Educational attainment & quitting smoking: A structural equation model approach

Preventive Medicine 116 (2018) 32–39

Contents lists available at ScienceDirect

Preventive Medicine journal homepage: www.elsevier.com/locate/ypmed

Educational attainment & quitting smoking: A structural equation model approach☆,☆☆ Ann Goding Sauera, a b

⁎,1

T

, Stacey A. Fedewaa,1, Jihye Kimb, Ahmedin Jemala, J. Lee Westmaasa

Intramural Research Department, American Cancer Society, 250 Williams Street NW, Atlanta, GA 30303, United States of America Bagwell College of Education, Kennesaw State University, 580 Parliament Garden Way, Kennesaw, GA 30144, United States of America

A R T I C LE I N FO

A B S T R A C T

Keywords: Cessation Tobacco use Smoking Vulnerable populations Socioeconomic factors

In the United States, disparities in smoking prevalence and cessation by socioeconomic status are well documented, but there is limited research on reasons why and none conducted in a national sample assessing multiple potential mechanisms. We identified smoking and cessation-related behavioral and environmental variables associated with both educational attainment and quitting success. We used a structural equation model of crosssectional data from respondents ≥25 years from the United States 2010–2011 Tobacco Use Supplement-Current Population Survey. Quitting success was defined as former (n = 2607) versus continuing smokers (n = 7636); categories of educational attainment were ≤high school degree, some college/college degree, and advanced degree. Results indicated that using nicotine replacement therapy (NRT) > 1 month and having a home smoking restriction were associated with both educational attainment and quitting success. Those with lower educational attainment versus those with an advanced degree were less likely to report using NRT > 1 month (≤high school: β = −0.50, p < 0.001; college: β = −0.24, p = 0.019). Use of NRT > 1 month, in turn, was positively associated with quitting success (β = 0.25, p < 0.001). Those with lower educational attainment were also less likely to report a home smoking restriction (≤high school: β = −0.42, p < 0.001; college: β = −0.21, p = 0.009). Having a home smoking restriction was positively associated with quitting success (β = 0.50, p < 0.001). Results were similar with income substituted for education. Using NRT > 1 month and having a home smoking restriction are two strategies that may explain the association between low education and lower cessation success; these strategies should be further tested for their potential ability to mitigate this association.

1. Introduction Despite substantial progress against cigarette smoking in the United States about 16% of adults currently smoke (Jamal et al., 2018; National Center for Health Statistics, 2017). The prevalence of smoking is higher (Jamal et al., 2018; National Center for Health Statistics, 2017) and cessation rates are lower (Zhuang et al., 2015) among those with lower educational attainment compared to those with higher educational attainment. The reasons for the education disparity in smoking cessation are unclear, though some possible factors have been identified. For example, smokers who are more tobacco dependent experience greater difficulty in quitting (Hymowitz et al., 1997); evidence suggests that those with less education are more tobacco dependent (Lund, 2015;

Siahpush et al., 2006). Moreover, smokers who live or work in environments where smoking is restricted are more successful at quitting (Fichtenberg and Glantz, 2002; Pizacani et al., 2004; Sorensen et al., 2004), and restrictions are less common in these environments for lower socioeconomic status persons (Dai and Hao, 2017; Gan et al., 2015; Homa et al., 2015). Support for quitting (e.g., quitline counseling or emotional support offered by family or friends) has also been associated with cessation (Westmaas et al., 2010) and those with more education are more likely to be influenced by members of their social network than those with less education (Christakis and Fowler, 2008). Smokers who receive advice from a clinician to quit smoking are more likely to quit and to use cessation aids (Fiore et al., 2008). However, people with lower educational attainment are more apt to delay or not seek medical care (National Center for Health Statistics, 2017).

☆

Conflicts of interest: none. Financial disclosures: AGS, SAF, and AJ are employed by the American Cancer Society as part of the Surveillance and Health Services Research Program, which received a grant from Merck, Inc. for intramural research; however, their salary is solely funded through American Cancer Society funds. ⁎ Corresponding author at: 250 Williams Street NW, Atlanta, GA 30303, United States of America. E-mail address: [email protected] (A. Goding Sauer). 1 Co-lead authors. ☆☆

https://doi.org/10.1016/j.ypmed.2018.08.031 Received 9 February 2018; Received in revised form 20 June 2018; Accepted 26 August 2018 Available online 29 August 2018 0091-7435/ © 2018 Elsevier Inc. All rights reserved.

Preventive Medicine 116 (2018) 32–39

A. Goding Sauer et al.

2.2.3. Demographic factors Demographic variables included in the analyses were age (continuous), sex (male, female), and race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, non-Hispanic other).

Some previous studies have attempted to determine what may account for smoking-related socioeconomic disparities. One study assessed smoking cessation resources, exposure to smoke at work and home, and peer smoking behaviors (Honjo et al., 2006). Their results suggested that smokers from higher social class are more likely to use effective cessation resources and have a smoking-restrictive home environment (Honjo et al., 2006). However, this study was limited by its small sample size (n = 481), and because it was state-based the findings may not be generalizable to the United States. Another study incorporated social support, neighborhood disadvantage, stress, craving for nicotine and self-efficacy into their conceptual model to test pathways from socioeconomic status to smoking cessation (Businelle et al., 2010). That study found neighborhood and social support factors as well as stress and self-efficacy to be significant mechanisms in the socioeconomic-cessation relationship (Businelle et al., 2010). Although, the results are insightful, the study population was limited to treatmentseeking smokers (n = 424). To our knowledge, there has been no nationwide population-based investigation of the education-cessation relationship that simultaneously examined multiple potential mechanisms, including the duration of nicotine replacement therapy (NRT) use. The present study used structural equation modeling (SEM) of a large national cross-sectional dataset to identify smoking and cessation-related behavioral and environmental variables associated with both lower educational attainment and less quitting success.

2.2.4. Tobacco dependence Tobacco dependence was measured by elapsed time until first cigarette after waking (a Fagerstrom Test for Nicotine Dependence (Heatherton et al., 1991) item—≤30 min, > 30 min), nighttime smoking (yes, no), and smoking initiation age (< 16 years, ≥16 years) (Baker et al., 2007; Bover et al., 2008; Breslau and Peterson, 1996; Scharf et al., 2008). These measures were used individually in descriptive analyses and used to form a latent tobacco dependence variable in SEM analyses. Cigarettes per day was not included in the measure of tobacco dependence as the TUS-CPS did not assess this for all former smokers for the time period immediately preceding cessation. 2.2.5. Smoking restrictions Home smoking restriction was assessed as allowing home smoking “in some places/times” (reference) versus “not at all.” Work smoking restriction was indicated by “work indoors/smoking restriction” (reference), “work indoors/no smoking restriction,” “other work environment,” and “retired/not working.” 2.2.6. Cessation support Cessation help from employer categories were “cessation help offered” (reference), “cessation help not offered,” “self-employed/work in home,” and “retired/not working.” Additionally, support for quitting from friends and family was defined as “used” or “not used” (reference).

2. Methods 2.1. Study population Data from the 2010–2011 Tobacco Use Supplement (TUS) to the Current Population Survey (CPS), a household survey administered by the United States Census Bureau, were used (U.S. Department of Commerce and Census Bureau, 2012). The person-level nonresponse rates for the May 2010, August 2010, and January 2011 waves were 37.7%, 38.4%, and 40.2%, respectively (U.S. Department of Commerce and Census Bureau, 2012). Participants of interest were civilian, selfrespondents age ≥ 25 years who were current or former smokers (n = 59,790; 10,243 who had made a recent quit attempt, as defined below). Analysis occurred in 2016 and 2017. Institutional review board approval was not required as deidentified publicly-available data were used (Health and Human Services and Office for Human Research Protections, 2016). Although more recent TUS-CPS data are available (2014–2015), that round did not include detailed questions regarding use of individual cessation methods as was necessary for the present study.

2.2.7. Healthcare utilization Healthcare utilization categories were “visited a medical doctor” versus “not” (reference) (past 12 months for continuing smokers, or 12 months prior to the last quit attempt for former smokers). 2.2.8. Cessation aid utilization Measurement of cessation aid use pertained to the last quit attempt for former smokers and the most recent quit attempt within the preceding year for continuing smokers. Use and duration of NRT was indicated by “not at all” (reference), “one month or less,” or “more than one month.” Behavioral counseling (quitline, individual or group counseling) was categorized as “use” versus “no use” (reference). Supplemental Table 1 presents pertinent survey questions. 2.3. Analytic plan

2.2. Measures

For descriptive statistics, weighted point estimates and 95% confidence intervals were generated using SAS-callable SUDAAN 11.0.1 (RTI International, 2016). To identify plausible mechanisms, bivariate analyses examined differences in the above-described smoking and cessation-related behavioral and environmental variables by educational attainment using F-tests. For SEM, direct effects of the independent variable, educational attainment, on the dependent variable, quitting success, controlling only for demographic variables were modeled first. A subsequent model included variables significantly associated with educational attainment in bivariate analyses with direct paths both from educational attainment to each of these variables, and from these variables to quitting success. The direct educational attainment-quitting success relationship was also included in this model. Given the limitations of including “other work environment” and “retired/not working” categories for workplace variables, we ran the model excluding these categories. The resulting model had poor fit (RMSEA = 0.062, TLI = 0.877, CFI = 0.911). Considering the overall advantages and disadvantages of including or excluding the observations linked to these categories, we retained them and identified this as an area in need of further research.

2.2.1. Outcome The primary outcome, quitting success, was successful versus unsuccessful quitting, defined as former versus continuing smokers. Former smokers quit smoking 6–24 months prior to survey completion but were once daily smokers for ≥6 months (n = 2607). Six months was chosen as the lower limit to increase the likelihood that cessation was sustained; 24 months was chosen as the upper limit to reduce recall bias and is in accordance with previous research (Smith et al., 2017). Continuing smokers were current everyday smokers who made a quit attempt in the 12 months preceding survey completion (n = 7636). The 12-month timeframe was based on available data. 2.2.2. Primary independent variable The primary independent variable, educational attainment, had three categories: high school degree or less (≤high school; n = 5463), some college/college degree (college; n = 4435), and graduate degree or higher (advanced; n = 345; reference). No participant was missing educational attainment data. 33

Preventive Medicine 116 (2018) 32–39

A. Goding Sauer et al.

Table 1 Descriptive statistics by education level, adults 25 years and older, Tobacco Use Supplement-Current Population Survey, United Statues, 2010–11. Characteristic

Age 25–34 35–44 45–54 55–64 65–74 75–84 85+ Missing Sex Male Female Missing Race/ethnicity NH white NH black Hispanic Other Missing Education level ≤HS degree Some college/college degree Advanced degree Missing Quitting success/smoking status Former smoker Continuing smoker Missing Smoking restriction at home Smoking allowed in at least some areas or at least at some times Complete restriction Missing Cessation help offered by employer (in past year) Yes - help offered No - help not offered Not working (includes retired) Self-employed/work in a home Missing Smoking restriction at work Yes - work indoors with smoking restriction No - work indoors with no smoking restriction Not working (includes retired) Other work environmenta Missing Doctor visit (in past year) Yes No Missing Time to first cigarette after awakening ≤30 min > 30 min Missing Night smoking Yes, cigarette at night No Missing Age at initiation < 16 years ≥16 years Missing Cessation aid useb

Overall (n = 10,243)

%

95% CI

n

100.0 27.0 21.0 25.5 17.6 6.9 1.9 0.2

(26.0–28.0) (20.1–22.0) (24.5–26.5) (16.8–18.4) (6.4–7.5) (1.6–2.2) (0.1–0.4)

2540 2142 2596 1913 806 218 28 0

100.0 52.0 48.0

(50.9–53.0) (47.0–49.1)

4879 5364 0

100.0 76.2 11.1 7.9 4.9

(75.2–77.2) (10.3–11.8) (7.2–8.6) (4.4–5.4)

100.0 53.5 43.2 3.4

(52.2–54.8) (41.9–44.4) (2.9–3.8)

8163 920 580 580 0 10,243 5463 4435 345 0

100.0 26.0 74.0

(25.0–27.1) (73.0–75.0)

2607 7636 0

100.0 43.1

(41.9–44.3)

4411

56.9

(55.7–58.1)

5606 226

100.0

≤HS degree (n = 5463)

Some college/college degree (n = 4435)

Advanced degree (n = 345)

%

95% CI

n

%

95% CI

n

%

95% CI

n

100.0 24.5 20.6 26.9 17.2 7.8 2.7 0.4

(23.1–25.9) (19.4–21.9) (25.6–28.3) (16.1–18.4) (7.0–8.6) (2.2–3.2) (0.2–0.6)

1217 1094 1463 1011 496 162 20 0

100.0 30.6 21.5 23.9 17.4 5.6 1.0 0.1

(29.0–32.3) (20.1–23.0) (22.5–25.3) (16.1–18.7) (4.9–6.4) (0.7–1.3) (0.0–0.1)

1257 975 1059 811 277 51 5 0

100.0 20.0 20.6 22.5 26.2 9.3 0.9 0.5

(15.5–25.5) (16.0–26.2) (17.8–28.2) (21.2–31.8) (6.3–13.3) (0.3–2.7) (0.1–2.8)

66 73 74 91 33 5 3 0

100.0 53.3 46.7

(51.8–54.8) (45.2–48.2)

2676 2787 0

100.0 50.1 49.9

(48.5–51.7) (48.3–51.5)

2025 2410 0

100.0 54.3 45.7

(48.5–60.1) (40.0–51.5)

178 167 0

100.0 74.0 12.4 9.4 4.2

(72.7–75.4) (11.4–13.5) (8.4–10.4) (3.6–4.9)

4262 549 373 279 0 – – – – –

100.0 78.5 9.8 6.2 5.6

(76.9–80.0) (8.8–10.9) (5.3–7.1) (4.8–6.5)

3611 353 194 277 0 – – – – –

100.0 81.1 6.1 5.4 7.4

(75.2–85.9) (3.7–10.0) (2.7–10.4) (4.6–11.8)

290 18 13 24 0 – – – – –

– – – – – 100.0

– – – – –

22.6 77.5

(21.4–23.8) (76.2–78.6)

1210 4253 0

100.0 47.9

(46.2–49.7)

2618

52.1

(50.4–53.8)

2731 114

100.0

– – – – – 100.0

– – – – –

29.1 70.9

(27.5–30.8) (69.2–72.5)

1260 3175 0

100.0 38.1

(36.6–39.6)

1683

61.9

(60.4–63.5)

2648 104

100.0

< 0.0001

0.0181

< 0.0001

– – – – – 100.0

– – – – –

41.7 58.3

(35.6–48.1) (51.9–64.4)

137 208 0

100.0 30.3

(24.9–36.4)

110

69.7

(63.6–75.1)

227 8

< 0.0001

< 0.0001

< 0.0001

100.0

11.7 38.2 42.1

(10.9–12.5) (37.0–39.5) (41.0–43.2)

1143 3699 4153

8.5 35.5 48.4

(7.6–9.4) (33.9–37.1) (46.8–49.9)

447 1829 2569

14.8 41.8 35.2

(13.6–16.1) (39.9–43.6) (33.6–36.7)

626 1746 1494

22.2 35.8 30.0

(17.2–28.1) (29.8–42.3) (24.4–36.4)

70 124 90

8.1

(7.4–8.8)

818

7.7

(6.8–8.7)

407

8.3

(7.4–9.2)

368

12.0

(8.2–17.2)

43

430 100.0 38.8

(37.6–39.9)

3.1

211

3881

100.0 31.2

(29.7–32.7)

(2.7–3.6)

321

2.9

41.4

(40.3–42.5)

4153

16.8

(15.9–17.7)

1624 264

100.0 67.5 32.5

(66.4–68.6) (31.4–33.6)

7009 3048 186

100.0

1674

(45.3–48.7)

(2.5–3.5)

155

3.3

47.7

(46.2–49.3)

2569

18.2

(16.9–19.5)

919 146

100.0 65.0 35.1

(63.3–66.5) (33.5–36.7)

3617 1751 95

100.0

53.9 46.1

(52.8–54.9) (45.1–47.2)

5457 4525 261

100.0 15.1 84.9

(14.3–16.0) (84.0–85.7)

1507 8588 148

100.0 29.3 70.7

(28.3–30.4) (69.6–71.7)

3032 7105 106

201 100.0 47.0

58.5 41.5

(57.1–59.9) (40.1–43.0)

3141 2189 133

100.0 16.8 83.2

(15.7–17.9) (82.1–84.3)

910 4472 81

100.0 32.6 67.4

(31.1–34.1) (65.9–68.9)

1816 3588 59

18 < 0.0001

2020

100.0 53.4

(46.5–60.3)

187

(2.7–4.0)

155

4.2

(2.1–8.3)

11

34.4

(32.9–36.0)

1494

29.1

(23.7–35.3)

90

15.3

(14.1–16.6)

655 111

13.2

(9.3–18.5)

50 7

100.0 70.1 29.9

(68.4–71.8) (28.3–31.6)

3136 1212 87

100.0 74.4 25.6

(68.2–79.8) (20.3–31.8)

256 85 4

100.0

< 0.0001

< 0.0001

100.0

48.8 51.2

(47.0–50.5) (49.5–53.0)

2154 2164 117

100.0 13.2 86.8

(12.1–14.4) (85.6–87.9)

555 3817 63

100.0 26.0 74.1

(24.5–27.4) (72.6–75.5)

1149 3243 43

p-Values

47.0 53.0

(39.9–54.3) (45.7–60.1)

162 172 11

100.0 14.0 86.0

(10.0–19.3) (80.7–90.0)

42 299 4

100.0 20.5 79.5

(15.9–26.0) (74.0–84.1)

67 274 4

< 0.0001

< 0.0001

(continued on next page) 34

Preventive Medicine 116 (2018) 32–39

A. Goding Sauer et al.

Table 1 (continued) Characteristic

Nicotine replacement therapy Not at all ≤1 month > 1 month Missing Friend/family support Yes No Missing Behavioral counselingc Yes No Missing

Overall (n = 10,243)

%

95% CI

n

100.0 75.7 18.7 5.6

(74.6–76.7) (17.8–19.7) (5.1–6.2)

7488 1955 586 214

100.0 32.4 67.6

(31.2–33.6) (66.4–68.8)

3328 6793 122

100.0 6.1 93.9

(5.6–6.6) (93.4–94.4)

715 9406 122

≤HS degree (n = 5463)

Some college/college degree (n = 4435)

Advanced degree (n = 345)

%

95% CI

n

%

95% CI

n

%

95% CI

n

100.0 76.8 18.4 4.8

(75.3–78.2) (17.2–19.8) (4.2–5.5)

4055 1023 270 115

100.0 74.7 19.0 6.2

(73.1–76.3) (17.7–20.5) (5.3–7.3)

3203 858 280 94

100.0 70.4 19.6 10.0

(64.3–75.9) (14.9–25.4) (6.9–14.2)

230 74 36 5

100.0 29.6 70.4

(28.1–31.1) (68.9–71.9)

1636 3766 61

100.0 35.7 64.3

(33.9–37.4) (62.6–66.1)

1578 2801 56

100.0 34.9 65.1

(28.9–41.4) (58.6–71.2)

114 226 5

100.0 5.5 94.5

(4.8–6.3) (93.7–95.2)

346 5057 60

100.0 6.7 93.3

(5.9–7.7) (92.3–94.1)

340 4038 57

100.0 7.2 92.8

(4.8–10.9) (89.1–95.3)

29 311 5

p-Values

0.0100

< 0.0001

0.1237

NH: non-Hispanic. HS: high school. CI: confidence interval. F-tests used to examine differences by educational attainment levels. Boldface indicates statistical significance at p < 0.05. a Other work environment includes: self-employed, travel and/or variable work environment, work in a home, work outdoors. b Cessation aid use pertains to the last quit attempt among former smokers and most recent attempt (within the past year) among continuing smokers. c Behavioral counseling includes quit line, individual or group counseling.

A final good-fitting model was selected after excluding nonsignificant paths and accounting for covariances that improved model fit based on modification indices. Robustness of the final model was examined using 1) replicate weights and 2) family income instead of educational attainment. Family income was defined as low (≤$49,999/ year), middle ($50,000-99,999/year), and high (≥$100,000/year; reference). The root mean square error of approximation (RMSEA, < 0.08), comparative fit index (CFI, > 0.90), and Tucker Lewis index (TLI, > 0.90), were considered acceptable fit indices (Cheung and Rensvold, 2002; Fabrigar et al., 1999; Hu and Bentler, 1999). Confidence intervals (90%) for RMSEA were provided. Mplus version 7.0 (Mplus, 2017) was used in SEM analyses (α = 0.05).

3.2. Structural equation model The model examining only the direct educational attainment-quitting success association controlling for sociodemographic factors had excellent fit (RMSEA = 0.000; CFI = 1.000; TLI = 1.000) but accounted for a small proportion (2.6%) of variance (Fig. 1). A subsequent model including paths from educational attainment to proposed explanatory variables and from those variables to quitting success, had acceptable fit (RMSEA = 0.061; 90% confidence interval = 0.059–0.062; CFI = 0.979; TLI = 0.972; R2 = 0.50). However, this model indicated, counterintuitively, that greater tobacco dependence was associated with a higher likelihood of quitting success. The model was revised to test whether this relationship could reflect smokers who are more tobacco-dependent seeking more intense cessation assistance (e.g., cessation aid use for longer duration). After including a path of this hypothesized relationship, excluding nonsignificant paths and adding co-variances that improved fit (based on modification indices), the previously significant positive association between greater tobacco dependence and likelihood of quitting was reduced to nonsignificance (β = 0.038, p = 0.201; Fig. 2). This final good-fitting model (RMSEA = 0.060; 90% confidence interval = 0.059–0.062; CFI = 0.980 TLI = 0.972) accounted for 44% of the variance in quitting success. The educational attainment-quitting success direct effects in the final model remained significant but the association was reduced compared to the initial model (≤high school: β = −0.50, p < 0.001 to β = −0.25, p = 0.002; college: β = −0.29, p < 0.001 to β = −0.17, p = 0.020; Figs. 1 and 2). The final model identified two factors that were significantly associated with both lower levels of educational attainment and quitting success: use of NRT > 1 month and having a home smoking restriction. Specifically, compared to respondents with an advanced degree, those with less education were significantly less likely to report having used NRT > 1 month (versus no NRT use) (≤high school: β = −0.50, p < 0.001; college: β = −0.24, p = 0.019). Use of NRT > 1 month, in turn, was positively associated with quitting success (β = 0.25, p < 0.001). The indirect effect of less education on decreased quitting success, as a result of not using NRT > 1 month was −0.126 for those with ≤high school degree (p < 0.001) and −0.060 for the college educated (p = 0.030), for a total indirect effect of −0.186. In addition, compared to respondents with advanced degrees, those with less education were less likely to report a home smoking restriction (≤high

3. Results 3.1. Descriptive statistics The study population was mostly < 65 years (91.1%) and white (76.2%) (Table 1). Approximately half (48.0%) were female and about half had ≤high school degree (53.5%). Respondents with college education were younger (76.0% < 55 years) than those with ≤high school degrees (72.0% < 55 years) and those with advanced degrees (63.1% < 55 years, p < 0.0001). Nearly one-third (32.6%) of those with ≤high school degrees reported starting smoking < 16 years versus 26.0% of those with college education and 20.5% with an advanced degree (p < 0.0001). The proportion of respondents reporting a home smoking restriction increased with advancing education (≤high school = 52.1%, college = 61.9%, advanced = 69.7%). Furthermore, only 31.2% of those with ≤high school degree reported work smoking restrictions versus 47.0% of those with college education and 53.4% with an advanced degree (p < 0.0001). Cessation help offered by employers was reported by only 8.5% of those with ≤high school degree compared to 14.8% of those with college education and 22.2% of those with an advanced degree (p < 0.0001). Those with the least education were also less likely to report using support from friends/ family (≤high school = 29.6%, college = 35.7%, advanced = 34.9%, p < 0.0001). Additionally, the least educated were less likely to report using NRT > 1 month (4.8%) compared to those with more education (college = 6.2%, advanced = 10.0%, p = 0.0100).

35

Preventive Medicine 116 (2018) 32–39

A. Goding Sauer et al.

-.20* -.50**

-.30**

R2 =0.026

Fig. 1. Diagram of structural equation model of direct effects of educational attainment on quitting smoking success, controlling for sex, race/ethnicity, and age, Tobacco Use Supplement-Current Population Survey, United States, 2010–11. RMSEA = 0.000; CFI = 1.000; TLI = 1.000; **p < 0.001; *p < 0.05; HS: high school; NH: nonHispanic. Note: Diagram presents standardized model results. Nonsignificant paths not depicted. Advanced degree used as reference group for educational attainment.

-.29** .03*

-.13**

school: β = −0.42, p < 0.001; college: β = −0.21, p = 0.009). Having a home smoking restriction was, in turn, significantly positively associated with quitting success (β = 0.50, p < 0.001). The indirect effects of less education on decreased quitting success by not having a home smoking restriction were − 0.21 for those with ≤high school degree (p < 0.001) and −0.10 for those with college education (p = 0.008) for a total indirect effect of −0.31. There was a significant association between respondents with ≤high school education and utilization of behavioral counseling. The relationship was not significant for those with college education. Unexpectedly, utilization of behavioral counseling was negatively associated with quitting success. Use of NRT ≤1 month and family/friend support were associated with quitting success but not associated with educational attainment. Workplace smoking restriction and cessation assistance were significantly associated with educational attainment but not with quitting success.

4. Discussion In our population-based sample of United States adults, not using NRT > 1 month and not having a home restriction were significantly and simultaneously associated with both levels of lower educational attainment and decreased quitting success. These factors may explain why smokers with lower educational attainment have greater difficulty quitting. These results build on prior studies that have documented relationships between less education and difficulty quitting by being the first to assess potential mechanisms of this relationship in a nationwide population-based sample that included duration of NRT use. Some experts have suggested that NRT use in the general population has not reduced cigarette consumption (Beard et al., 2011); our results and others suggest that utilization duration may be an important consideration (Schlam et al., 2016; Siahpush et al., 2015). Herein, the results indicated that those with lower educational attainment (versus advanced degree) were less likely to have used NRT > 1 month, which in turn was significantly associated with decreased quitting success. Previous studies have observed a relationship between socioeconomic status and use of cessation resources (Honjo et al., 2006) but none, to our knowledge, assessed variability in NRT utilization duration. Our results raise the question of whether the cost of prolonged NRT use is prohibitive for lower socioeconomic status smokers. Several forms of NRT are available over-the-counter and the current Affordable Care Act (ACA) provisions include coverage of all Food and Drug Administration-approved cessation aids (111th and Congress, 2010). However, overall, there is no evidence of change in cessation aid utilization despite this ACA provision (Babb et al., 2017). Educating individuals about insurance coverage of cessation aids and emphasizing the costs of continuing to smoke versus extended NRT use could be one way to promote longer NRT use in vulnerable populations. Evidence also suggests that gaining insurance increases the odds of quitting success for vulnerable populations (Bailey et al., 2016); thus, expanding insurance to those previously uninsured could be an additional avenue to increase NRT utilization. Because having a home smoking restriction was also significantly associated with both educational attainment and quitting success, it is possible that instituting a home smoking restriction may improve quitting success among those with lower educational attainment. Additionally, those with higher educational attainment are more likely to believe second- and third-hand smoke is dangerous to household residents and are more likely to have home smoking bans versus those with lower educational attainment (Winickoff et al., 2009). However,

3.3. Supplemental analyses The final model described above was also tested using the survey's replicate weights. The associations of NRT > 1 month and home smoking restriction with both educational attainment and quitting success were similar to those of the unweighted model (data not shown). With large samples such as ours, results of SEM using population-weighted data are typically similar to those of an unweighted model (Methuen, 2017). The robustness of the final model, without replicate weights, was further tested using family income instead of educational attainment. This model had good fit (RMSEA = 0.059; 90% confidence interval: 0.058–0.061; CFI = 0.979; TLI = 0.970) and explained 43% of the variance in quitting success (Supplemental Fig. 1). Unlike the final education model, when potential mechanism variables were included, the direct effects of family income on quitting success were no longer statistically significant (≤$49,999: β = −0.105, p = 0.032 to β = 0.026, p = 0.634; $50,000–$99,999: β = −0.351, p < 0.001 to β = 0.018, p = 0.751), suggesting that mediating variables accounted entirely for the initially significant direct effect of family income on quitting success. Similar to the final education model (Fig. 2), associations with using NRT > 1 month and having a home smoking restriction were significant (Supplemental Fig. 1).

36

Preventive Medicine 116 (2018) 32–39

A. Goding Sauer et al.

Fig. 2. Diagram of structural equation model of relationship between educational attainment and quitting smoking success, Tobacco Use Supplement-Current Population Survey, United States, 2010–11. RMSEA = 0.060 (90% CI: 0.059–0.062); CFI = 0.980; TLI = 0.972; **p < 0.001; *p < 0.05; HS: high school; NRT: nicotine replacement therapy; NH: nonHispanic; QA: quit attempt; cig: cigarette. Note: Diagram presents standardized model results. Nonsignificant paths not depicted. Advanced degree used as reference group for educational attainment. n = 9333; 910 excluded due to missing data for one or more variables.

cessation support (Kaiser Family Foundation and Health Research and Educational Trust, 2017). Moreover, information regarding workplace smoking restrictions in the current study was limited to those working indoors. Future research should attempt to better define the role of work environment in the educational attainment-quitting success relationship. The initial model showed an association between greater tobacco dependence and greater quitting success. This unexpected finding was explained by more tobacco dependent individuals being more likely to use NRT. Still, use of behavioral counseling was associated, counterintuitively, with less quitting success and post-hoc analyses could not determine the reason. This may reflect the complex associations among tobacco dependence, quit attempts, methods of quitting, and the length of time such methods are employed. Incorporating different measures of tobacco dependence (i.e., cigarettes/day) in future research may provide insight. Using a more detailed definition of behavioral counseling may also be beneficial. For example, information regarding number and length of counseling sessions was not available in the

establishing home smoking bans may be more challenging with multiple resident smokers. Future research that assesses the implementation of home smoking restrictions and subsequent abstinent rates would best determine whether this strategy is a true mediator of the education-cessation relationship. Experimental research assigning smokers to restriction implementation versus no implementation, would further determine the role of restrictions in improving quitting success. Although respondents with less education were less likely to report workplace smoking restrictions and cessation support, results from this study did not show associations between these two highly correlated workplace variables and quitting success, which is inconsistent with previous reports (Seidel et al., 2017). The lack of association between work-related variables and quitting success in the present study may reflect the fact that the proportion of those unemployed was relatively high for the least educated group. It is also possible that those with less education are less aware of what cessation assistance is offered by their employer. Additionally, those with less education are more likely to work at smaller companies (Headd, 2000) which are less likely to offer 37

Preventive Medicine 116 (2018) 32–39

A. Goding Sauer et al.

timeframe of former smokers, particularly with regards to cessation medication use (Borland et al., 2012). Our analysis included smoking and cessation-related behavioral and environmental variables excluded from previous studies, yet, not all potential mediators of the educational attainment-quitting success relationship were included herein. Future research may further elucidate this relationship by investigating depression/anxiety and insurance as potential mediators.

present study, but variation in such factors has been shown to affect cessation success (Fiore et al., 2008). As expected, less education predicted less healthcare utilization prior to quitting, possibly due to lower rates of insurance among the less educated (Kaplan et al., 2017). However, in this study, having seen a medical doctor in the 12 months prior to quitting was associated with less quitting success (i.e., continuing smoker). Continuing smokers were more likely to report visiting a doctor in the preceding year, possibly due to more smoking-related health problems. Despite the healthcare encounter, tobacco-related discussions may not occur during a medical visit aimed at treating comorbidities (Rojewski et al., 2016). Evidence suggests, however, that smokers who receive assistance and follow-up from a primary care provider have greater odds of successfully quitting (Park et al., 2015). The final income-quitting success model indicated that after accounting for potential mediators, there were no significant direct effects between income and quitting success, suggesting that the difficulty in quitting experienced by those with lower income is attributable to not using NRT > 1 month and not having a home smoking restriction. These results support the above suggestions that lower socioeconomic status smokers should be encouraged to extend NRT use and establish home smoking restrictions. Although this study employed cross-sectional data, the results are consistent with previously reported path analyses using longitudinal data (Honjo et al., 2006). SEM results herein, are also consistent with recent research on home smoking bans (Kegler et al., 2015; Williams et al., 2016). Furthermore, our results suggest that other variables included in the final model may not be mediators in the low educationless quitting relationship as these variables were associated with either educational attainment or quitting success but not both. Such variables included: workplace smoking restrictions, cessation help offered by employer, and use of family/friend support. While these strategies can improve quitting success, as demonstrated by previous research, the focus of the current study was on whether they could be plausible candidates for explaining the education-quitting disparity, while simultaneously controlling for other variables.

5. Conclusions Vulnerable populations, such as those with less education or income, continue to demonstrate higher smoking prevalence (Drope et al., 2018; Jamal et al., 2018; National Center for Health Statistics, 2017). Understanding the causes of the disparities in smoking prevalence and cessation success is necessary for tobacco control efforts. The current study points to two strategies, using NRT > 1 month and having a home smoking restriction, that could potentially mitigate the association between low education and lower cessation success. Supplementary data to this article can be found online at https:// doi.org/10.1016/j.ypmed.2018.08.031. Acknowledgements Ms. Goding Sauer; Drs. Westmaas, Jemal, and Fedewa are supported by the American Cancer Society's Intramural Research Department. Dr. Kim is supported by Kennesaw State University's Bagwell College of Education. References 111th Congress, 2010. The Patient Protection and Affordable Care Act, Public Law 111–148. pp. 906. Babb, S., Malarcher, A., Schauer, G., Asman, K., Jamal, A., 2017. Quitting smoking among adults - United States, 2000–2015. MMWR Morb. Mortal. Wkly Rep. 65, 1457–1464. Bailey, S.R., Hoopes, M.J., Marino, M., Heintzman, J., O'Malley, J.P., Hatch, B., Angier, H., Fortmann, S.P., DeVoe, J.E., 2016. Effect of gaining insurance coverage on smoking cessation in community health centers: a cohort study. J. Gen. Intern. Med. 31, 1198–1205. Baker, T.B., Piper, M.E., McCarthy, D.E., Bolt, D.M., Smith, S.S., Kim, S.Y., Colby, S., Conti, D., Giovino, G.A., et al., 2007. Time to first cigarette in the morning as an index of ability to quit smoking: implications for nicotine dependence. Nicotine Tob. Res. 9 (Suppl. 4), S555–S570. Beard, E., McNeill, A., Aveyard, P., Fidler, J., Michie, S., West, R., 2011. Use of nicotine replacement therapy for smoking reduction and during enforced temporary abstinence: a national survey of English smokers. Addiction 106, 197–204. Borland, R., Partos, T.R., Cummings, K.M., 2012. Systematic biases in cross-sectional community studies may underestimate the effectiveness of stop-smoking medications. Nicotine Tob. Res. 14, 1483–1487. Bover, M.T., Foulds, J., Steinberg, M.B., Richardson, D., Marcella, S.W., 2008. Waking at night to smoke as a marker for tobacco dependence: patient characteristics and relationship to treatment outcome. Int. J. Clin. Pract. 62, 182–190. Breslau, N., Peterson, E.L., 1996. Smoking cessation in young adults: age at initiation of cigarette smoking and other suspected influences. Am. J. Public Health 86, 214–220. Burns, E.K., Levinson, A.H., 2008. Discontinuation of nicotine replacement therapy among smoking-cessation attempters. Am. J. Prev. Med. 34, 212–215. Businelle, M.S., Kendzor, D.E., Reitzel, L.R., Costello, T.J., Cofta-Woerpel, L., Li, Y., Mazas, C.A., Vidrine, J.I., Cinciripini, P.M., et al., 2010. Mechanisms linking socioeconomic status to smoking cessation: a structural equation modeling approach. Health Psychol. 29, 262–273. Cheung, G.W., Rensvold, R.B., 2002. Evaluating goodness-of-fit indexes for testing measurement invariance. Struct. Equ. Model. 9, 233–255. Christakis, N.A., Fowler, J.H., 2008. The collective dynamics of smoking in a large social network. N. Engl. J. Med. 358, 2249–2258. Dai, H., Hao, J., 2017. The prevalence of exposure to workplace secondhand smoke in the United States: 2010 to 2015. Nicotine Tob. Res. 19, 1300–1307. Drope, J., Liber, A.C., Cahn, Z., Stoklosa, M., Kennedy, R., Douglas, C.E., Henson, R., Drope, J., 2018. Who's still smoking? Disparities in adult cigarette smoking prevalence in the United States. CA Cancer J. Clin. 68 (2), 106–115. https://doi.org/10. 3322/caac.21444. (Epub 2018 Jan 31). Fabrigar, L.R., MacCallum, R.C., Wegener, D.T., Strahan, E.J., 1999. Evaluating the use of exploratory factor analysis in psychological research. Psychol. Methods 4, 272–299. Fichtenberg, C.M., Glantz, S.A., 2002. Effect of smoke-free workplaces on smoking behaviour: systematic review. BMJ 325, 188. Fiore, M.C., Jaen, C.R., Baker, T.B., et al., 2008. Treating tobacco use and dependence: 2008 update. In: Clinical Practice Guideline.

4.1. Limitations The use of self-reported data is subject to recall bias and differences in question interpretation (e.g., friend/family support for cessation). Additionally, use of cross-sectional data precluded conclusions about directions of causality. Nevertheless, it is unlikely that using NRT > 1 month or having a home smoking restriction could cause lower educational attainment. Instituting a home smoking restriction may follow quitting; however, previous research supports the direction modeled in this analysis (Kegler et al., 2015; Williams et al., 2016). Additionally, when members of one's social network quit smoking, the individual is also more likely to quit (Christakis and Fowler, 2008), suggesting that if smoking is banned at home, cessation attempts by any member of the household are more likely to be successful. It is also possible that those who used NRT ≤1 month ceased using NRT following a relapse. However, evidence suggests that smokers with lower versus higher educational attainment, are more likely to stop using NRTs for reasons other than successfully quitting (Burns and Levinson, 2008). The sample size for the advanced degree category was small relative to the other education categories. However, given the disparities in smoking prevalence by education level (Jamal et al., 2018; National Center for Health Statistics, 2017), we felt it was important that those with advanced degrees were separated. Future research with education groups of more similar size may provide more robust results. In the present study, the timeframes used to define former (quit 6–24 months ago) and continuing smokers (quit attempt within past 12 months) differed. The data from the more recent timeframe of continuing smokers may be subject to less recall bias than that of the longer 38

Preventive Medicine 116 (2018) 32–39

A. Goding Sauer et al.

JAMA Intern. Med. 175, 1509–1516. Pizacani, B.A., Martin, D.P., Stark, M.J., Koepsell, T.D., Thompson, B., Diehr, P., 2004. A prospective study of household smoking bans and subsequent cessation related behaviour: the role of stage of change. Tob. Control. 13, 23–28. Rojewski, A.M., Baldassarri, S., Cooperman, N.A., Gritz, E.R., Leone, F.T., Piper, M.E., Toll, B.A., Warren, G.W., 2016. Exploring issues of comorbid conditions in people who smoke. Nicotine Tob. Res. 18, 1684–1696. RTI International, 2016. SUDAAN 11.0.1. Scharf, D.M., Dunbar, M.S., Shiffman, S., 2008. Smoking during the night: prevalence and smoker characteristics. Nicotine Tob. Res. 10, 167–178. Schlam, T.R., Fiore, M.C., Smith, S.S., Fraser, D., Bolt, D.M., Collins, L.M., Mermelstein, R., Piper, M.E., Cook, J.W., et al., 2016. Comparative effectiveness of intervention components for producing long-term abstinence from smoking: a factorial screening experiment. Addiction 111, 142–155. Seidel, S.E., Metzger, K., Guerra, A., Patton-Levine, J., Singh, S., Wilson, W.T., Huang, P., 2017. Effects of a tobacco-free work site policy on employee tobacco attitudes and behaviors, Travis County, Texas, 2010–2012. Prev. Chronic Dis. 14, E133. Siahpush, M., McNeill, A., Borland, R., Fong, G.D., 2006. Socioeconomic variations in nicotine dependence, self-efficacy, and intention to quit across four countries: findings from the International Tobacco Control (ITC) Four Country Survey. Tob. Control. 15 (Suppl. 3), iii71–5. Siahpush, M., Shaikh, R.A., McCarthy, M., Sikora Kessler, A., Tibbits, M., Singh, G.K., 2015. Association between duration of use of pharmacotherapy and smoking cessation: findings from a national survey. BMJ Open 5, e006229. Smith, A.L., Carter, S.M., Dunlop, S.M., Freeman, B., Chapman, S., 2017. Revealing the complexity of quitting smoking: a qualitative grounded theory study of the natural history of quitting in Australian ex-smokers. Tob. Control. 1–9. Sorensen, G., Barbeau, E., Hunt, M.K., Emmons, K., 2004. Reducing social disparities in tobacco use: a social-contextual model for reducing tobacco use among blue-collar workers. Am. J. Public Health 94, 230–239. U.S. Department of Commerce, Census Bureau, 2012. National Cancer Institute-sponsored Tobacco Use Supplement to the Current Population Survey (2010–11). Data Files and Technical Documentation. Westmaas, J.L., Bontemps-Jones, J., Bauer, J.E., 2010. Social support in smoking cessation: reconciling theory and evidence. Nicotine Tob. Res. 12, 695–707. Williams, R.S., Stollings, J.H., Bundy, L., Haardorfer, R., Kreuter, M.W., Mullen, P.D., Hovell, M., Morris, M., Kegler, M.C., 2016. A minimal intervention to promote smoke-free homes among 2-1-1 callers: North Carolina Randomized Effectiveness Trial. PLoS One 11, e0165086. Winickoff, J.P., Friebely, J., Tanski, S.E., Sherrod, C., Matt, G.E., Hovell, M.F., McMillen, R.D., 2009. Beliefs about the health effects of “thirdhand” smoke and home smoking bans. Pediatrics 123, e74–e79. Zhuang, Y.L., Gamst, A.C., Cummins, S.E., Wolfson, T., Zhu, S.H., 2015. Comparison of smoking cessation between education groups: findings from 2 US National Surveys over 2 decades. Am. J. Public Health 105, 373–379.

Gan, W.Q., Mannino, D.M., Jemal, A., 2015. Socioeconomic disparities in secondhand smoke exposure among US never-smoking adults: the National Health and Nutrition Examination Survey 1988–2010. Tob. Control. 24, 568–573. Headd, B., 2000. The characteristics of small-business employees. In: Monthly Labor Review, pp. 6. Health and Human Services, Office for Human Research Protections, 2016. Regulations & Policy. (Rockville, MD). Heatherton, T.F., Kozlowski, L.T., Frecker, R.C., Fagerstrom, K.O., 1991. The Fagerstrom Test for Nicotine Dependence: a revision of the Fagerstrom Tolerance Questionnaire. Br. J. Addict. 86, 1119–1127. Homa, D.M., Neff, L.J., King, B.A., Caraballo, R.S., Bunnell, R.E., Babb, S.D., Garrett, B.E., Sosnoff, C.S., Wang, L., 2015. Vital signs: disparities in nonsmokers' exposure to secondhand smoke—United States, 1999–2012. MMWR Morb. Mortal. Wkly Rep. 64, 103–108. Honjo, K., Tsutsumi, A., Kawachi, I., Kawakami, N., 2006. What accounts for the relationship between social class and smoking cessation? Results of a path analysis. Soc. Sci. Med. 62, 317–328. Hu, L., Bentler, P.M., 1999. Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Struct. Equ. Model. Multidiscip. J. 6, 1–55. Hymowitz, N., Cummings, K.M., Hyland, A., Lynn, W.R., Pechacek, T.F., Hartwell, T.D., 1997. Predictors of smoking cessation in a cohort of adult smokers followed for five years. Tob. Control. 6 (Suppl. 2), S57–S62. Jamal, A., Phillips, E., Gentzke, A.S., Homa, D.M., Babb, S.D., King, B.A., Neff, L.D., 2018. Current cigarette smoking among adults - United States, 2016. MMWR Morb. Mortal. Wkly Rep. 67, 53–59. Kaiser Family Foundation, Health Research & Educational Trust, 2017. Employer Health Benefits 2017 Annual Survey. pp. 222. Kaplan, R.M., Fang, Z., Kirby, J., 2017. Educational attainment and health outcomes: data from the Medical Expenditures Panel Survey. Health Psychol. 36 (6), 598–608. https://doi.org/10.1037/hea0000431. (Epub 2017 Apr 6). Kegler, M.C., Bundy, L., Haardorfer, R., Escoffery, C., Berg, C., Yembra, D., Kreuter, M., Hovell, M., Williams, R., et al., 2015. A minimal intervention to promote smoke-free homes among 2-1-1 callers: a randomized controlled trial. Am. J. Public Health 105, 530–537. Lund, M., 2015. Social inequality in cigarette consumption, cigarette dependence, and intention to quit among Norwegian smokers. Biomed. Res. Int. 2015, 835080. Methuen, 2017. In: Kim, J. (Ed.), Goodness-of-fit Statistics Using Replicate Weights in Mplus. Mplus, 2017. Structural Equation Modeling. National Center for Health Statistics, 2017. Health, United States, 2016: With Chartbook on Long-term Trends in Health. (Hyattsville, MD). Park, E.R., Gareen, I.F., Japuntich, S., Lennes, I., Hyland, K., DeMello, S., Sicks, J.D., Rigotti, N.A., 2015. Primary care provider-delivered smoking cessation interventions and smoking cessation among participants in the National Lung Screening Trial.

39