Socioeconomic Disparities in Telephone-Based Treatment of Tobacco Dependence

Abstract

Objectives. We examined socioeconomic disparities in tobacco dependence treatment outcomes from a free, proactive telephone counseling quitline.

Methods. We delivered cognitive–behavioral treatment and nicotine patches to 6626 smokers and examined socioeconomic differences in demographic, clinical, environmental, and treatment use factors. We used logistic regressions and generalized estimating equations (GEE) to model abstinence and account for socioeconomic differences in the models.

Results. The odds of achieving long-term abstinence differed by socioeconomic status (SES). In the GEE model, the odds of abstinence for the highest SES participants were 1.75 times those of the lowest SES participants. Logistic regression models revealed no treatment outcome disparity at the end of treatment, but significant disparities 3 and 6 months after treatment.

Conclusions. Although quitlines often increase access to treatment for some lower SES smokers, significant socioeconomic disparities in treatment outcomes raise questions about whether current approaches are contributing to tobacco-related socioeconomic health disparities. Strategies to improve treatment outcomes for lower SES smokers might include novel methods to address multiple factors associated with socioeconomic disparities.

In the United States, the prevalence of daily smoking among lower socioeconomic status (SES) groups is 3 to 4 times higher than that of higher SES groups and a leading contributor to socioeconomic health disparities.^1–5 Comprehensive tobacco control programs can reduce these disparities by providing all smokers with effective treatment for tobacco dependence; however, significant socioeconomic disparities in treatment outcomes are observed in many treatment settings, raising concerns about contributing to or at least maintaining existing disparities with these approaches.^6–14 Treatment delivered through telephone quitlines has become widely available in the United States and the United Kingdom.¹⁵ Proactive quitlines attract a large proportion of lower SES smokers^16–18 and smokers with different demographic and clinical characteristics than in-person, community-based treatments.^16,19,20 Because of their ubiquitous nature and because they appear to be especially accessible and attractive to lower SES smokers,^16,17,18 quitlines have the potential to attenuate tobacco-related disparities; however, if quitlines also demonstrate socioeconomic disparities in treatment outcomes, then this would strengthen concerns about current approaches contributing to or maintaining these disparities.

SES ideally incorporates the social and economic factors that influence what position individuals or groups hold in a societal structure.^21,22 In health research, SES is a broad construct describing relative access to basic resources required to achieve or maintain good health.^23,24 Consistent with leading conceptual models of health disparities,^23–26 SES is empirically related to smoking cessation through complex reciprocal relations among clinical, environmental, and treatment utilization factors including stress, coping resources, psychological factors, exposure to other smokers, and use of treatment resources.^6,27–32

Cognitive–behavioral treatment (CBT) provided through proactive quitlines is a practical innovation that attracts a promising number of lower SES smokers.^15–17,33 Although not targeted to or tailored for lower SES groups, CBT, when delivered appropriately, addresses individuals’ treatment-related clinical characteristics (e.g., stress, coping, dependence level, motivation, self-efficacy, environmental challenges). Nonetheless, significant disparities have been found in CBT treatment outcomes in many tobacco treatment settings.^6–14 Quitlines treat smokers with different characteristics than in-person treatment,^16,19,20 however, and thus might not demonstrate the same disparities as in-person CBT treatment.^6–13

We investigated socioeconomic disparities in tobacco dependence treatment outcomes using data from a proactive quitline in Arkansas in operation from 2005 to 2008. We used statistical modeling of abstinence at the end of treatment (EOT) and 3 and 6 months after treatment to examine the independent contribution of SES to treatment outcomes controlling for other factors. Consistent with findings from community-based treatment, we hypothesized that after accounting for demographic, clinical, environmental, and treatment utilization factors, the lowest SES participants would be least likely to achieve long-term abstinence.

METHODS

All English-speaking smokers who received at least 1 telephone treatment contact from the Arkansas quitline from 2005 to 2008 were included (n = 6626).

Tobacco Dependence Treatment

All behavioral treatment was delivered by telephone. The evidence-based behavioral treatment consisted of manual-driven, individual 20- to 30-minute telephone sessions of CBT scheduled with participants weekly for 6 weeks.^{7,11,16,34–36} A lay overview of the biopsychosocial underpinnings of tobacco dependence was presented, including physiological components (e.g., tolerance and withdrawal) and learning components (e.g., triggers or cues, tobacco use as reinforcement) and use of tobacco to cope with nicotine-related (e.g., lowered plasma nicotine levels) and nicotine-unrelated (e.g., managing stress) events.

Treatment components included setting a quit date, guided rate reduction, self-monitoring, stimulus control, problem solving, conflict management, cigarette refusal training, enhancing social support, goal setting, relapse prevention, and stress management. Treatment session content, tracked separately from treatment contacts (number of telephone contacts with participants in which treatment was provided), progressed sequentially (sessions 1 through 6). We provided free nicotine patches in 2-week increments by mail with physician approval and session attendance. Twelve weeks of patch use was encouraged. Treatment fidelity was addressed with structured treatment protocols, clinical supervision of selected recorded cases, and weekly clinical supervision.

Procedures

The Arkansas quitline was part of a state-funded treatment system that used quitline media promotion, health care provider training,³⁷ a smoke-free workplace program, and a faxed referral program³⁸ to recruit participants into 2 treatment modalities (in person and by telephone) with equivalent treatment content.¹⁶ Tobacco treatment specialists certified by the University of Mississippi Medical Center ACT Center delivered the treatment. Quitline tobacco treatment specialists were located in a call center on the University of Arkansas for Medical Sciences campus.

All data were collected by telephone. Demographic, clinical, and tobacco use data were collected on first contact by intake coordinators who entered data directly into the electronic data management system and scheduled telephone treatment appointments with tobacco treatment specialists. Reminder calls were made the day before all scheduled telephone appointments. At the scheduled time, tobacco treatment specialists contacted participants by telephone, delivered treatment sessions, scheduled the next session, and entered treatment data into structured treatment notes in the electronic data management system. If participants were unavailable at scheduled times, the specialists left messages and attempted recontact within 5 minutes. Intake coordinators attempted to reschedule participants who missed scheduled calls. At least 6 and as many as 10 attempts were made to contact participants to reschedule treatment sessions depending on participant response and circumstances. This system is similar to that used in in-person treatment settings.

Outcome data were collected by outcome assessment interviewers specially trained to interview participants in a standardized manner. The 4-week outcome assessment window opened 2 weeks before the outcome assessment due date and closed 2 weeks after the due date. Two weeks before the window opened, a postcard was sent to participants to remind them of the upcoming assessment call. At least 6 and as many as 10 attempts were made to contact participants for outcome assessment within the 4-week window depending on participant response and circumstances. This system is similar to that used in other settings.³⁹

Measures

Socioeconomic status.

Educational achievement is perhaps the most widely used proxy for SES in the United States because it is strongly associated with future occupation and earning potential^21,22; however, household income is often the best measure of material resources, especially for non–wage-earning family members.^21,22,23,25 Composite indices are more accurate proxies for SES than either of these measures alone.^21,22 We used a composite index of SES using household income and educational level.^6,16,22 Household income was assessed with 6 categories used by the US Census Bureau (< $10 000, $10 000–$14 999, $15 000–$24 999, $25 000–$34 999, $35 000–$49 999, ≥ $50 000). Educational level was assessed with years of completed education grouped into 4 categories (< 12 years of education, 12 years of education, 13–14 years of education, ≥ 15 years of education). Values were assigned to income level (lowest = 1, highest = 6) and educational category (lowest = 1, highest = 4). Adding the income- and educational-level values resulted in a discrete analog SES scale (range = 2–10). This SES composite scale was then collapsed into 3 SES levels: SES1 (2–4), SES2 (5–7), and SES3 (8–10).

Fagerström Test for Nicotine Dependence.

Scores from the Fagerström Test for Nicotine Dependence (FTND), a 6-item measure (range  = 0–10), were grouped into 2 levels of dependence (0–4 = lower, 5–10 = higher).^40,41

Perceived Stress Scale.

Scores from the Perceived Stress Scale–4 (PSS-4), a 4-item measure (range = 0–16), were grouped into 2 levels (0–5 = low, 6–16 = high). According to previous studies, mean levels for smokers range from 4.8 to 5.9.^42,43

Transtheoretical Model stages of change.

The Transtheoretical Model (TTM) provides a framework for assessing readiness to quit in smokers. Smokers in the process of quitting were classified as being in action, preparation (planning to quit in the next 30 days), contemplation (planning in the next 6 months), and precontemplation (not planning in the next 6 months).⁴⁴

Motivation, self-efficacy, and concern about weight gain.

We measured these constructs on a 0 to 10 scale (0 = not at all, 10 = most ever) using the following questions: “How much do you want to quit smoking?” “How confident are you that you can quit using tobacco and stay quit for good?”^45–47 and “How concerned are you about gaining weight after you quit?”⁴⁸ These 11-point scales were grouped into 3 categories (low = 0–4, medium = 5–7, high = 8–10).

Smoking policy at work and at home.

We assessed smoking policies in participants’ workplaces and homes with the following options: (1) no smoking anywhere inside or outside; (2) no smoking inside, but smoking is allowed outside; (3) smoking is allowed in certain areas inside; or (4) smoking is allowed anywhere inside.

Treatment use.

Treatment use was measured with (1) number of contacts in which treatment content was delivered (did not include assessment or intake), (2) amount of treatment content (sessions 1–6), (3) whether nicotine replacement therapy patches were dispensed, (4) number of nicotine replacement therapy patches dispensed, and (5) whether behavioral treatment was completed (completed 5 of 6 sessions of content).

Outcome assessment.

We used smoking status at the last treatment contact as the EOT outcome. Three and 6 months after the EOT, outcome assessment interviewers attempted contact with all participants by telephone and asked, “How many cigarettes are you smoking on a usual day?” followed by “Have you smoked any cigarettes in the past 7 days?” Participants were considered abstinent if they answered “0” and “no,” respectively. Self-reported, 7-day point prevalence is an appropriate, valid, and reliable method for assessing abstinence in this type of program.^49–51

Data analysis.

We analyzed data with PASW version 20 (SPSS Inc., Chicago, IL). We calculated descriptive statistics and conducted analysis of variance and χ² analysis to examine differences among participants among the 3 SES levels and among those contacted and lost to follow-up 6 months after treatment. We used standardized residuals (R = {observed − expected} / square root of expected) to identify sources of significant differences (standardized residual > 2.00) for categorical variables.⁵² Comparisons among SES groups and SES levels for all outcome assessment points were adjusted for the effects related to abstinence in the respective models. Level of significance was set at .05.

We calculated abstinence rates for the three outcome assessment points: EOT and 3 and 6 months after treatment. Consistent with recommendations,^7,53 we calculated abstinence rates using 2 methods to accommodate missing data: (1) modified intention to treat (ITT), simply imputing all participants lost to follow-up as smoking, and (2) complete-case analysis (CCA), eliminating participants lost to follow-up from the analysis.^54,55

We modeled abstinence using logistic regression and generalized estimating equations (GEE) and report outcomes as adjusted odds ratios (AORs). GEE is an extension of logistic regression that incorporates all outcome data points in 1 model, corrects for missing data, corrects standard errors of estimates by a working correlational structure based on observed data, and incorporates the effects of time.^54,56–58 We used continuous variables in all models whenever available. Logistic regression models used EOT, 3-month, and 6-month ITT abstinence as the dependent variables. The GEE model used overall abstinence as the dependent variable. All models accounted for all variables found to differ significantly by SES group listed in Table 1. The GEE model also included time as a within-subjects categorical variable represented by the 3 outcome assessment points (EOT and 3 and 6 months), and a robust covariance matrix estimator and an exchangeable working correlation matrix. In all models, the procedure for selecting variables was the same. Variables were analyzed in 3 blocks:

• Demographics: age, sex, ethnicity, partnered status, work status, and health care insurance status;

• Treatment use: maximum amount of treatment content and completed treatment; and

• Clinical and environmental: partner smoking status, last quit attempt, sought help in past, ready to set a quit date, work supports quitting, smoking policy in the home, smoking policy at work, psychiatric diagnosis, referral type, smokes menthol, cigarettes per day, age started smoking, years smoking, FTND, motivation level, confidence level, and PSS-4 level.

TABLE 1—

Characteristics of Participants and Differences Among the 3 Socioeconomic Status Groups: Arkansas, 2005–2008

Variable and Level, Category, or Range	Sample Size, No.	% or Mean (SD)	SES1 (n = 2430), % or Mean (SD)	SES2 (n = 2468), % or Mean (SD)	SES3 (n = 1069), % or Mean (SD)
Demographic
Age (range = 14–89), y***	6615	43.07 (13.2)	43.9 (13.8)d	42.4 (12.9)d	42.8 (11.5)
Sex***	6626
Mena		30.3	26.0	30.0	42.0
Womena		69.7	74.0	70.0	58.0
Partnered***	6620
Yesa		60.6	47.5	66.8	81.5
Noa		39.4	52.5	33.2	18.5
Ethnicity***	6605
White		80.2	78.1	81.1	84.2
African Americana		15.6	17.7	14.9	11.1
Other		4.2	4.2	3.9	4.7
Education, y	6606
0–41		12.28 (2.3)
< 12		22.8
12		39.6
13–14		25.9
≥ 15		11.7
Work status***	6613
Full or part timea		43.9	22.9	53.2	77.2
Retired		5.5	5.7	5.8	4.5
Disableda		27.2	41.9	19.9	6.5
Unemployeda		12.9	20.2	8.4	3.0
Homemakera		10.5	9.3	12.8	8.9
Household income, $	6382
≤ 10 000		29.6
10 000–14 999		17.4
15 000–24 999		19.2
25 000–34 999		13.2
35 000–49 999		9.9
≥ 50 000		10.6
Health care insurance status***	6602
Medicaid or Medicarea		29.6	45.5	22.3	5.4
Private insurancea		33.5	10.7	40.2	77.4
No insurancea		36.9	43.8	37.5	17.2
SES scaleb
2–10		5.15 (2.2)
SES1		44.4
SES2		38.8
SES3		16.8
Treatment use
Amount of treatment content***	6626
1–6		3.13 (1.9)	3.0 (1.9)d,e	3.2 (2.0)d	3.4 (1.9)e
≤ session 1a		30.9	33.5	29.8	25.0
≤ session 2		16.3	15.7	17.1	16.5
≤ session 3		13.6	13.6	13.0	15.2
≤ session 4		8.7	9.6	7.7	9.1
≤ session 5		9.2	9.4	8.7	9.4
≤ session 6a		21.4	18.3	23.8	24.8
Completed treatment***	6626
Yesa		30.5	27.7	32.5	34.2
No		69.5	72.3	67.5	65.8
Patches dispensed	6626
Yes		29.2	29.5	30.2	26.7
No		70.8	70.5	69.8	73.3
No. of patches dispensed	6626
0–364		16.6 (36.0)	16.5 (37.1)	17.3 (36.4)	15.8 (33.3)
0 patches		70.8	70.5	69.8	73.3
14 patches		6.5	7.4	6.5	4.2
28 patches		4.4	4.6	4.4	3.3
42 patches		4.1	4.0	4.6	3.6
≥ 56 patches		14.2	13.5	14.7	15.5
No. of treatment contacts	6626
1–62		5.10 (5.2)
1		24.6	25.9	24.4	20.7
2		15.2	15.3	14.9	15.7
3		11.3	10.0	12.1	12.4
4		8.4	8.8	8.2	8.5
5		6.7	7.1	5.8	7.9
6		6.5	6.9	5.5	7.9
> 6		27.2	26.0	29.1	26.9
Clinical and environmental
Partner smoking status***	6576
Partner smokesa		38.1	34.7	41.8	40.2
Partner does not smokea		27.5	19.5	29.1	43.9
Last quit attempt**	6576
Nevera		8.7	9.8	8.1	7.0
≤ 30 d ago		13.5	14.3	13.4	12.2
1–3 mo ago		11.0	11.7	10.3	9.9
3–6 mo ago		9.6	9.3	9.2	11.5
> 6 mo ago		57.2	54.9	59.0	59.4
Sought help in past***	6558
Yesa		24.5	20.0	26.5	32.7
Noa		75.5	80.0	73.5	67.3
Ready to set a quit date**	6573
Yes		90.6	89.2	91.3	92.6
Noa		9.4	10.8	8.7	7.4
Smokes menthol cigarettes*	3942
Yesa		30.9	32.0	31.2	25.9
No		69.1	68.0	68.8	74.1
Any use of smokeless tobacco	6588
Yes		1.8	82.2	1.6	1.2
No		98.2	97.8	98.4	98
Work outside home supports quitting***	2840
Yes		76.6	70.5	76.8	81.9
Noa		23.4	29.5	23.2	18.1
Smoking policy at home***	6597
None inside/outside		1.5	1.7	1.1	1.6
None insidea		34.6	26.5	35.6	53.5
Allowed insidea		63.9	71.8	63.4	44.9
Smoking policy at work outside the home***	2934
None inside/outsidea		11.6	10.4	10.3	14.8
None inside		67.4	67.5	66.8	68.3
Allowed insidea		21.0	22.1	22.9	17.0
Psychiatric diagnosis***	5740
Yesa		14.7	18.5	12.8	9.8
No		85.3	81.5	87.2	90.2
Referral type***	6573
Health care providera		46.2	56.2	41.5	31.4
TV or radioa		24.8	18.4	27.8	34.0
Word of mouth		18.2	17.3	19.4	17.8
Print media, Web site, or other		8.0	7.4	8.1	8.8
Workplacea		2.9	0.8	3.2	8.0
Cigarettes per day (range = 0–120)***	6610	24.3 (13.2)	25.2 (13.8)d	24.3 (12.7)d	22.6 (12.5)d
Age started smoking (range = 3–71), y***	6551	16.8 (5.7)	16.2 (5.8)d	17.0 (5.5)d	17.8 (5.4)d
Years smoking (range = 0–67)***	6531	25.42 (13.3)	26.9 (13.7)d,e	24.5 (13.1)d	24.0 (12.0)e
FTND (range = 0–10)***	6479	5.67 (2.3)	6.0 (2.2)d	5.6 (2.2)d	5.0 (2.3)d
Motivation levelc (range = 0–10)***	6569	9.47 (1.2)	9.5 (1.2)d	9.5 (1.1)e	9.3 (1.3)d,e
Confidence levelc (range = 0–10)***	6569	7.59 (2.5)	7.7 (2.5)d	7.6 (2.5)d	7.3 (2.4)d
PSS–4 (range = 0–16)***	6459	7.86 (2.5)	8.2 (2.5)d	7.8 (2.5)d	7.1 (2.5)d
Concern about weight gainc (range = 0–10)	2859	5.00 (4.3)	5.0 (4.4)	4.9 (4.3)	5.1 (4.2)
BMI (range = 12–76)	2443	28.40 (7.3)	28.6 (7.8)	28.4 (7.1)	27.6 (5.8)

Note. BMI = body mass index (defined as weight in pounds times 703 divided by height in inches squared); FTND = Fagerstrom Test for Nicotine Dependence; other = Multiethnic (1.0%), American Indian (2.2%), Hispanic or Latino (0.5%), other (0.4%), or Asian/Pacific Islander (0.2%); partnered = married or living with significant other; PSS–4 = Perceived Stress Scale 4 items; SES = socioeconomic status; SES1 = 2–4; SES2 = 5–7; SES3 = 8–10. Percentages might not add to 100% because of rounding.

^aStandardized residual > 2.0 indicating that the category identified contributed significantly to differences among SES groups.

^bSES was measured with a composite index incorporating values for annual household income and educational level (range = 2 [lowest] to 10 [highest]).

^cMeasured on a scale of 0–10 where 0 = none at all and 10 = most possible.

^d,eSignificant differences among means of the same variable. Differences between means indicated by presence of the same superscript.

*P < .05. **P < .01. ***P < .001.

We manually eliminated variables with significance levels greater than .1 in a stepwise manner within blocks, entered the results from each block analysis into the model, and repeated the elimination procedure for the final model. Although smoking menthol cigarettes was associated with abstinence in preliminary analyses in the logistic regression models, we eliminated it from the clinical and environmental block because a large amount of missing data caused a more than 40% reduction in the overall number of cases, significantly reducing overall representativeness. No special procedures were needed to eliminate menthol use in the GEE model.

RESULTS

Participants (n = 6626) were primarily of lower to middle SES, White, and employed. Although participants were highly motivated and moderately confident about quitting, they were also highly nicotine dependent and smoked, on average, 24 cigarettes per day; had started smoking in mid-adolescence; and had smoked for more than 25 years. Although most were ready to make a quit date, most had not made a quit attempt in more than 6 months, and few had previously obtained professional help with quitting. About one third of participants smoked menthol cigarettes. (Note: Data collection for menthol use was instituted in mid-2006 and was thus collected from just 60% [n = 3942] of participants.) Levels of perceived stress were high. Of those who were employed, many reported that their workplace supported efforts to quit, but 21.0% reported that smoking was allowed inside their workplaces. Nearly two thirds allowed smoking inside their homes, and 38.1% lived with a partner who smoked. Most were referred by health care providers (46.2%), followed by TV or radio advertisements (24.8%) and word of mouth (18.2%). Participants completed a mean of 3.1 (SD = 1.9) sessions of content in a mean of 5.1 treatment contacts, with about one third completing treatment. Nicotine patches were dispensed to 29.2% of participants, with a mean of 16.6 patches dispensed per participant (Table 1). Unadjusted abstinence rates were as follows: EOT, 31.9%; 3 months, 29.3% CCA and 17.7% ITT; and 6 months, 28.6% CCA and 15.5% ITT.

Data were available for 100% of participants at the EOT, for 60.3% at 3 months after treatment, and for 54.3% at 6 months after treatment. Those successfully contacted for outcome assessment were more likely to be of higher SES; to be retired or disabled; to have Medicaid, Medicare, or private health insurance; to have a psychiatric diagnosis; and to have a partner who smoked. Those contacted were more likely to have sought help in the past, to report support from their work, to have smoked for more years, to have less confidence in quitting, to report lower stress levels, and to be referred by a health care provider (Table 2).

TABLE 2—

Differences Among Participants Successfully Contacted and Lost to Follow-Up 6 Months After Treatment: Arkansas, 2005–2008

	Successfully Contacted, Mean (SD) or %	Lost to Follow-Up, Mean (SD) or %
Demographic
Age, y	45.2 (13.0)	40.5 (12.9)
Education: ≥ 15 y	59.6	40.4
Work status
Retired	69.9	30.1
Disabled	58.7	41.3
Unemployed	43.0	57.0
Socioeconomic status scalea	5.3 (2.2)	5.0 (2.1)
SES1	51.6	48.4
SES3	59.1	40.9
Household income, $
≤ 10 000	51.1	48.9
≥ 50 000	61.1	38.9
Health care insurance status
Medicaid/Medicare	57.7	42.3
Private insurance	58.3	41.7
No insurance	47.9	52.1
Treatment use
Amount of treatment content (up to and including)	3.5 (2.0)	2.6 (1.8)
Session 1	41.0	59.0
Session 2	49.3	50.7
Session 4	61.8	38.2
Session 5	62.1	37.9
Session 6	72.0	28.0
Completed treatment: yes	69.0	31.0
Patches dispensed: yes	65.5	34.5
No. of patches dispensed	21.4 (41.1)	10.9 (27.7)
0	49.6	50.4
42	66.4	33.6
≥ 56	70.2	29.8
No. of treatment contacts	6.2 (5.9)	3.8 (3.9)
1	38.9	61.1
2	46.8	53.2
3	46.3	53.7
5	63.1	36.9
6	63.4	36.6
> 6	71.0	29.0
Clinical and environmental
Partner smokes: yes	51.1	48.9
Last quit attempt: never	47.5	52.5
Sought help in past: yes	59.5	40.5
Smokes menthol: yes	47.9	52.1
Work supports quittingb: yes	54.3	45.7
Psychiatric diagnosisc: yes	60.0	40.0
Years smoking	27.4 (13.3)	23.0 (13.0)
Confidence level	7.5 (2.5)	7.7 (2.5)
PSS-4	7.8 (2.5)	7.9 (2.6)
How referred; health care provider	51.3	48.7

Note. PSS–4 = Perceived Stress Scale 4 items. Only variables and categorical groups with significant differences are displayed (α = .05).

^aSocioeconomic status was measured with a composite index that incorporated values for household income and educational level (range = 2–10 [lowest–highest]); SES1 included values from 2 to 4; SES2, from 5 to 7; SES3, from 8 to 10.

^bOf those employed.

^cSelf-reported diagnosed with a psychiatric disorder.

Socioeconomic Differences

The lowest SES group (SES1) was older, less likely to be male or working, more likely to be African American, and less likely to be partnered, but more likely, if partnered, to have a partner who smoked than the 2 higher SES groups. SES1 received less treatment content and were less likely to complete treatment than the higher SES groups. SES1 was more likely to have never made a quit attempt and less likely to have sought help in the past, but more likely to be ready to set a quit date than were the higher SES groups. SES1 was more likely to report that their work did not support quitting and less likely to be protected by smoking restrictions at home or at work. SES1 was twice as likely as SES3 to report a psychiatric diagnosis. SES1 was more likely to smoke menthol, smoke more cigarettes per day, start smoking earlier, smoke for more years, be more heavily dependent, and be more stressed, but was also more confident about quitting and motivated to quit than were the higher SES groups. SES1 was more likely to be referred by a health care provider and less likely to be referred by TV or radio or the workplace than were the higher SES groups (Table 1).

Unadjusted comparisons among SES groups revealed that SES1 was the least likely to be abstinent at every outcome assessment point using ITT and CCA methods. At the EOT, SES1 was 28.3% abstinent, SES2 was 33.3% abstinent, and SES3 was 38.5% abstinent (χ²[2] = 40.9; P ≤ .001). For CCA at 3 months, SES1 was 24.2% abstinent, SES2 was 30.8% abstinent, and SES3 was 37.7% abstinent (χ²[2] = 45.8; P ≤ .001). For ITT at 3 months, SES1 was 14.0% abstinent, SES2 was 18.7% abstinent, and SES3 was 24.6% abstinent (χ²[2] = 63.1; P ≤ .001). For CCA at 6 months, SES1 was 25.1% abstinent, SES2 was 28.9% abstinent, and SES3 was 36.7% abstinent (χ²[2] = 29.2; P ≤ .001). For ITT at 6 months, SES1 was 12.9% abstinent, SES2 was 16.0% abstinent, and SES3 was 21.7% abstinent (χ²[2] = 45.9; P ≤ .001).

Modeling Abstinence

In the logistic regression models, with demographic, clinical, environmental, and treatment use factors accounted for, SES was not significantly associated with EOT abstinence, but it was significantly associated with long-term abstinence. The EOT model retained 75.6% of participants (n = 5010). The odds of achieving abstinence at the EOT increased by a factor of 1.02 (95% CI = 0.98, 1.07; P = .38) for every 1-unit increase in the SES scale. Adjusted comparisons revealed no significant differences among SES groups (SES1/SES3: AOR = 1.12, 95% CI = 0.87, 1.46; SES2/SES3: AOR = 1.06, 95% CI = 0.93, 1.21) and between the highest and the lowest SES level. The odds of abstinence for participants at the highest SES level (SES = 10) were 1.18 times those of the participants at the lowest SES level (SES = 2; 95% CI = 0.81, 1.72).

The 3-month model retained 90.9% of participants (n = 6024). The odds of achieving abstinence 3 months after treatment increased by a factor of 1.08 (95% CI = 1.04, 1.11; P ≤ .001) for every 1-unit increase in the SES scale. Adjusted comparisons among SES groups revealed significant differences (SES1–SES3: AOR = 1.51; 95% CI = 1.25, 1.81; SES2–SES3: AOR = 1.24; 95% CI = 1.12, 1.36). The odds of abstinence 3 months after treatment of the participants at the highest SES level (SES = 10) were 1.81 times those of the participants at the lowest SES level (SES = 2; 95% CI = 1.38, 2.36).

The 6-month model retained 79.5% of participants (n = 5258). The odds of achieving abstinence 6 months after treatment increased by a factor of 1.06 (95% CI = 1.02, 1.10; P = .004) for every 1-unit increase in the SES scale. Adjusted comparisons among SES groups revealed significant differences (SES1–SES3, AOR = 1.37; 95% CI = 1.10, 1.70; SES2–SES3, AOR = 1.18; 95% CI = 1.05, 1.32). The odds of abstinence 6 months after treatment of the participants at the highest SES level (SES = 10) were 1.58 times those of the participants at the lowest SES level (SES = 2; 95% CI = 1.15, 2.16).

The GEE model of overall abstinence retained 81.2% of participants (n = 5382). The odds of abstinence increased by a factor of 1.07 (95% CI = 1.05, 1.10; P ≤ .001) for every 1-unit increase in the SES scale. The odds of abstinence for the highest-SES participants were 1.75 times those of the lowest SES participants (95% CI = 1.44, 2.13). The final GEE model results are presented in Table 3.

TABLE 3—

Model of Overall Abstinence, Generalized Estimating Equations Accounting for Socioeconomic Differences in Demographic, Clinical and Environmental, and Treatment Utilization Factors: Arkansas, 2005–2008

Variable and Category	AORb (95% CI)
Socioeconomic statusa,***	1.07 (1.05, 1.10)
Demographic
Sex**
Female	0.85 (0.76, 0.95)
Male (Ref)	1.00
Treatment use
Amount of treatment contenta,***	1.69 (1.58, 1.80)
Completed treatment*
Yes	1.28 (1.01, 1.61)
No (Ref)	1.00
Clinical and environmental
Cigarettes per daya,***	0.99 (0.98, 1.00)
Fagerstrom Test for Nicotine Dependencea,***	0.94 (0.91, 0.97)
Confidence levela,***	1.06 (1.03, 1.08)
Motivation levela,**	1.07 (1.02,1.12)
Psychiatric diagnosis**
Yes	0.83 (0.71, 0.96)
No (Ref)	1.00
Time***
6 mo after treatment	0.31 (0.28, 0.34)
3 mo after treatment	0.37 (0.34, 0.41)
End of treatment (Ref)	1.00
Intercept***	0.03 (0.02, 0.06)

Note. CI = confidence interval; AOR = adjusted odds ratio.

^aA difference of +1 in the measure increases the odds of abstinence by the AOR.

^bWith all other factors accounted for in the model.

*P < .05. **P < .01. ***P < .001.

DISCUSSION

We found significant socioeconomic disparities in long-term treatment outcomes among smokers who received telephone treatment through a free, proactive quitline. We found no significant disparity in treatment outcomes at the EOT, similar to in-person treatment,⁶ but 3 to 6 months after treatment, the odds of abstinence for the highest SES participants were significantly greater than the odds of abstinence for the lowest SES participants. CBT for tobacco dependence delivered over the telephone through quitlines appears to be particularly attractive to smokers of lower SES; however, these findings raise concerns that current approaches contribute to or at least maintain the existing significant socioeconomic disparities in treatment outcomes despite improvements in access to care. Because tobacco-related disparities are a significant contributor to socioeconomic health disparities, this is a significant concern to public health.

Consistent with conceptual and empirical models and with previous research,^6,23–26 SES was significantly associated with numerous demographic, clinical, environmental, and treatment use differences associated with abstinence (Table 1). Compared with higher SES participants, lower SES participants reported significantly higher levels of stress and nicotine dependence and a higher frequency of current psychiatric diagnoses, placing higher demands on coping resources as well as limited material resources. Lower SES participants were also less likely than higher SES participants to be protected by smoking policies at home and at work. Although we found no significant differences among SES groups in the use of nicotine patches and number of treatment sessions, lower SES participants were exposed, on average, to less treatment content—a factor that is highly associated with treatment outcomes.^6,16,59 In our models, these differences were accounted for and significant disparities in treatment outcomes remained, indicating that there is much we don’t yet understand about how these disparities emerge. The findings were similar between the logistic regression and the GEE models, reinforcing our confidence in the findings.

There are several similarities and differences among these findings and the findings of our in-person, community-based study.⁶ Using logistic regression, the odds of abstinence 6 months after treatment increased by a factor of 1.06 (95% CI = 1.02, 1.10) for every 1-unit increase in the SES scale in this study compared with 1.12 (95% CI = 1.06, 1.18) in the in-person treatment study. With all other factors accounted for, the model in the in-person study retained some of the same variables as this study (i.e., cigarettes per day, confidence level) but also retained variables that were not retained in this study (i.e., smoking policy in the home, referral source, stress level). The final GEE model in this study retained variables that were not retained in the in-person study (i.e., sex, psychiatric diagnosis, FTND, motivational level, and time). Amount of treatment content, a highly significant and modifiable factor, was retained in all the models in both studies. We speculate that the differences among models are derived from differences in the characteristics of the participants as well as differences in the statistical procedures used to develop the models (logistic regression vs GEE). Nonetheless, these findings highlight the importance of identifying common factors, such as amount of treatment content, to serve as modifiable prognostic indicators and potential targets for enhancing long-term treatment outcomes among all treatment modalities.

Efforts to reduce tobacco use and eliminate tobacco disparities would benefit from identifying factors associated with disparities in tobacco treatment outcomes. Given the manner in which the disparity emerges in the months after treatment, perhaps treatment that specifically targets posttreatment clinical and environmental challenges encountered by lower SES groups might improve long-term effectiveness for lower SES smokers, reduce tobacco disparities, and improve the effectiveness of comprehensive tobacco control programs. In addition to treatment content, targets for reducing disparities in treatment outcomes might include addressing psychiatric issues, particularly the most prevalent mood and anxiety disorders, and addressing issues particular to women as well as reducing dependence levels and increasing motivation and confidence levels before treatment. Our findings here and elsewhere⁶ suggest that bolstering treatment with innovative methods delivered over a longer period of time might demonstrate a positive impact on lower SES long-term treatment outcomes.

This study used a rich data set from a diverse group of smokers. Nonetheless, the Arkansas quitline differed from other quitlines from 2005 to 2008 in that it did not limit enrollment by readiness to quit and provided a more comprehensive and standardized treatment than that provided by most state-sponsored quitlines.¹⁵ Although the rates of participants lost to follow-up were similar to or better than those in community-based and other telephone treatment studies,^{6,8,14,51,60,61} the interpretations are limited by the number of participants lost to follow-up. Lack of biochemical validation of abstinence is a limitation as well; however, studies have suggested low rates of deception,⁵¹ and self-report is acknowledged as an appropriate assessment method in a study of this type.⁴⁹ Our outcomes can thus be viewed with a reasonable degree of confidence. Finally, lower SES smokers in this study should not be assumed to represent all lower SES smokers, some of whom do not have the resources to access telephone or in-person treatment.

We found significant socioeconomic disparities in long-term treatment outcomes among smokers who received telephone treatment through a free, proactive quitline. The odds of long-term abstinence for the highest SES participants were significantly greater than the odds of abstinence for the lowest SES participants in both the logistic regression and GEE models. This disparity wasn’t apparent immediately after treatment, but emerged in the 6 months after treatment. CBT for tobacco dependence delivered over the telephone through quitlines is particularly attractive to many lower SES smokers; however, these findings raise concerns that current approaches contribute to or at least maintain disparities in treatment outcomes despite improvements in access to care. Tobacco-related disparities are a significant contributor to socioeconomic health disparities and this is a significant concern for tobacco control efforts and public health.