Communicating the Risks of Tobacco and Alcohol Co-Use

Abstract

Background:

While tobacco and alcohol co-use is highly prevalent across the United States, little experimental research has examined ways to counter such dual use. We developed and tested messages about the risks of co-using tobacco and alcohol among adults who used a combustible tobacco product and drank alcohol within the past 30 days.

Methods:

In an online experiment, 1,300 participants were randomly assigned to read different messages about tobacco and alcohol co-use (e.g., Alcohol and tobacco cause throat cancer). Three between-subjects experiments manipulated the presence of: 1) a marker word (e.g., Warning), 2) text describing the symptoms of health effects and a quitting self-efficacy cue, and 3) an image depicting the health effect. Participants rated each message using a validated Perceived Message Effectiveness (PME) scale. We used independent sample t-tests to examine differences between experimental conditions. Results include effect sizes (Cohen’s d) to compare standardized mean differences.

Results:

Our sample was 64% male, 70% white, 23% Black, and 17% Hispanic/Latino with a mean age of 42.4 (SD=16.4) years. Messages that described the symptoms of the health effect (d=0.17, p=0.002) and included an image (d=0.11, p=0.04) were rated significantly higher in PME compared with messages that did not describe symptoms and were text-only. We found no significant effects of a marker word or self-efficacy cue on PME.

Conclusions:

Messages that describe the symptoms of health effects and include text and images may be particularly effective for communicating the risks of tobacco and alcohol co-use and decreasing adverse health effects from co-use.

1.0. INTRODUCTION

Using tobacco products and drinking alcohol cause over a third of cancer deaths in the United States (US), and these behaviors synergistically impact the risk of multiple cancers in the aero-digestive tract (Anantharaman et al., 2011). However, few interventions have been developed to inform people about the risks of tobacco and alcohol co-use, which is problematic since in 2020, 8.1% of all US adults smoked cigarettes and drank alcohol in the past 30 days—representing 19.2 million adults—and among current US smokers, 57.0% drank alcohol in the past month (Centers for Disease Control and Prevention, 2020). The few tobacco and alcohol co-use interventions that have been developed show promise in reducing use of both substances (Ames, Pokorny, Schroeder, Tan, & Werch, 2014), but no published research, to our knowledge, has rigorously developed or evaluated messages about the risks of tobacco and alcohol co-use.

Several message components are known to increase the effectiveness of messages. First, messages that elicit a strong emotional response are often more effective than other types of messages (Allen et al., 2015; Durkin, Brennan, & Wakefield, 2012). Typically, this can be accomplished by highlighting negative health effects caused by the behavior. Health messages are also more effective if they are combined with text to promote self-efficacy (e.g., you can quit), since people are more likely to change behavior if they believe they can succeed (Sheeran, Harris, & Epton, 2014; Strahan et al., 2002). In addition, messages that begin with a marker word (e.g., Warning) may prime participants to notice a message, inform people about the danger of products, and provide credibility if the marker word references a government agency (e.g., Surgeon General) (Grummon, Hall, Taillie, & Brewer, 2019). Finally, research shows that messages with images depicting health effects are more effective than messages without images (Hammond, 2011).

The primary objective of the current study was to develop and evaluate messages about the risks of co-using tobacco and alcohol on perceived message effectiveness (PME). PME refers to the extent to which a person believes that a message will be effective in changing their behavior (Baig et al., 2018). Longitudinal research from tobacco education campaigns has shown that PME predicts quit intentions and cessation behavior, making it a useful construct to evaluate co-use messages (Noar, Barker, Bell, & Yzer, 2020). We hypothesized that co-use messages that included text about specific symptoms associated with health effects (Durkin et al., 2012), a self-efficacy cue (Strahan et al., 2002), a marker word (Mahood & World Health, 2003), and an image (Hammond, 2011) would have higher PME ratings than messages without these elements.

2.0. METHODS

2.1. Participants

We used Qualtrics to recruit participants through their existing panels for social science research. Qualtrics recruited 1,300 participants for our study from June to July 2021. Eligible participants were 18 years or older, spoke English, lived in the US, used cigarettes, cigars, or waterpipe tobacco in the past 30 days, and consumed alcohol in the past 30 days.

2.2. Procedures

The University of North Carolina at Chapel Hill Institutional Review Board approved the study. We conducted three separate between-subjects randomized experiments to manipulate different elements of co-use messages (see Figure 1 and Supplementary Files A–C). Our three experiments were:

1. Marker word: Participants were randomized to view one of four messages with a “Surgeon General” marker word, a “Danger” marker word, a “Warning” marker word, or no marker word.

2. Symptoms and self-efficacy: In a 2x2 factorial experiment, participants were randomized to view messages with a statement about the symptoms of a health effect (present vs absent) and a statement designed to cue self-efficacy for quitting (present vs absent). All statements included the text: “Alcohol and tobacco cause oropharyngeal cancer.” The text about the symptoms of health effect was: “Oropharyngeal cancer can cause pain and bleeding in the mouth.” The text with the self-efficacy cue was: “Quit tobacco and reduce alcohol to prevent oropharyngeal cancer.” We selected symptoms that were accurately caused by the health effect, based on consultation with an MD on our study team and understandable to a lay person.

3. Format x health effect: In a 2x2 factorial experiment, participants were randomized to view messages with different health effects (throat cancer vs. heart disease) and format (text-only statement vs. text + image). We selected images depicting personal suffering from the health effect that performed well in a previous study (Anonymous, 2021).

In total, participants saw three messages and answered PME items after each message. For instance, in Experiment 1, participants saw one of four potential messages. In Experiment 2, participants saw one of four potential messages, and so on.

Figure 1.

Experimental stimuli

2.3. Measures

2.3.1. PME

Our primary outcome was PME, which we assessed with three items, adapted from a previously validated and reliable scale (Baig et al., 2018). The three PME items assessed the ability of messages to change beliefs about consequences, attitudes about behaviors, and motivation regarding substance use. Specifically, the survey asked participants: “How much does this statement…: 1) “make you concerned about the health effects of using alcohol and tobacco?” (beliefs about consequences), 2) “make your use of alcohol and tobacco seem unpleasant?” (attitude about behavior), 3) “discourage you from wanting to use alcohol and tobacco?” (motivation). The response scale ranged from not at all (1) to a great deal (5). We averaged responses to the items (Cronbach’s alpha was 0.93).

2.3.2. Demographics and substance use

Participants reported their age, gender identity, sexual orientation, race, ethnicity, education, income, tobacco use and frequency, alcohol use, frequency, and quantity, and nicotine and alcohol dependence.

2.4. Data analysis

We conducted chi-square tests and ANOVAs to compare experimental conditions on demographic characteristics, tobacco use, and alcohol use variables, which revealed no differences (all p-values > 0.05). The interactions between manipulations in the 2x2 experiments were not significant; therefore, we only analyzed main effects. We used ANOVAs and independent samples t-tests to compare experimental conditions on PME and calculated effect sizes with 95% confidence intervals (CIs) as eta-squared (for the ANOVA) or Cohen’s d (for the t-tests). There were no missing data in our models.

We conducted a sensitivity analysis where we controlled for exposure to prior experiments (i.e., carryover effects) and found that exposure to prior experiments did not change any results in significance, direction, or magnitude. We also examined whether nicotine dependence, alcohol dependence, tobacco frequency, and alcohol frequency and quantity moderated effects of the experimental manipulations on PME and found no evidence of moderation. We conducted analyses in SAS (version 9.4) and used two-tailed tests with a critical alpha of 0.05.

3.0. RESULTS

3.1. Participants

Our sample (N=1300) was 64% male and 35% female. In terms of race/ethnicity, 70% were white, 23% were Black, and 17% were Hispanic/Latino. Participants’ mean age was 42.4 (SD=16.4) years (Supplementary File D). Participants resided in all 50 US states and Washington D.C. On average, participants’ nicotine dependence scores were 2.4 (SD=1.9) on a scale of 0-5 and participants’ alcohol dependence scores were 4.9 (SD=2.4) on a scale of 1-12.

3.2. Marker word

Exposure to a message with the Surgeon General, Danger, or Warning marker word did not affect PME compared with no marker word (p=0.67), eta-squared = 0.00; 95% CI: 0.00, 0.01) (Table 1).

Table 1.

Effects of the experimental manipulations on PME, N=1300

Experimental manipulations	PME Mean (SD)	p	Effect size (95% CI)
Experiment 1: Marker Word a
[None]	3.07 (1.22)	0.67	0.00 (0.00, 0.01)
Warning:	3.19 (1.24)
Danger:	3.17 (1.30)
Surgeon General:	3.16 (1.26)
Experiment 2: 2x2 Symptoms x self-efficacy b,c
Symptoms
Symptoms absent	3.34 (1.21)	0.002	0.17 (0.06, 0.28)
Symptoms present	3.55 (1.20)
Self-efficacy
Self-efficacy message absent	3.40 (1.26)	0.13	0.08 (−0.03, 0.19)
Self-efficacy message present	3.50 (1.16)
Experiment 3: 2x2 Format vs. health effect b,c
Text vs. text + image
Text-only	3.63 (1.17)	0.04	0.11 (0.01, 0.22)
Text + image	3.76 (1.18)
Health effect
Throat cancer	3.66 (1.17)	0.23	0.07 (−0.04, 0.18)
Heart disease	3.74 (1.19)

Perceived message effectiveness (PME)

^ap-value calculated using ANOVA and effect size calculated as eta-squared

^bp-values calculated using independent samples t-tests and effect sizes calculated as Cohen’s d

^cInteractions among conditions were not significant so only main effects were examined.

3.3. Symptoms & Self-Efficacy

Exposure to a message describing the symptoms of health effects increased PME compared with a control message that did not describe symptoms (p=0.002, Cohen’s d = 0.17; 95% CI: 0.06, 0.28). Including a self-efficacy cue did not affect PME (p=0.13, Cohen’s d = 0.08; 95% CI: −0.03, 0.19).

3.4. Format x Health Effect

Exposure to a text + image message increased PME compared with a text-only control (p=0.04, Cohen’s d = 0.11; 95% CI: 0.01, 0.22). The health effect referenced in the message did not affect PME (p=0.24, Cohen’s d = 0.07; 95% CI: −0.04, 0.18).

4.0. DISCUSSION

Communication interventions are an effective strategy for reducing tobacco and alcohol use (U.S. Department of Health and Human Services, 2012; Young et al., 2018), but no research, to our knowledge, has rigorously evaluated messages about the risks of co-using alcohol and tobacco. As expected and consistent with the existing literature, we found that messages with images were more effective than text-only messages without images. It is important to note that both images used in our experiment depicted personal suffering. There are other types of images that could be used, such as graphic depictions of the health effect (e.g., diseased organ) or a symbolic representation of the health effect (e.g., ball and chain to represent addiction), which could be evaluated in future research, particularly with important subgroups (e.g., young adults).

The messages used in our experiments highlighted negative consequences of alcohol and tobacco co-use as a way to provoke an emotional response and we found that adding text emphasizing specific symptoms (pain and bleeding in the mouth) improved PME scores. Research shows that emphasizing negative health effects may be more effective than other types of message themes (e.g., social norms, anti-industry) (Durkin et al., 2012), although more work is needed in this area for co-use messages, including 1) if messages evoking certain emotions (e.g., fear) are more effective than other emotions (e.g., hope) and 2) the extent to which messages elicit psychological reactance (Thompson, Pearce, & Barnett, 2007).

Contrary to our hypotheses, the self-efficacy cue did not enhance PME scores (Strahan et al., 2002). One way to enhance self-efficacy is to provide statements about quitting efficacy, as we did in this experiment (e.g., quit tobacco, reduce alcohol). Another way to enhance self-efficacy is to provide information about quitting (e.g., quit tips, websites for quitting). It is possible that messages with quitting information could be effective for co-use messages, but more research is needed. Importantly, increased access to resources (e.g., nicotine replacement therapy, behavioral therapy) to help people reduce or quit tobacco and alcohol consumption is needed to further bolster self-efficacy for behavioral change.

Finally, our study found that a marker word did not enhance the effectiveness of co-use messages. Tobacco and alcohol warning labels in the US currently carry marker words such as “Surgeon General’s Warning,” but few experiments have evaluated whether a marker word enhances message impact. One study on sugar-sweetened beverages found that a marker word increased PME (Grummon et al., 2019) and another on e-cigarettes found that a marker word increased warning recall (King et al., 2020). Our study suggests that co-use messages may not need a marker word to be effective.

It is important to note that effect sizes for our experimental manipulations on PME were small (d=0.11 and 0.17 for the image and symptoms manipulations, respectively). However, even small effects of communication interventions can have broad population-level reach and therefore, large impacts on public health. Notably, this reach can also include socially disadvantaged populations that may be difficult to reach with other types of interventions, such as intensive, in-person programs.

4.1. Limitations

Strengths of our study include use of a randomized experimental design and a well-validated outcome measure (Noar et al., 2020). Limitations include 1) the use of a one-time experiment with a cross-sectional study design that precluded measurement of future behavior, 2) that we assessed PME for both alcohol and tobacco within the same items rather than with separate items, which makes it hard to determine if effects occurred for one vs. both substances, and 3) the use of a convenience sample, which means that results may not generalize to other populations with different alcohol or tobacco use patterns. To this last point, however, research shows that well-designed experimental studies using convenience samples can yield results that are consistent with those found in representative samples (Jeong et al., 2019). Finally, more research evaluating PME for other substances beyond tobacco is needed, including how PME may predict co-use of substances.

5.0. CONCLUSIONS

In this study, we developed and tested messages about the risks of co-using tobacco and alcohol and found that messages that describe the symptoms of health effects and incorporate both text and images may be particularly effective at communicating risks.