Anti-Smoking Media Campaigns and Disparities in Smoking Cessation in the United States, 2001–2015

Abstract

Purpose:

To evaluate sociodemographic differences in the relationship between state and national anti-smoking media campaigns and cessation behaviors among adult smokers in the U.S.

Design:

Repeated cross-sectional analysis.

Setting:

U.S. nationally representative survey of adults ages 18 and older, 2001–2015.

Subjects:

76,278 year-ago smokers from the 2001–2015 Tobacco Use Supplement to the Current Population Survey.

Measures:

Area-level exposure to State-sponsored and “Tips from former smokers” anti-tobacco media campaigns was the primary predictor of this study. Outcome variables included: quit attempt in the past 12 months, past 30-day smoking cessation, and past 90-day smoking cessation among year-ago smokers.

Analysis:

We conducted modified Poisson regression models to examine the association between media campaign exposure and cessation behaviors. We also examined effect modification on the additive scale by sex, race/ethnicity, income, and education using average marginal effects.

Results:

Year-ago smokers with greater exposure to media campaigns were more likely to report 30-day (Prevalence Ratio [PR]: 1.18, CI: 1.03, 1.36) and 90-day cessation (PR: 1.18, CI: 1.00, 1.41) compared to respondents with less campaign exposure. We found no evidence of effect modification by sociodemographic variables.

Conclusion:

Exposure to anti-smoking media campaigns were associated with year-ago smokers’ cessation behaviors. However, there were no differences in the association by sex, race/ethnicity, income, or education, indicating that broadly focused media campaigns may be insufficient to reduce smoking cessation among priority populations, and thus health disparities generally.

Purpose

Tobacco use is the leading preventable cause of various chronic diseases and death in the U.S.¹ Media campaigns have commonly been used as a strategy to curb tobacco use,^2–4 as there is substantial evidence that they can promote quit attempts^5–11 and cessation^12,13 among adult smokers. However, little is known about potential differences in campaign effects on cessation behaviors across sociodemographic groups. This study aims to evaluate sociodemographic differences in state and national media campaign effects on cessation behaviors among adult smokers in the U.S.

There are considerable disparities in smoking cessation and quit attempts.^14–16 In general, males,^14,15 individuals with higher income,^14,16 and individuals with higher educational attainment^14,16 are more likely to make a quit attempt and to successfully quit. Regarding race/ethnicity, non-Hispanic Blacks are more likely to make a quit attempt than non-Hispanic Whites and Hispanics, but are less likely to successfully quit.¹⁴

Several studies have evaluated whether the effectiveness of anti-tobacco media campaigns, including the “Tips from Former Smokers” campaign (hereafter Tips),^5,8,9 the “Become an EX” campaign,¹⁰ State-sponsored campaigns,^13,17 and a campaign sponsored by the American Legacy Foundation,¹³ differs by sociodemographic characteristics, but have come to varying conclusions. For example, some studies found that media campaigns were more effective in promoting quit attempts and 30-day cessation among smokers with lower socioeconomic status (i.e., lower education and income).^9,10,13 Other studies have found the opposite pattern.^8,17 Moreover, media campaigns have been found to promote quit attempts among only non-Hispanic Blacks,¹⁰ only non-Hispanic Whites,⁸ or both,⁹ while no effects were found for Hispanics.^8–10 Another study found no differences in campaign effects by race/ethnicity or education.⁵ These contrasting findings across studies may be due, to some extent, to differing campaign content or media buys; however, they may also be due to differences in study design.

Importantly, most studies analyzing the differential effect of campaign exposure on cessation across different sociodemographic subgroups have estimated exposure using self-reported recall,^8–10,17 which is subject to recall bias.¹⁸ Relatively fewer studies on this topic have measured campaign exposure using Gross Rating Points (GRPs),^5,13,18 which provide an exogenous estimate of campaign exposure given media buys in a particular media market. Thus, investigations into examining the association between GRP measured exposure to anti-smoking media campaigns and cessation across sociodemographic subgroups would provide comparatively objective evidence for the effect of media campaigns on cessation behaviors.

We used a nationally representative sample of year-ago adult smokers in the U.S. from 2001–2015 to examine the association between GRP exposure to Tips and state-sponsored anti-smoking media campaigns and cessation behaviors, including recent quit attempts, 30-day smoking cessation, and 90-day smoking cessation. In addition, we tested for effect modification of these relationships by sex, race/ethnicity, household income, and educational attainment.

Methods

Design

We analyzed data from the 2001–2015 Tobacco Use Supplement (TUS) to the Current Population Survey (CPS). The CPS is conducted monthly, and selects 60,000 households using a two-stage sampling method, and surveys approximately 108,000 individuals 15 years or older in a given month.¹⁹ The TUS questionnaire is administered as a supplement to the CPS in selected years and months, and contains detailed questions on tobacco product use and cessation among adults 18 years and older. We used data from five TUS waves in 2001–2002, 2003, 2006–2007, 2010–2011, and 2014–2015 provided by the Integrated Public Use Microdata Series (IPUMS) at the University of Minnesota.²⁰

Sample

Of the original sample of 2,192,026 participants in the TUS dataset obtained from the IPUMS that were surveyed between 2001–2015, we excluded members of the longitudinal cohort (n = 62,711). Next, we excluded proxy respondents (n = 1,253,578) to minimize information bias on self-reported smoking measures. In addition, we excluded respondents younger than 25 years old (n = 89,756) to restrict the study sample to adults who have established smoking patterns and have likely completed their educational attainment.

We restricted the study sample to participants who smoked cigarettes a year ago, excluding never smokers, former smokers who quit smoking more than a year ago, and participants who were not smoking one year prior to the interview. Finally, we matched media campaign exposure data from each Designated Market Area (DMA), which are standard units used when buying ad time for traditional media campaigns, with corresponding TUS data based on metropolitan statistical area (MSA)/core-based statistical area (CBSA).

Because DMAs and metropolitan areas do not perfectly overlap, we classified a metropolitan area as within a given DMA if at least 70% of the population within that metropolitan area was located within the DMA, according to county population data from the U.S. Census Bureau.²¹ Our final analytic sample consisted of 76,052 year-ago smokers.

Of the final analytic sample, 6.7% had missing values for annual household income. Respondents in 2001–2007 had missing income values, but respondents in 2010–2011 and 2014–2015 waves did not because TUS provides imputed income data. In order to impute income data for 2001–2007, we employed multiple imputation using the chained equations method with IVEware version 0.3.²² Five data sets were imputed using respondents’ age, race, education, sex, marital status, family size, employment status, and longitudinally ascertained family income observations, as well as survey year.

Measures

Outcome Variables

Quit attempt, 30-day smoking cessation, and 90-day smoking cessation were the main outcome variables, and were analyzed only among year-ago smokers. Respondents were considered to have made a quit attempt if they stopped or tried to stop smoking one or more days in the 12 months prior to the interview. Participants who reported having quit smoking for at least 30 days at the time of the survey were classified as achieving successful cessation. Our 90-day cessation outcome was defined the same way, and represents longer-term cessation.^23,24 We did not include 6-month smoking cessation because our sample is limited to people who were smoking a year ago, leaving limited time for individuals to achieve 6-month cessation. However, previous work has demonstrated that the vast majority of smokers who have successfully quit for 90 days or more continue to abstain for at least one year.²⁵

Independent Variable

Our independent variable was exposure to Tips and state-sponsored anti-smoking media campaigns. Media campaign data came from Nielsen Media Research, and measured television advertising in households across the U.S.²⁶ Households in the U.S. are divided into 210 Designated Market Areas (DMAs); Nielsen provides data for the top 75 DMAs in the U.S.,^6,27 which accounts for about 78% of all viewing households.²⁶ Nielsen assigns television advertisements gross ratings points, or GRPs, which capture the percentage of households exposed to a given advertisement in a specific DMA over the course of a month.²⁶ Our analyses used 12-month GRP sums, which were divided by 100 to represent the average ad exposure rating per person over the prior 12 months.

Our combined exposure variable pooled GRPs of state-sponsored media campaigns from 2001 to 2015 and the Tips campaign from 2014–2015. Tips GRPs were zero from 2001–2013 as the campaign did not start until 2012 and TUS respondents were not surveyed in 2012 and 2013. Aggregate campaign exposure was dichotomized as <48 GRPs and ≥ 48 GRPs, as prior work has established 48 GRPs per year as a threshold level of exposure to media campaign for behavioral changes such as cessation among adult smokers.²⁸

Covariates

Key sociodemographic variables included sex (male or female), age (continuous, ≥25), race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, non-Hispanic Other), annual household income (<$15k; $15k-$29,999; $30k-$49,999; $50k-74,999; ≥$75k), and education level (<high school, high school diploma or GED, some college, 4 year college or graduate school). To control for state-level covariates, we included the yearly proportion of the population unemployed and living below the poverty line from the University of Kentucky’s Center for Poverty Research,²⁹ the yearly percent Hispanic and Black state populations from the Survey of Epidemiology and End Results (SEER),³⁰ and the state-level percent of college graduates among the population age 25 and older from the U.S. Census Bureau’s American Community Survey.³¹ Finally, we controlled for state-level average cigarette price per pack³² and the four-category census region the respondent resided in (Northeast, Midwest, South, and West).³³

Analysis

First, we calculated weighted prevalence estimates for the overall analytic sample. We also calculated the weighted percentage of all outcome variables, and provided weighted estimates of quit attempts, 30-day, and 90-day cessation by all relevant sociodemographic variables of interest. We then conducted modified Poisson regression analysis to examine the association between campaign exposure and outcome variables,³⁴ adjusting for covariates and fixed year effects. We used Poisson regression instead of logistic regression to estimate prevalence ratios instead of odds ratios.³⁵ Regression models estimated the association between media campaign exposure quit attempts, 30-day cessation, and 90-day cessation among respondents who were smoking one year prior. In additional analyses, we tested for effect modification of the association between media campaign exposure and each outcome variable by sex, race/ethnicity, income, and education, on the additive scale using average marginal effects.^36,37 To adjust for multiple testing, we used the Benjamini-Hochberg correction set at 5%.³⁸

We also conducted four sensitivity analyses. Our first sensitivity analysis examined a scaled continuous GRP predictor variable instead of the dichotomized GRP exposure, as an alternative modeling of the functional form relative to our dichotomous definition based on a cut-point (<48 GRPS, ≥48 GRPs) established in prior literature.²⁸ Second, we tested the functional form of GRPs as quadratic and cubic terms, and, separately, square root terms, to assess if there were non-linearities in the association between campaign exposure and smoking outcomes. Third, we examined multilevel Poisson regression with mixed state effects to explicitly account for clustering by outcomes within states. Finally, we tested the extent to which the relationship between anti-smoking media campaigns and the three outcome variables changed over time using interactions between year and the exposure. Separate analyses were conducted defining year continuously and categorically when checking for effect modification of media campaigns over time.

All analyses were adjusted for survey weights to account for the survey design and conducted in Stata SE, version 15. This research was considered to be exempt from human subjects regulations due to the use of publicly available, de-identified data.

Results

Table 1 presents descriptive statistics of the study sample. The sample was comprised mainly of males (54.4%) and non-Hispanic Whites (71.0%). Respondents most commonly reported having an annual household income of between $30,000 and $50,000 (23.5%) and had a high school education or GED (36.8%). During the study period, 66.3% of the sample was exposed to less than 48 GRPs over the prior 12 months, which included the 7.5% that were not exposed to any Tips or state-sponsored anti-smoking media campaigns. Among year-ago smokers, 41.1% had attempted to quit smoking, 10.0% reported cessation for 30 days or more, and 8.4% reported cessation for 90 days or more. Non-Hispanic Blacks reported higher proportions of quit attempts than all other racial/ethnic groups, but lower 30- and 90-day cessation. Respondents in the highest income group had the highest percentage of quit attempts (43.8%), 30-day smoking cessation (13.5%), and 90-day cessation (11.2%) compared to lower income groups, in a graded fashion. Individuals with at least a four-year college degree also had the highest percentage of quit attempts (45.3%) and the greatest proportion of 30- and 90-day cessation (15.0% and 12.6%, respectively) compared to lower education groups, again in a graded fashion. Respondents were generally more successful in quitting in recent years, with quit attempts (45.3%), 30-day cessation (12.3%), and 90-day cessation (10.5%) all at their highest in the 2014–2015 survey wave. Finally, the South and Midwest had lower proportions of individuals who made a quit attempt, and lower proportions of individuals reporting 30- and 90-day cessation than those in the West or the Northeast.

Table 1.

Descriptive statistics of study sample (year-ago smoker), Tobacco Use Supplement to Current Population Survey, 2001–2015

				Quit attempt (n=76,052)	30-day cessation (n=76,023)	90-day cessation (n=76,023)
Variable		n	% or Mean (SD)	%	%	%
Total		76,052		41.1	10.0	8.4
Age (≥ 25 years)		76,052	45.2 (12.5)
Sex
	Male	37,621	54.4	40.4	10.0	8.2
	Female	38,431	45.6	42.0	10.1	8.6
Race/ethnicity
	Non-Hispanic White	56,868	71.0	40.2	10.4	8.7
	Non-Hispanic Black	9,139	13.2	44.0	7.7	6.3
	Hispanic	6,199	10.6	42.2	10.7	8.7
	Non-Hispanic Other	3,846	5.3	43.8	10.4	8.4
Annual household income ($)
	0–15K	13,500	17.9	39.8	7.7	6.5
	15–30K	14,763	19.4	39.5	8.5	6.9
	30–50K	17,971	23.5	41.1	9.4	7.7
	50–75K	14,514	18.8	41.1	11.0	9.5
	≥75K	15,304	20.3	43.8	13.5	11.2
Education
	<High school	11,681	15.7	36.2	7.0	5.8
	High school or GED	28,365	36.8	38.5	8.1	6.7
	Some college	23,013	30.5	44.5	11.1	9.3
	≥4 year college	12,993	17.0	45.3	15.0	12.6
Survey wave (year)
	2001–2002	19,504	22.9	42.2	8.8	7.2
	2003	15,651	18.9	38.8	9.3	7.8
	2006–2007	15,893	21.8	39.1	9.9	8.3
	2010–2011	13,971	19.4	40.7	10.4	8.6
	2014–2015	11,033	17.0	45.3	12.3	10.5
Anti-smoking media campaign exposure (GRPs)
	GRP sum (0–309.65)	76,052	37.7 (38.5)
	<48	52,586	66.3	40.6	9.6	8.0
	≥48	23,466	33.7	42.2	11.0	9.2
Census region
	Northeast	17,253	22.3	42.4	10.3	8.4
	Midwest	18,239	23.0	40.6	9.2	7.7
	South	24,599	33.7	39.5	9.5	8.1
	West	15,961	21.0	42.8	11.5	9.6
State cigarette price ($ per pack)		76,052	4.2 (1.6)
State % unemployment		76,052	6.1 (1.9)
State % poverty		76,052	12.9 (2.7)
State % Black		76,052	12.7 (6.7)
State % Hispanic		76,052	15.3 (11.4)
State % college grad (age 25+)		76,052	27.6 (4.4)

Note. n = unweighted sample size. % = weighted percentage. M = weighted mean. SD = weighted standard deviation. GRP = gross rating point.

Table 2 reports the regression results for the models examining the relationship between anti-tobacco media campaign and quit attempts, 30-day cessation, and 90-day cessation. In bivariate analyses, individuals who were exposed to at least 48 GRPs were more likely to report making a quit attempt (Prevalence Ratio [PR]: 1.04, 95% CI: 1.02–1.06) and quitting smoking for at least 30 days (PR: 1.15, CI: 1.09–1.21) and 90 days (PR: 1.15, CI: 1.09–1.22), compared to individuals exposed to less than 48 GRPs. In fully adjusted regression models, individuals exposed to at least 48 GRPs were more likely to achieve successful 30-day (PR: 1.07, CI: 1.01–1.12) and 90-day cessation (PR:1.08, CI: 1.02–1.14) relative to individuals with less than 48 GRPs of exposure. Adjusted models showed no significant association between media campaign exposure and quit attempts.

Table 2.

Main effects of GRP measured exposure to anti-smoking media campaign on Quit attempts, 30-day cessation, and 90-day cessation among year-ago adult smokers (age ≥ 25 years)

			Quit attempt	30-day cessation	90-day cessation
			PR (95% CI)	PR (95% CI)	PR (95% CI)
Unadjusted model (bivariate association)
GRPs
		<48	Reference	Reference	Reference
		≥48	1.04 (1.02, 1.06)	1.15 (1.09, 1.21)	1.15 (1.09, 1.22)
Adjusted model a
GRPs
		<48	Reference	Reference	Reference
		≥48	1.01 (0.98, 1.04)	1.07 (1.01, 1.12)	1.08 (1.02, 1.14)
Age			0.97 (0.97,0.98)	0.91 (0.90, 0.92)	0.91 (0.90, 0.92)
Age2			1.00 (1.00, 1.00)	1.00 (1.00, 1.00)	1.00 (1.00, 1.00)
Female (vs. Male)			1.05 (1.03, 1.07)	1.05 (0.99, 1.10)	1.07 (1.02, 1.13)
Race/ethnicity (vs. White)
	Non-Hispanic Black		1.15 (1.10, 1.19)	0.89 (0.79, 1.00)	0.87 (0.76, 1.00)
	Hispanic		1.06 (1.03, 1.09)	1.11 (1.02, 1.22)	1.08 (0.99, 1.19)
	Non-Hispanic Other		1.04 (0.98, 1.11)	0.87 (0.75, 1.02)	0.84 (0.71, 0.99)
Income (vs. 0–15K)
	15–30K		0.99 (0.95, 1.03)	1.04 (0.94, 1.16)	1.00 (0.90, 1.11)
	30–50K		1.02 (0.98, 1.05)	1.12 (1.04, 1.22)	1.09 (0.99, 1.18)
	50–75K		1.01 (0.97, 1.04)	1.27 (1.16, 1.40)	1.29 (1.18, 1.41)
	≥75K		1.06 (1.02, 1.10)	1.48 (1.37, 1.59)	1.43 (1.33, 1.54)
Education (vs. <High school)
	High school or GED		1.06 (1.02, 1.11)	1.13 (1.05, 1.21)	1.11 (1.02, 1.22)
	Some college		1.21 (1.16, 1.26)	1.47 (1.33, 1.61)	1.46 (1.32, 1.62)
	≥4 year college		1.22 (1.15, 1.29)	1.82 (1.67, 1.99)	1.83 (1.67, 2.01)
Census regions (vs. West)
	Northeast		1.01 (0.96, 1.07)	0.97 (0.86, 1.09)	0.98 (0.88, 1.09)
	Midwest		1.02 (0.96, 1.08)	0.94 (0.83, 1.07)	0.95 (0.84, 1.07)
	South		1.00 (0.94, 1.08)	0.96 (0.84, 1.10)	0.97 (0.85, 1.11)
State cigarette price			1.01 (0.99, 1.02)	0.99 (0.95, 1.03)	0.97 (0.93, 1.01)
% of state unemployment			0.86 (0.20, 3.61)	0.21 (0.01, 3.18)	0.15 (0.01, 3.01)
% of state below poverty			1.47 (0.77, 2.80)	0.71 (0.16, 3.24)	0.54 (0.12, 2.47)
% of state that is Hispanic			1.03 (0.88, 1.21)	1.38 (1.01, 1.89)	1.50 (1.13, 1.99)
% of state that is Black			0.68 (0.50, 0.94)	0.80 (0.41, 1.56)	0.83 (0.44, 1.55)
% of state that is college graduates			1.01 (1.00, 1.01)	1.01 (0.99,1.02)	1.01 (0.99, 1.02)

Note. PR = Prevalence ratio. CI = confidence interval. GRP = gross rate point. Boldface numbers indicate statistically significant PRs (p < .05).

^a.regression model was adjusted for age, age² sex, race/ethnicity, income, education, state cigarette price, state unemployment rate, state poverty rate, state-level percent Hispanic population, state-level percent Black population, state-level percent college graduates among the population age 25 and older, census regions and fixed year effects.

Additive p-values for all models with two-way interaction terms are summarized in Table 3. After adjusting for multiple testing using the Benjamini-Hochberg correction set at 5%, we found no statistically significant interactions between media campaigns and gender, race/ethnicity, household income, or education for any of the outcomes.

Table 3.

Unadjusted p-values^a for joint test of additive interactions between anti-smoking media campaign and gender, race/ethnicity, household income, education, and time for Quit attempts, 30-day cessation, and 90 day cessation among year-ago adult smokers (age ≥ 25 years)

Interaction	Quit attempt	30-day cessation	90-day cessation
Media×Male	0.768	0.695	0.829
Media×Race	0.066	0.725	0.103
Media×Income	0.538	0.284	0.642
Media×Education	0.082	0.602	0.443
Media×Timeb	0.670	0.867	0.585

Note. GRP = gross rating points.

^a.Each interaction is estimated from a separate Poisson model with all main effects and a single interaction term between media campaign exposure and either gender, race/ethnicity, income, education, or year variable. All models controlled for age, age² sex, race/ethnicity, income, education, state cigarette price, state unemployment rate, state poverty rate, state-level percent Hispanic population, state-level percent Black population, state-level percent college graduates among the population age 25 and older, census regions and fixed year effects.

^b.Time reflects year, defined continuously

In sensitivity analyses, we tested different functional forms of the exposure variables. Analyses using continuous GRPs as our primary exposure showed no significant findings for quit attempts, 30-day cessation or 90-day cessation. Further, fitting higher order polynomial models for GRP exposure (quadratic, cubic, square-root) provided no statistically significant evidence of non-linear trends.

In the third sensitivity analysis, we examined whether results from multilevel models adjusting for mixed state effects were consistent with results from modified Poisson regression models adjusting for the state-level demographic measures. We found that the results were similar in magnitude and direction: Individuals who were exposed to at least 48 GRPs were more likely to achieve successful 30-day cessation (PR: 1.09, CI: 1:03–1.16) and 90-day cessation (PR: 1.10, CI: 1.03–1.18) compared to individuals exposed to less than 48 GRPs (Appendix Table 1).

Analyses investigating if there was effect modification on smoking cessation behaviors by time found a significant difference in the impact of campaign exposure on quit attempts in years 2006 and 2015 when using time as a categorical variable (Appendix Figure 1), but no differences when time was a linear variable. We found no other time interactions for any of the other outcomes after correcting for multiple testing.

Discussion

Using a nationally representative sample of smokers between 2001–2015 in the U.S., we found that in areas with greater exposure to Tips and state-sponsored anti-smoking media campaigns, year-ago smokers were more likely to have stopped smoking for both 30 and 90 days, but not more likely to have tried to quit. We found no evidence of disparities in the strength of the association between anti-smoking media campaign exposure and cessation behaviors, even though cessation behaviors varied by smokers’ sociodemographic characteristics.

Regarding quit attempts, several studies have shown that exposure to the Tips campaign was associated with a greater likelihood of trying to quit,^5,8,9 while studies looking at the impact of state-sponsored campaigns have shown mixed results,^6,17,39,40 which could be due to variability in the effectiveness of state-specific ads. Our findings agree with the evidence that suggests state-sponsored anti-tobacco media campaigns do not increase the likelihood of a quit attempt.^6,17 The similarity of our findings to the literature on state-sponsored campaigns rather than the Tips campaign may be because the Tips campaign did not begin until 2012, so individuals surveyed in the first four waves (or roughly 85% of respondents) were only exposed to state-sponsored campaigns. However, even in the years when Tips media campaigns were included, there was no relationship with quit attempts, as observed in our sensitivity analysis examining effect modification by time (Appendix Figure 1).

Our finding that media campaign exposure increases the likelihood of stopping smoking for 30 and 90 days is consistent with several studies;^12,13 however, the few studies to address this issue have shown mixed results.^12,13,17 Specifically, one study found greater levels of Tips exposure to be associated with an increased probability of successfully quitting, although the association was only significant for respondents that had been exposed to the third highest level of campaign exposure, when grouped into five categories.¹² Another study looking at exposure to state or American Legacy Foundation media campaigns found that individuals with higher levels of GRPs were more likely to report cessation for 30 days or more,¹³ although a different study looking exclusively at state-sponsored campaigns found no association between media campaign exposure and quitting.¹⁷

We found no evidence that the impact of the Tips and state-sponsored media campaigns differed by sex, race/ethnicity, income, or education. Several other studies have analyzed whether anti-tobacco media campaigns have differential effects on quit attempts and cessation for individuals of various sociodemographic backgrounds, and have come to mixed conclusions ^{5,8–10,13,17} Two studies that used interactions to explore potential differential relationships between Tips exposure and quit attempts by sociodemographic factors echoed our findings that there were no significant differences by race or educational attainment.^5,9 Several studies evaluating Tips campaign exposure also presented stratified quit attempt results, but interactions were either not tested⁸ or non-significant.⁹ One study found Tips exposure to be associated with an increased probability of quit attempts among only non-Hispanic Whites and individuals with some college education,⁸ while another study using Tips data from 2012–2013 (prior to Tips data included in our study) found that higher ad exposure was associated with an increased probability of quit attempts among non-Hispanic Whites and non-Hispanic Blacks, relative to Hispanics, as well as for individuals of lower educational attainment.⁹ No studies reported results as to the differential impact of Tips exposure by sex, race/ethnicity, or socioeconomic status on successful cessation.

Only one study tested for differential effects of state-sponsored ad campaigns with respect to making a quit attempt, and found that media campaigns were less likely to promote quit attempts in individuals of lower educational attainment compared to those with college degree.¹⁷ However, this study was limited only to ads aired in Wisconsin in 2003–2004, while our study tested exposure to a combined ads of state and Tips among a nationally representative year-ago smoker sample in 2001–2015. One other study analyzed the differential impact of a separate media campaign, the EX-campaign, between 2003–2006, and found that confirmed awareness of EX ads was associated with a higher likelihood of attempting to quit, but only among non-Hispanic Blacks and individuals with less than a high school education.¹⁰

Regarding cessation, very few studies have examined the differential impact of media campaigns on cessation by sociodemographic factors. One study found that the impact of highly emotional ads – or ads that generally show a personal story or the harmful health impacts of smoking – sponsored by states or the American Legacy Foundation in 2001–2004 on the higher odds of 30-day cessation was greater among individuals of low or middle SES relative to individuals of high SES defined by education and income levels.¹³ Another study analyzing Wisconsin’s state-sponsored ad campaign found no differential association between the impact of ads on 1-year cessation by educational attainment.¹⁷ Differences in the type of media campaign, year of data, type of media campaign, and study sample between previous studies and our study might explain the differences in the results between our studies.

Disparities in smoking cessation are readily observable in the descriptive breakdown of our study population, which reveal a considerable gap between the proportion of non-Hispanic Black participants who make a quit attempt compared to those who are successful in quitting, relative to other racial/ethnic groups. Likewise, far lower proportions of less educated and lower income individuals report successful 30- and 90-day cessation compared to their higher SES counterparts. These findings were corroborated in the adjusted regression models, and are consistent with other studies that have demonstrated non-Hispanic Blacks⁴¹ and individuals of lower SES⁴² are less likely to quit. Because individuals of lower income and education are more likely to smoke in the first place,⁴³ disparities in smoking cessation exacerbate disparities in smoking prevalence.

Given our findings that Tips and state-sponsored anti-tobacco media campaigns are not more effective among racial/ethnic minority and low SES populations, other strategies must be employed to reduce disparities in cessation. There is considerable evidence that highly emotional ad campaigns^13,40 and the use of graphic images or personal testimonials^40,44 might be an effective way of reaching low and middle SES tobacco users. Furthermore, some research has demonstrated that digital segmentation, a strategy that specifically markets to buyers with distinct needs or characteristics,⁴⁵ could play an integral role in ensuring media campaigns with tailored messages reach racial/ethnic minority and lower SES individuals.⁴⁵ Digital segmentation has become a common component of health promotion campaigns, including youth anti-smoking campaigns (e.g., Truth campaign) and campaigns that target individuals of lower SES.⁴⁵ Media campaigns tailored specifically to adults in racial/ethnic minority populations may also be of benefit to reducing cessation-related disparities, though more research is needed to identify themes most salient and impactful to this group.

This study has several limitations. First, this study did not consider the effects of other forms of media, specifically online advertisements, which can affect smokers’ cessation behaviors. We were unable to measure exposure to online advertisement, so cannot determine the extent to which those advertisements influenced our findings. However, we did not observe effect modification by time in models predicting 30-day and 90-day abstinence, and the effect modification by time for quit attempts showed no discernible patterning; hence, it is unlikely that our findings are affected by the increase in other forms of media. Second, the analysis of nation-wide media campaign exposure data did not allow for the evaluation of advertisement themes on smoking related behaviors, although differences in campaign effectiveness by advertisement type have been documented.^40,46 Third, while using exogenous exposure measures like GRPs allows us to limit confounding due to recall bias, it is defined by area-level media buys, so there is a chance that individuals residing in a specific area where Tips or state-sponsored media campaigns were featured were not exposed to the extent their GRP score might have indicated. Lastly, we may have lacked statistical power to fully explore effect modification. For example, the additive interaction between race/ethnicity, as well as education, and media campaigns on quit attempts was marginally significant prior to adjusting for multiple testing, and we cannot fully rule out differential associations.

Our study found that higher levels of exposure to Tips and state-sponsored anti-tobacco media campaigns between 2001 and 2015 were associated with higher likelihood of 30- and 90-day smoking cessation, but not quit attempts. We found no evidence of effect modification of these relationships by sex, race/ethnicity, income, or educational attainment. While the successes of Tips and state-sponsored anti-tobacco media campaigns in aiding in cessation is encouraging, additional efforts, including targeted media campaigns, are needed to address cessation disparities among racial/ethnic minorities and individuals of lower SES.

SO WHAT?

What is already known on this topic?

While anti-tobacco media campaigns have been shown to be effective in reducing tobacco use and aiding in smoking cessation, the majority of literature is older or does not analyze how campaigns differ in effectiveness by sociodemographic group.

What does the article add?

We provide an in-depth look at how recent anti-tobacco media campaigns impact quit attempts and cessation in the U.S., and explore whether their effectiveness varies by sex, race/ethnicity and socioeconomic status. We found that higher levels of Tips and state-sponsored campaign exposure was associated with a greater probability of making a quit attempt and having quit in the past 30 or 90 days. Campaigns did not appear to differ in effectiveness by and sociodemographic markers of interest.

What are the implications for health promotion practice or research?

Media campaigns show a considerable impact on smoking cessation, and should be leveraged moving forward. Strategies utilizing digital segmentation and highly emotional ads should be further explored as options to reach vulnerable populations