tobacco use among smokers of low socioeconomic status (SES) than among those of higher SES, which points to the need for targeted tobacco cessation interventions. Documentation of smoking status in electronic health records (EHRs) provides the tools for health systems to proactively offer tobacco treatment to socioeconomically disadvantaged smokers.
To evaluate a proactive tobacco cessation strategy that addresses sociocontextual mediators of tobacco use for low-SES smokers.
Design, Setting, and Participants
This prospective, randomized clinical trial included low-SES adult smokers who described their race and/or ethnicity as black, Hispanic, or white and received primary care at 1 of 13 practices in the greater Boston area (intervention group, n = 399; control group, n = 308).
We analyzed EHRs to identify potentially eligible participants and then used interactive voice response (IVR) techniques to reach out to them. Consenting patients were randomized to either receive usual care from their own health care team or enter an intervention program that included (1) telephone-based motivational counseling, (2) free nicotine replacement therapy (NRT) for 6 weeks, (3) access to community-based referrals to address sociocontextual mediators of tobacco use, and (4) integration of all these components into their normal health care through the EHR system.
Main Outcomes and Measures
Self-reported past-7-day tobacco abstinence 9 months after randomization (“quitting”), assessed by automated caller or blinded study staff.
The intervention group had a higher quit rate than the usual care group (17.8% vs 8.1%; odds ratio, 2.5; 95% CI, 1.5-4.0; number needed to treat, 10). We examined whether use of intervention components was associated with quitting among individuals in the intervention group: individuals who participated in the telephone counseling were more likely to quit than those who did not (21.2% vs 10.4%; P < .001). There was no difference in quitting by use of NRT. Quitting did not differ by a request for a community referral, but individuals who used their referral were more likely to quit than those who did not (43.6% vs 15.3%; P < .001).
Conclusions and Relevance
Proactive, IVR-facilitated outreach enables engagement with low-SES smokers. Providing counseling, NRT, and access to community-based resources to address sociocontextual mediators among smokers reached in this setting is effective.
clinicaltrials.gov Identifier: NCT01156610
Although tobacco use in the United States has declined,1 substantial socioeconomic and racial and/or ethnic disparities in smoking prevalence, risk of addiction, and tobacco-related disease remain.2,3 Despite relatively similar rates of tobacco use, blacks experience a higher burden of tobacco-related disease, particularly lung cancer, than whites.3 Importantly, smokers of low socioeconomic status (SES) and/or of a racial or ethnic minority also have more difficulty quitting for several reasons, including more limited access to treatment,4- 7 misinformation about risks and benefits of nicotine replacement therapy (NRT),8 lack of social support,9 neighborhood disadvantage,10 discrimination,11 and other life stressors.12,13 Widening socioeconomic disparities in mortality in the United States are largely explained by slower declines in tobacco use among low-SES smokers than among higher-SES smokers.14,15
While the majority of smokers visit a primary care clinician (PCC) each year, PCCs do not have adequate time or training to provide tobacco treatment16; therefore, it is important to offer systematic opportunities for tobacco treatment beyond the clinician’s office. The broad dissemination of electronic health records (EHRs), with coded data about smoking status, represents an important tool to help reach out to smokers.17,18 Another important tool, interactive voice response (IVR), is a telephone technology that allows a computer to detect voice responses during a call and so provides an efficient way to proactively reach large populations, such as patients identified in the EHRs as smokers. The IVR scripts can be translated into other languages, facilitating outreach to diverse populations. Researchers have used IVR as part of multicomponent smoking cessation programs to provide reminders and to facilitate or sustain treatment delivery.19- 21 This technology can also be used to engage smokers by providing direct linkage to a tobacco treatment specialists (TTS) and other resources.
Despite growing disparities in tobacco use and tobacco-related disease, few trials have specifically examined smoking cessation interventions in low-SES populations.22 Because of the substantial burden of tobacco in these populations, the objective of the present study was to develop and evaluate a proactive approach to tobacco treatment for low-SES smokers that addressed broader sociocontextual mediators of tobacco use. While conceptual models of smoking cessation stress the importance of addressing the broader context of smoking,23 we do not know of other empirical studies that have incorporated referrals to community resources as part of a cessation program. The intervention was designed so that it could be incorporated into the health system through IVR outreach and EHR documentation.
The protocol for the present study, Project CLIQ (Community LInk to Quit), was reviewed and approved by the institutional review board (IRB) of Partners HealthCare. The IRB waived written informed consent and instead approved review of a letter describing the study and how to opt out. The baseline IVR call briefly reviewed the study purpose and protocol; oral, implied consent was presumed if subjects completed the IVR call. Participants were informed that they could withdraw from participation at any time.
We undertook this prospective, randomized clinical trial (RCT) enrolling low-SES smokers to compare usual care delivered by a patient’s health care team with a proactive treatment program that included (1) a series of telephone-based motivational counseling calls with a TTS based in the health care system, (2) access to free NRT patches, (3) personalized, community-based referrals to reduce sociocontextual mediators of tobacco use, and (4) integration of all these components into the participants’ normal health care through updated documentation in the EHR system. The EHR system was used to identify low-SES smokers who described their race and/or ethnicity as white, black, or Hispanic; IVR was used for recruitment.
Smokers were recruited from 13 primary care practices affiliated with Partners HealthCare, a large health care delivery system in eastern Massachusetts. Participating primary care practices included 6 community health centers, 2 community-based practices, 4 hospital-based practices, and 1 medical home. These sites did not have on-site tobacco treatment programs during the study period, but clinicians could refer patients to the Massachusetts Smokers’ Helpline (http://makesmokinghistory.org). Practices shared a web-based, fully functional EHR that allowed coded documentation of smoking status; this allowed identification of smokers by querying a data repository, thus providing an opportunity for outreach. These tools and services provided the basis for “usual care” for tobacco treatment in these practices.
Persons eligible for study participation were adults (age ≥18 years) who visited a PCC at a participating clinic in the month before outreach. Eligibility also required smoker coding in the EHR, race and/or ethnicity documented as white, black, or Hispanic, English or Spanish language preference noted, and residence in a low (<$45 000) or moderate ($45 000-$67 050) median household income census tract. These income thresholds were chosen to reflect the income characteristics of Massachusetts. In the 2008-2012 period, the median household income in Massachusetts was $66 658.24 These income thresholds also divided our population into thirds; the highest income group (>$67 050) was not eligible. We used ArcMap 10.0 (Esri) to geocode participants’ mailing addresses to append median household income estimates based on 2010 census tract as a proxy for socioeconomic status, since income is not captured in the EHR.25 We targeted individuals shortly after a clinic visit to ensure current documentation of smoking status and contact information. Because of the proactive, population-based design, we reached out to all individuals who met these eligibility criteria regardless of their interest in quitting. We excluded patients without a telephone number.
Recruitment for this study occurred between November 7, 2011, and June 3, 2013. Every 2 weeks, the EHR generated a list of individuals who met the eligibility criteria. Within 1 month of a PCC visit, EHR-identified smokers received an informational letter that described the study and included a toll-free telephone number to call if they wished to opt out. Patients in both study groups received identical mailings. We contacted patients who did not opt out within 2 weeks using an IVR script that called up to 15 times over a 2-week period on different days of the week and times of the day. Patients could choose to hear the IVR script in English or Spanish. The IVR platform did not leave messages. If all initial call attempts failed to make contact, patients remained eligible for an additional attempt at contact 4 months later. Participants were recruited solely by automated telephone scripts.
Smokers identified through EHRs were randomly allocated to intervention or control status at roughly a 1:1 ratio. Randomization was performed in batches based on the date of the clinic visit. The first patient randomized in each batch was randomized to intervention status; batches with an odd number of participants therefore resulted in an imbalance in the size of the intervention and control groups (Figure 1). Randomization occurred before consent for logistical reasons (IVR caller could not randomize in real time).
After patients who answered the IVR call confirmed their identity, they heard a standard informed consent script. All individuals who agreed to participate then completed a brief smoking history. We excluded patients who reported that they had not smoked even a puff of a cigarette in the prior week or who had not smoked at least 100 cigarettes in their lifetime (ie, they were considered to be a nonsmoker). The baseline IVR call was identical for intervention and control groups until the final question. After completion of the smoking history, participants in the intervention group were asked if they wanted to speak with a TTS for advice about how to quit smoking and to receive a free 6-week supply of NRT patches. Participants in the control group were not offered treatment.
We designed the intervention based on the Chronic Care Model26,27 and the Social Contextual Model for Reducing Tobacco Use.23 The IVR system sent an automated e-mail to the TTS when a participant requested contact. Participants in the intervention group were eligible for up to 4 counseling calls (a total of approximately 75-100 minutes for all calls over 8-10 weeks), with additional calls scheduled between the participant and TTS at the request of the participant. The TTS used motivational interviewing techniques, which help a participant resolve ambivalence about behavior change regardless of readiness to quit.28 The counseling calls included standard content as well as content tailored to the individual based on intent and confidence to quit, and participants could select optional modules based on their needs (eg, stress, weight gain, menthol use).
The TTS encouraged participants to receive a 6-week supply of free NRT patches. Participants who smoked 10 or more cigarettes per day were offered a 2 weeks’ supply of 21-mg/d patches, 2 weeks of 14-mg/d patches, and 2 weeks of 7-mg/d patches; participants who smoked fewer than 10 cigarettes per day offered 4 weeks of 14-mg/d patches and 2 weeks of 7-mg/d patches. The TTS instructed participants on the use of NRT and addressed knowledge gaps around safety and efficacy.
To address the sociocontextual mediators of tobacco use, the TTS also encouraged participants to receive a personalized referral from a community resource database, HelpSteps.com, a web-based referral system for the greater Boston area. HelpSteps.com is designed to help users select referrals to over 1700 local health and human service agencies categorized into 13 social resource domains (eg, food, education, employment) and has resources related to over 90 different types of services (eg, literacy classes, domestic violence hotline).29,30 At each call, the TTS encouraged participants to pick 2 domains and then selected referrals near the participant’s address and mailed a personalized referral document to the participant. Costs for these referrals, if any, were determined by the particular agency or service and were paid for by the participants (ie, not paid for by the study). The majority of these resources are free. Finally the TTS coordinated additional treatment needs (eg, prescription medications) with the participant’s PCC. We tracked use of intervention components (speaking with the TTS, use of NRT, request for and use of a HelpSteps.com referral).
Our primary outcome measure was self-reported 7-day tobacco abstinence at 9 months following randomization.31 We assessed outcomes by IVR call with live follow-up of nonrespondents (52.9% of outcome calls were completed by IVR with the remainder completed by blinded study staff). Participants were considered to have quit smoking if they responded “no” to the question “Have you smoked a cigarette, even a puff, in the past 7 days?” The IVR platform automatically sent updated information about smoking status to the EHR for the clinician’s review. We also obtained self-reported information about the use of tobacco treatment during the study period by asking: “Did you use any counseling over the phone or in person to help you to quit smoking?” and “Have you used any nicotine products to help you to quit smoking, like the patch, gum, or inhaler?” Participants in both the intervention and the control group who completed the outcome assessment were eligible for a monthly drawing for 1 of 2 $100 gift cards.
The primary analysis used an intent-to-treat approach. We compared participants’ characteristics by group using 2-sample t tests, Wilcoxon tests, and χ2 tests. For the primary analysis, we assumed that nonrespondents at follow-up were smokers. We conducted a sensitivity analysis using multiple imputation of both baseline characteristics and the outcome variable using the FCS option (fully conditional specification) in SAS PROC MI (version 9.3; SAS Institute Inc).32 We explored the effect of the intervention in subgroups based on sex, race and/or ethnicity, SES, and baseline tobacco use. To assess whether there was a differential effect of the intervention for subgroups based on race and/or ethnicity and SES, we constructed logistic regression models that included interaction terms between variables representing the subgroups and intervention status.
During the enrollment period, we attempted to contact 8544 individuals identified by EHR data as smokers (Figure 1). Of these, 455 were not eligible for the study (348 did not have a valid telephone number or address, and 107 reported that they were not a smoker when reached). Of the 8089 potentially eligible adults, 5008 (61.9%) were never reached by the IVR system (no answer and/or answering machine to all of the call attempts) and 2374 (29.3%) declined participation. Overall, 707 (8.3% of those eligible) agreed to participate; 66% of participants completed the outcome assessment call 9 months after enrollment.
Overall, the median age of the participants was 50 years; 68% were women, 20% self-reported Hispanic ethnicity, and 28% black; 36% reported their highest level of educational attainment to be high school or less, and 35% had Medicaid (Table 1). Common comorbidities included hypertension (10%), depression or anxiety (8%), high cholesterol levels (5%), and diabetes (5%). At enrollment, participants were daily smokers, a median of 15 cigarettes per day; 88% reported smoking within 30 minutes of waking, and 77% said that they planned to quit in the next 30 days. The majority had made a quit attempt within the prior year. There were no significant differences between the enrolled intervention and control smokers in demographic, comorbidity, or baseline tobacco use characteristics. Demographic information from the EHR allowed us to compare the participants with nonparticipants: compared with participants, nonparticipants were younger (median age, 47 years), less likely female (54.9%), more likely Hispanic (24.7%), and less likely black (17.1%) (P < .001 for all comparisons).
Individuals in the intervention group were significantly more likely to report quitting at the time of outcome assessment than individuals in the control group (Figure 2). In the primary intention-to-treat analysis, 17.8% of intervention participants quit compared with 8.1% in the control group (P < .001; odds ratio [OR], 2.5; 95% CI, 1.5-4.0). The number needed to treat (NNT) was 10. The sensitivity analysis, using multiple imputation, confirmed these findings, with quit rates of 26.8% and 12.4% respectively (P < .001; OR, 2.6; 95% CI, 1.7-3.9; NNT, 7). Intervention participants were more likely than control participants to respond affirmatively that they had used any counseling for tobacco treatment during the follow-up period (49.6% vs 8.4%; P < .001) or had used NRT (63.6% vs 41.8%; P < .001).
We examined the odds of quitting for intervention vs control participants for subgroups specified by demographic characteristics and baseline tobacco use (Figure 3). Women (OR, 3.0; 95% CI, 1.6-5.4; P < .001), blacks (OR, 3.9; 95% CI, 1.4-10.8; P = .001), whites (OR, 2.1; 95% CI, 1.2-3.9; P = .02), individuals with more than a high school education (OR, 2.7; 95% CI, 1.4-5.0; P = .002), and those who lived in a low-income census tract (OR, 3.5; 95% CI, 1.7-7.2; P = .001) were all significantly more likely to quit in the intervention vs control group. The odds of Hispanics quitting in the intervention vs control groups were not significant. The intervention was effective in all baseline tobacco use subgroups, whether or not the participant planned to quit in the next 30 days. There were no significant interactions between intervention status and any of the demographic or baseline smoking characteristics.
Among individuals in the intervention group, we examined use of each intervention component (ie, speaking to the TTS, receiving NRT, request or use of a community referral) and whether use of a specific component of the intervention was associated with quitting (Table 2). Of those in the intervention arm (n = 399), 274 (68.7%) spoke with the TTS at least once. Of these 274 individuals, 79.6% received NRT; 46.7% requested a HelpSteps.com referral; and 20.1% reported using this referral. The domains most commonly requested for community resources included physical activity (n = 379), educational opportunities (n = 235), and job counseling (n = 69).
Among individuals in the intervention arm, those who spoke to the TTS were more likely to quit than those who did not (21.2% vs 10.4%; P = .01). There was no difference in quitting by use of NRT. Quitting status did not differ by a request for a community referral, but individuals who reported using their referral for the community resource were much more likely to quit than those who did not (43.6% vs 15.3%; P < .001). In a multivariate model that included the intervention components, the only intervention component associated with quitting was use of a community referral (OR, 5.4; 95% CI, 2.5-12.0).
Project CLIQ, one of the first RCTs that specifically targets low-SES smokers, demonstrates that a proactive, systematic, telephone-based intervention including counseling, NRT, and referrals to community resources to address sociocontextual mediators of tobacco use doubles smoking cessation rates compared with usual care for this population with access to primary care. Among participants, use of counseling and the community referrals were important components of this intervention; referral to community resources was one of the novel aspects of this treatment program.
We designed our intervention to take advantage of effective strategies for smoking cessation and address individual sociocontextual mediators that promote tobacco use. First, we recruited individuals from primary care. While 70% of smokers see a PCC each year, minority and low-SES smokers are less likely to report tobacco counseling.4,5 The use of a common EHR allowed us to identify smokers outside of the context of a visit, independent of their interest in quitting.
Second, the intervention used automated calls to provide a systematic, proactive outreach to link interested smokers to treatment. Third, the telephone-based program allowed for the delivery of established effective treatments, including personalized counseling and subsidized NRT.33
Finally, our intervention was designed to consider the sociocontextual challenges to tobacco cessation faced by minority and low-SES smokers.34 Among participants, community resources for physical activity, educational opportunities, and job counseling were the most common referrals requested, and use of these referrals was associated with quitting among participants, which suggests that attempts to address the broader life experiences of smokers was an important component of our intervention. Our findings suggest that the use of these community resources was one of the important mediators of the effectiveness of our program.
For the 8554 registry-identified smokers in our clinics, the design and implementation of this intervention was estimated to cost $283 023 ($33.13 per registry-identified smoker).
The rates of tobacco abstinence in the present study were similar to those measured in other proactive treatment programs not specifically targeting low-SES smokers35 and higher than those found in interventions involving visit-based clinician advice for smoking cessation.36 The Veterans Victory over Tobacco Study (VVTS),37 a pragmatic RCT, used a registry of current smokers at 4 Veterans Affairs medical centers to test a proactive outreach with either telephone or in-person cessation services compared with usual care. At 12 months after randomization, the 6-month abstinence rate was 13.5% in the outreach group vs 10.9% in the usual care group. The VVTS Project CLIQ had a similar proportion of black participants, but VVTS included few Hispanics, and most VVTS participants were men. Project CLIQ extends the result of a previous study38 in which smokers from 1 health center were proactively contacted by mail to offer telephone counseling and free NRT; that study found increased self-reported quit rates at 3-month follow-up compared with usual care (5.3% vs 1.1%).38
Electronic health records allow for identification and management of a population of smokers in a health system.17,39 Coupled with IVR or other technology for systematic outreach (eg, texting), this infrastructure allows for the possibility of large-scale linkage to care teams for tobacco treatment.40 Systematic intervention facilitated by EHRs may be particularly important for low-SES smokers who experience substantial barriers to treatment.19,41 While only 8.3% of the potentially eligible population in Project CLIQ responded to the proactive outreach, this is not surprising because only a minority of smokers may be interested in hearing about cessation at any given time, and there are difficulties inherent in reaching people by telephone (eg, screening of calls, intermittent service particularly among low-SES groups). Our participation rate was higher than that of quitlines,42 which achieve, at best, 1% to 2% reach into populations.
Population-based approaches may reach a relatively small proportion of smokers, but settings with a large number of smokers will still realize a large number of quits. Future implementation of this type of program could consider an ongoing schedule of outreach informed by qualitative work to understand how to better engage men, younger participants, and Hispanics, the groups that proved less likely to participate in this study. This model could be generalized to other health systems with EHRs, which are increasingly promoted to improve the safety and quality of health care.
This study has several limitations. It was a pragmatic trial, conducted within a primary care network. However, this is also a strength in that our findings speak more directly to clinical effectiveness. Our randomization protocol led to an imbalance of the group sizes; despite this limitation, the groups were well balanced on measured characteristics.
Our outcome was measured by self-report and was not biochemically verified, an approach similar to other population-based interventions.31 Project CLIQ was designed to be generalizable to clinical care where clinicians use self-reported smoking status to guide treatment and risk assessment.
While our outcome survey completion rate was good (66%), there is the potential for nonresponse bias. Because we had a higher outcome assessment rate in the control group, the assumption that nonrespondents were smokers would bias against finding an effect. Our sensitivity analysis also suggests that our findings are robust.
Our intervention was designed to address barriers to smoking cessation experienced by low-SES participants, but we needed to use a census tract–based proxy for income assessment. Electronic health records should include measures of individual SES to enable interventions to reduce health disparities. Widening socioeconomic disparities in tobacco use and mortality support the importance of the broad collection of individual socioeconomic measures in EHRs to support interventions designed to promote health equity.14
Project CLIQ demonstrates that proactive, systematic, telephone-based interventions to provide counseling, pharmacotherapy, and access to community-based resources to address the social context of smoking can promote tobacco cessation in disadvantaged populations. Interventions to reduce tobacco use for these populations may reduce disparities in preventable deaths in the United States, an important public health goal.
Accepted for Publication: August 10, 2014.
Corresponding Author: Jennifer S. Haas, MD, MSc, Division of General Medicine and Primary Care, Brigham and Women’s Hospital, 1620 Tremont St, Boston, MA 02120 (firstname.lastname@example.org).
Published Online: December 15, 2014. doi:10.1001/jamainternmed.2014.6674.
Author Contributions: Dr Haas had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Haas, Linder, Park, Rigotti, Klinger, Kontos, Zaslavsky, Williams.
Acquisition, analysis, or interpretation of data: Haas, Linder, Gonzalez, Klinger, Kontos, Zaslavsky, Brawarsky, Marinacci, St Hubert, Fleegler, Williams.
Drafting of the manuscript: Haas, Linder, Park, Gonzalez, Klinger, Brawarsky.
Critical revision of the manuscript for important intellectual content: Haas, Linder, Rigotti, Klinger, Kontos, Zaslavsky, Marinacci, St Hubert, Fleegler, Williams.
Statistical analysis: Linder, Zaslavsky, Brawarsky, Marinacci.
Obtained funding: Haas, Linder, Kontos, Williams.
Administrative, technical, or material support: Haas, Linder, Gonzalez, Klinger, Kontos, Marinacci, St Hubert, Fleegler, Williams.
Study supervision: Haas, Park, Rigotti.
Conflict of Interest Disclosures: Dr Rigotti receives royalties from UpToDate and serves as an unpaid consultant to Pfizer outside the scope of the present work. Dr Fleegler is project director for HelpSteps.com. No other conflicts are reported.
Funding/Support: This work was conducted with support from the Lung Cancer Disparities Center at the Harvard School of Public Health (funded by National Cancer Institute grant P50 CA148596) and the Harvard Catalyst and from the Harvard Clinical and Translational Science Center (funded by National Institutes of Health [NIH] grant 1 UL1 RR 025758-01 and financial contributions from participating institutions). In addition, Drs Haas and Rigotti and Mss Klinger and St Hubert report receiving grants from the NIH during the conduct of this study. Dr Fleegler reports receipt of a Home for Little Wanderers Service Grant and grants from the Aerosmith Endowment Fund for Prevention and Treatment of AIDS and HIV Infections, the Verizon Foundation, Thrive in 5/Boston Mayor’s Office Ready Families Program, the Highland Street Foundation, and the Boston Foundation to fund the initial development of HelpSteps.com, for which Dr Fleegler serves as project director. Dr Fleegler also receives ongoing support from Boston Children’s Hospital and the Boston Public Health Commission.
Role of the Funder/Sponsor: The funding institutions had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclaimer: The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of Harvard University and its affiliated academic health care centers or the NIH.
Additional Information: HelpSteps.com is freely available on the web for general use. There are no intended patents.
Correction: This article was corrected on March 25, 2015, to update the cost of the design and implementation of the intervention.