Design Cues for Tobacco Communication: Heuristic Interpretations and Usability of Online Health Information about Harmful Chemicals

Abstract

Objective:

Many people have a poor understanding of the numerous chemicals in tobacco products that cause severe health harms. The US government must publicly display a list of these harmful chemicals to the public. Online disclosures are one promising solution, but evidence is needed for effective design strategies to encourage interpretation and use of information as intended.

Method:

To examine the impact of website designs for the activation of heuristics and usability perceptions, a national probability sample of US adolescents and adults (n=1441) was randomized in a 3 (chemical format) x 2 (webpage layout) between-subjects online experiment. Chemicals were displayed as names only, with a visual risk indicator, or with numerical ranges. Layouts displayed health harms at the top of the webpage separate from chemicals or the chemicals grouped by associated health harms. Participants viewed a webpage and reported activated heuristics, usability (perceived ease of use and usefulness), and intentions to use the website.

Results:

Displaying risk indicators increased website usability by encouraging users to rely on colors to interpret the risk of the chemicals (all p<.01). Website designs that grouped chemicals with harms allowed users to link the chemicals to harms they cause and increased perceived usability and intentions to use the website (all p<.001).

Conclusion:

Assessing heuristics gives insights for how US adolescents and adults interpret chemical information and the impact of design strategies on usability. Public disclosures of chemicals in tobacco products could be optimized with color-coded risk indicators and layouts placing chemicals near the harms they cause.

1. Introduction

Smoking remains the number one cause of preventable death in the US and of many acute health harms.^1–3 The US Food and Drug Administration (FDA) is legally required to communicate about harmful and potentially harmful constituents (chemicals) in tobacco products “by brand and by quantity.”⁴ These chemical disclosures must also be “understandable and not misleading to a lay person.” While not explicitly stated, the legal requirements appear to reflect intentions for chemical disclosures to educate and potentially discourage tobacco use.⁵ Yet, the public has little understanding of these chemicals and their health harms, and risks are often misunderstood in the US and globally.^6–10 Engaging users with clear chemical information can increase awareness and potentially change behavior.^11–13

Optimizing chemical disclosures online is one solution to inform the public.¹⁴ Americans regularly consult the internet for health information.¹⁵ However, for online chemical disclosures to be effective, people must desire to use them and interpret the information as intended. According to the technology acceptance model (TAM),^16,17 user adoption (intentions-use) is best explained by usability beliefs.¹⁸ Users are more likely to adopt chemical disclosure websites if perceived as easy to use (requires little effort) and useful (enhances one’s performance) to learn about chemicals in cigarette smoke. Although website designs influence technology acceptance,^19,20 the TAM does not include guidance for visualizations or strategies to increase usability beliefs or intentions. Nor does this framework account for how audiences are interpreting designs.

To identify effective designs, it is fruitful to look to cues that activate easy-to-process heuristics.²¹ People are bombarded with thousands of messages daily, and rely on heuristics, or mental shortcuts, to process information quickly.²² Design cues can activate specific heuristics, or ways of thinking,²³ and influence use of online information.^24,25 Understanding heuristics could improve tobacco communication; matching people’s mental models can encourage understanding and influence behavior.¹²

Color cues in chemical disclosures could activate heuristics and provide easy-to-interpret context.²⁶ Around the world, red is an active color linked to danger through traffic symbols and warnings.^26,27 Many individuals heuristically link red=danger.^27–29 Visual risk indicators using red to convey risk, along with text descriptions, may increase the likelihood information is interpreted as intended (Hypothesis 1, H1); colors and descriptions (e.g., dangerous) are attention-getting and easily processed to facilitate intuitive understanding.^29–31

Numbers could inform the public and increase the usefulness of online chemical disclosures by conveying precise and verifiable information.³⁰ Numbers may activate the heuristic of more=bad (H2). Yet, presenting numbers assumes the public will use their education or familiarity,³⁰ despite many individuals struggle to understand and apply numbers for health decisions.³² Numbers can be difficult to interpret without sufficient understanding of health contexts (e.g., reference points), and people often prefer descriptive, text-based information over numbers for tobacco chemical communication.^33,34 Numbers may also be misleading – lower amounts incorrectly perceived as safer or inconsequential and amounts vary by industry testing conditions.^33,35–38 Numerical ranges can convey variations in actual amounts. However, showing ranges could activate an unintended heuristic, more=confusing (H3), and decrease usability and use of online chemical disclosures.

Where objects are placed can convey meaning. According to Gestalt psychology, individuals believe objects closer together belong together.³⁹ Website layout, through structural proximity of objects, can facilitate understanding⁴⁰ and potentially increase the usability and acceptance of online information.⁴¹ Placement of chemical information near harms may increase usability; close=related (H4).⁴²

This study examined whether users “decode” chemical disclosures with predicted heuristics and whether color, number, and layout cues encourage greater usability and adoption. This study also assessed whether heuristic evaluations of cues explain (mediate) usability and adoption (RQ1). To do so, a between-subjects online experiment was conducted with a nationally representative sample of US adolescents and adults.

2. Materials and Methods

A 3 (format) x 2 (layout) experiment examined the influence of the webpage designs on activated heuristics, usability perceptions, and use intentions. The first factor – chemical display format – had three levels: chemical names only, with visual risk indicators, or with numerical ranges. The second factor – webpage layout – had two levels: health effects shown separate from chemicals (Layout A) or chemicals organized by health effects (Layout B). Participants were randomly assigned to view one of six versions.

2.1. Participants

A national probability sample of US adults and adolescents (n=1441) was recruited by the Carolina Survey Research Laboratory as part of a larger study to understand public reactions to chemical information.⁴³ Participants were selected using random-digital-dial and list-assisted sampling frames, including landlines and cellphones, with oversampled counties with higher prevalence of smokers and low-income individuals. All participants were 13 years old or older and spoke/read English. The response rate was 82%.

2.2. Stimuli

Static webpages with chemicals in a fictitious cigarette brand (Brentfield Gold) were created (examples in Figure 1). Webpages had a title, chemical information header, health harms, and footnotes with applicable keys. Health harms included five FDA-identified categories: cancers, permanent breathing problems, heart attack and stroke, reproductive organ damage, and addiction. Each webpage had 20 chemicals, 18 from FDA’s abbreviated list of harmful and potentially harmful constituents and two other familiar chemicals.^44,45 Six webpages were created: three chemical formats shown in two layouts.

Figure 1.

Example Chemical Disclosure Webpage Designs

Note: All numerical ranges and level of risk were provided by a toxicologist to reflect an expected range for reported chemicals in cigarette smoke.

Formats included: chemicals as names only, with visual risk indicators, or with numerical ranges. Chemicals were shown as names only, without amounts, in the control condition. Chemical amounts were shown qualitatively with visual risk indicators, a color-coded system indicating three levels of risk: green for “Safe: does not cause health problems,” light red for “Risky: puts you at risk to develop health problems,” and dark red for “Dangerous: can cause immediate damage to your body.” The amount of each chemical was shown with numbers (e.g., 6-29 ng), in the ranges condition. A toxicologist provided risk levels and quantities to accurately reflect current cigarette smoke. Layouts included Layout A with health effects near the top of the webpage and the chemicals listed underneath and Layout B with chemicals grouped by associated health harms.

2.3. Procedure

Following consent, participants were randomly assigned to one between-subjects condition and shown a static webpage with outcome measures below. Participants were compensated $45. The Institutional Review Board at the University of North Carolina approved the study.

2.4. Measures

Response options ranged from “strongly disagree” (coded as 1) to “strongly agree” (5). Newly developed heuristic items also included “does not apply to this webpage” (coded as 0, excluded from analyses) to allow participants to indicate no opinion (e.g., if numbers do not appear); “does not apply” was added after items were piloted with a national convenience sample of US adults (n=254) from Amazon Mechanical Turk.

2.4.1. Color Heuristic.

The survey assessed a color heuristic, red=danger, with the item, “When looking at the webpage, the color red alerts me to chemicals that are dangerous” (M=4.04, SD=1.26).

2.4.2. Amount Heuristics.

An intended heuristic, more=bad, was assessed with, “When looking at the webpage, larger numbers next to chemicals show me which ones are more harmful” (M=3.25, SD=1.40). An unintended amount heuristic, more=confusing, was assessed with, “When looking at the webpage, seeing a lot of numbers stops me from wanting to understand the webpage” (M=2.71, SD=1.32).

2.4.3. Location Heuristic.

The survey assessed a location heuristic, close=related, with, “When looking at the webpage, the location of the chemicals shows me the health problems they cause” (M=3.77, SD=1.38).

2.4.4. Perceived Ease of Use.

How easy it would be to use the website was assessed with three adapted items: “This webpage would be easy to use,” “My interaction with this webpage would be clear and understandable,” and “It would be easy to get this webpage to do what I want” (M=3.77, SD=1.00, α=.86).^16,46

2.4.5. Perceived Usefulness.

Perceived usefulness was assessed with three adapted items: “Using this webpage would increase how much I know about chemicals in cigarette smoke,” “I would find this webpage useful to learn about chemicals in cigarette smoke,” and “Using this webpage would enhance my knowledge about chemicals in cigarette smoke” (M=4.25, SD=.87, α=89).^16,46

2.4.6. Intentions to Use the Webpage.

Intentions were assessed with the adapted item: “Assuming I had access to this webpage, I predict that I would use it” (M=3.28, SD=1.24).¹⁶

2.5. Data Analysis

To determine the impact of chemical format and layout on each heuristic, an analysis of variance (ANOVA) and Bonferroni adjusted post hoc comparisons were conducted. The predictor variable was chemical format (name only, visual risk indicator, range) for ANOVAs for color and amount heuristics. The predictor variable was layout (A, B) for the ANOVA for the location heuristic. A 3 (format) x 2 (layout) multivariate analysis of variance (MANOVA), ANOVAs, and post hoc comparisons were conducted to examine effects on perceived ease of use, usefulness, and intentions to use the website. Exploratory analyses were conducted to determine if age (adolescent vs. young adult 18-25 vs. adult), smoking status (current smoker vs. nonsmoker), or numeracy (perfect score/missed one item vs. missed more than one) moderated the effect of cues on heuristics or TAM outcomes. Next, PROCESS models were run to determine whether activated heuristics mediated significant effects of perceive ease of use, perceived usefulness, and use intentions, controlling for age, smoking status, and numeracy.⁴⁷ Format was recoded into two dummy variables to compare against the other conditions (risk indicator vs. ranges & names only; ranges vs. risk indicator & names only). Single mediator models were run with the color heuristic (format) and location heuristic (layout). Parallel multiple mediator models were run for the amount heuristics (format) to account for associations between intended and unintended heuristics. All analyses were conducted in SPSS v24.

3. Results

Participants (n=1441) were between 13 and 90 years old (M=32.47, SD=18.69), White (76%), African American (15%), Asian (3%) and non-Hispanic (94%); see Table 1. Roughly a third (31%) of participants had a college degree. Almost one in five (18%) identified as current smokers. For each heuristic, fewer than 8% selected “does not apply to this webpage;” these participants were excluded from respective analytic samples.

Table 1.

Participant Demographics (n = 1441)

	n (%)
Age (M = 32.47, SD = 18.69)
Adolescents (13-17)	427 (30%)
Young adults (18-25)	320 (22%)
Adults (26 or older)	692 (48%)
Gender
Female	776 (54%)
Male	654 (45%)
Transgender/Other	10 (1%)
Race
White	1095 (76%)
Black or African American	218 (15%)
American Indian or Alaska Native	22 (2%)
Asian	36 (3%)
Pacific Islander	8 (1%)
Other	61 (4%)
Ethnicity
Non-Hispanic	1347 (94%)
Hispanic	94 (7%)
Education
Less than high school	462 (32%)
High school diploma	253 (18%)
Some college	273 (19%)
Associate’s degree	94 (7%)
Bachelor’s degree	222 (15%)
Graduate or professional degree	135 (9%)
Smoking status
Smoker	262(18%)
Nonsmoker	1179 (82%)
Numeracy (4 items)
Missed four (0)	64 (4%)
Missed three (.25)	197 (14%)
Missed two (.5)	377 (26%)
Missed one (.75)	448 (31%)
Perfect score (1)	355 (25%)

Note: Current smokers were defined as people who have smoked at least 100 cigarettes in their lifetime and smoke every day or some days. Numeracy was assessed with four items requiring participants to select the biggest risk shown in numbers. Numeracy items as correct (1) or incorrect (0; missing responses were coded as incorrect) and averaged them for a total numeracy score.

3.1. Impact of Chemical Format

Participants who saw the visual risk indicator were more likely to use a color heuristic (red=danger) than people who names only or ranges, F(2,1282)=47.21, p<.001. Means and standard deviations are shown in Table 2.

Table 2.

Impact of Chemical Display for Webpages about Chemicals in Cigarette Smoke

Construct (n)	1 Chemical name only M (SD)	2 Chemical with a risk indicator M (SD)	3 Chemical with ranges M (SD)	F	η2	1-2	1-3	2-3
Color heuristic (red = danger, 1285)	3.72 (1.33)	4.46 (1.01)	3.85 (1.30)	47***	.07	***		***
Amount heuristic – intended (more = bad, 1169)	3.06 (1.35)	3.23 (1.43)	3.43 (1.39)	7.18**	.01		***
Amount heuristic – unintended (more = confusing, 1273)	2.56 (1.31)	2.60 (1.31)	2.93 (1.31)	11 ***	.02		***	**
Perceived ease of use (1441)	3.81 (1.02)	3.94 (.91)	3.57 (1.04)	17***	.02		**	***
Perceived usefulness (1441)	4.25 (.89)	4.35 (.83)	4.16 (.89)	5.85**	.01			**
Intentions to use the website (1441)	3.28 (1.25)	3.34(1.24)	3.21 (1.24)	1.51	.00

^**p < .01;

^***p < .001

Showing ranges led to greater use of an intended amount heuristic (more = bad) than names only, F(2,1166)=7.18, p=.001; there were no differences between ranges and the risk indicator. Showing ranges also led to greater use of an unintended heuristic (more = confusing) than for names only or the risk indicator, F(2,1270)=11.08, p<.001. In two instances, format had a larger impact among participants with higher numeracy (perfect score/missed one, n=803) than lower numeracy (n=638): The visual risk indicator (vs. ranges and names) led to greater endorsement of the color heuristic, F(2, 1279)=6.43, p=.002, among those with higher numeracy, 95% CI for difference: .53, 1.05 and .73, 1.27, than those with lower numeracy, 95% CI for difference: .06, .67 and .15, .73. Showing ranges (vs. risk indicator and names) only led to significant endorsement of the unintended amount heuristic, F(2, 1267)=3.56, p=.029, among those with higher numeracy, 95% CI for difference: .11, .68 and .30, .88.

Webpages with visual risk indicators or names only were easier to use than webpages with ranges, F(2,1435)=17.27, p<.001. Visual risk indicators did not make websites easier to use than chemical names only. Visual risk indicators led to greater perceived usefulness compared to ranges, F(2,1435)=5.85, p=.003, but not more than names only. Chemical format did not influence intentions to use the website, F(2,1435)=1.51, p=.221. Age and smoking status did not moderate the effects of format on heuristics, perceived ease of use, usefulness, or use intentions.

Activation of the color heuristic explained (i.e., mediated) the relationship between showing a visual risk indicator and perceived ease of use (Figure 2). Use of the color heuristic was positively associated with perceived ease of use, unstandardized beta (b)=.17, SE=.03, 95% CI: .12, .22. Similarly, positive associations for the color heuristic explained the effect for greater perceived usefulness, b=.17, SE=.02, 95% CI: .12, .21. Controlling for the color heuristic reduced the visual risk indicator effect on perceived usefulness to non-significance, p=.622.

Figure 2.

Heuristic Mediated Effects of Design Cues on Usability and Intentions Outcomes

Note: Path values are unstandardized beta coefficients (b). *p < .05, **p < .01, ***p < .001.

Activation of amount heuristics mediated the effect of ranges on usability, albeit in different directions. The intended amount heuristic (more=bad) was associated with greater perceived ease of use, b=.14, SE=.02, 95% CI: .09, .18, and usefulness, b=.09, SE=.02, 95% CI: .05, .13; however, use of the unintended amount heuristic (more=confusing) was associated with less ease of use, b=−16, SE=.02, 95% CI: −.20, −.11, and usefulness, b=−.08, SE=.02, 95% CI: −.12, −.04. Controlling for amount heuristics, the negative effect of ranges remained statistically significant on perceived ease of use and usefulness, however the total indirect effects were not significant.

3.2. Impact of Webpage Layout

Participants who saw chemicals grouped by health harms (Layout B), were more likely to use a location heuristic (close=related) than people who saw harms separate from chemicals (Layout A), F(1,1324)=276.96, p<.001 (Table 3). Webpages with Layout B (vs. A) led to greater perceived ease of use, F(1,1435)=16.64, p<.001, usefulness, F(1,1435)=15.28, p<.001, and use intentions, F(1,1435)=11.71, p=.001. Layout B (vs. A) had a larger impact among participants with higher, 95% CI for difference: 1.33, 1.67, vs. lower numeracy, 95% CI for difference: .50, .90 for the location heuristic, F(1,1322)=34.77, p<.001, where stratified analyses remained significant, and for perceived ease of use, F(1,1437)=8.41, p=.004, and perceived usefulness, F(1,1437)=11.36, p=.001, where Layout B only impacted those with higher numeracy, 95% CI for difference: .21, .48 and .19, .43 (Table 4). There were no significant interactions between the webpage layout and format, age, or smoking status for any outcomes, all p>.05.

Table 3.

Impact of Layout for Webpages about Chemicals in Cigarette Smoke

Construct (n)	Layout A Health harms at the top M(SD)	Layout B Grouped by the health harms M(SD)	F	η2
Location heuristic (1326)	3.17 (1.50)	4.32 (.99)	277***	.17
Perceived ease of use (1441)	3.67 (1.06)	3.88 (.93)	17***	.01
Perceived usefulness (1441)	4.16 (.92)	4.34 (.81)	15***	.01
Intentions to use the website (1441)	3.17 (1.26)	3.39(1.21)	11**	.01

^**p < .01;

^***p < .001

Table 4.

Impact of Chemical Display and Layout by Numeracy Score

	Higher numeracy					Lower numeracy
				F	${\eta }_p^2$				F	${\eta }_p^2$
Chemical Display	Name only M (SE)	Risk indicator M (SE)	Ranges M (SE)			Name only M (SE)	Risk indicator M (SE)	Ranges M (SE)
Color heuristic (red = danger)	3.54a (.08)	4.54a,b (.07)	3.75b (.08)	47***	.07	3.92c (.09)	4.37c,d (.09)	4.00d (.09)	7.57*	.01
Amount heuristic – unintended (more = confusing)	2.45e (.09)	2.65f (.09)	3.04e,f (.08)	13***	.02	2.67g (.09)	2.55h (.10)	2 79g,h (.09)	1.54	.00
Layout	Layout A M (SD)	Layout B M (SD)			Layout A M (SD)	Layout B M (SD)
Location heuristic	2.91i (.07)	4.42i (.06)	—	272***	.17	3.49j (.07)	4.19j (.07)	—	48***	.04
Perceived ease of use	3.58k (.05)	3.93k (.05)	—	24***	.02	3.79 (.06)	3.82 (.06)	—	.24	.00
Perceived usefulness	4.08l (.04)	4.39l (.04)	—	27***	.02	4.27 (.05)	4.28 (.05)	—	.01	.00

Note. Table shows effect sizes for the analyses where chemical display or layout interacted with numeracy. Means that share a superscript differ by p<.05.

^**p < .01;

^***p < .001

Activation of the location heuristic explained the relationship between layout and usability and intentions to use the webpage. The layout heuristic (close=related) was associated with greater perceived ease of use, b=.24, SE=.02, 95% CI: .21, .29; usefulness, b=.20, SE=.02, 95% CI: .16, .24; and use intentions, b=.29, SE=.03, 95% CI: .24, .34. Controlling for the location heuristic reduced the effect of the layout on perceived ease of use, usefulness, and use intentions to non-significance, all p>.05. The effects of design cues on intentions mediated through heuristic and usability outcomes are in Appendix A.

4. Discussion

Websites can communicate about cigarette chemicals and harms;^48,49 however, to make a public health impact, they must be used. Users quickly evaluate websites,^50–52 where impressions are influential for the perceived usability and use of online health information.^20,53 These findings indicate design cues impact how audiences process and evaluate online tobacco information – with predictable heuristics, or reflexive rules of thumb that bypass (or take the place of) analytical thinking.²²

Layouts with chemicals near associated health harms led to greater perceived ease of use, usefulness, and use intentions. When information placement illustrated associations, adolescents and adults were more likely to engage. Visual processing and evaluations unfold in a series of sensory and cognitive stages.^54–56 Object and group parameter detection happens quickly, often without conscious awareness.^54,57 Objects are then evaluated affectively, aesthetically, and finally, cognitively.^52,58,59 Using website layout to connect chemicals to harms encouraged use of accessible “close=related” heuristics across all participants, and positive downstream usability assessments among those with higher numeracy.

Color-coding of the visual risk indicator gave meaning to the chemical information and activated a “red=danger” heuristic. Repeat exposure codifies the red-danger link – stop signs, sirens, problematic situation descriptors (e.g., red handed) – and contributes to strong, implicit associations.²⁷ Societal conditioning coupled with biological predispositions for danger cues (e.g., angry faces, blood) makes color-coding an effective strategy for risk visualizations. Visual risk indicator webpages were perceived as easier to use and more useful than webpages with ranges, mediated by the color heuristic. Using design cues – color and placement – aids interpretation and improves usability for online engagement. This is promising across user motivations; greater perceived ease of use is influential for hedonic website use, while perceived usefulness is a predictor for utilitarian uses.⁶⁰

Websites with numerical ranges were less easy to use or useful. The impact of two competing heuristics – more=bad versus more=confusing – help explain why. Adolescents and adults endorsed both heuristics when shown ranges. Intended risk intuition was negated by inability, or lack of motivation, to process unfamiliar quantities;^12,30 however, individuals with higher numeracy were the only group to report greater confusion. This mirrors international evidence that adult smokers and nonsmokers find chemical information more comprehensible and informative when harm descriptions (e.g., cause cancer) are given instead of numbers.^33,35,61 Many individuals struggle with complex risk evaluations; using visual aids with clear descriptive text (versus numbers) could improve communication and health decision making.⁶²

Assessing heuristics allows for understanding how designs are interpreted. Individuals often rely on automatic, easily accessible mental short cuts to process, interpret, and potentially act on what they see.^50,54,63 Designs activated intended heuristics with positive downstream effects for perceived usability of and desire to engage with online tobacco education. These strategies may also increase the effectiveness of social media infographics and other visual-based campaigns for tobacco control.

Study strengths include a randomized experiment with a large national sample. This study is limited to a few design strategies to communicate risk and increase acceptance of fictional webpages. Other designs or interactivity could have greater effects on usability and use. Heuristics were assessed with newly developed items for activated mental shortcuts to process a design cue. While participants could disagree to use of heuristics without the cues, it is not known how participants interpreted these items in absence of a cue. Future studies are needed to establish the measurement properties of the heuristic items. With insights for interpretations, this study highlights promising design strategies for online disclosures. Future studies of interactive website are needed to determine if use increases gist risk understanding (e.g., chemicals=dangerous) and discourages smoking among the public, as intended.

5. Conclusion

Tobacco chemical disclosures should be developed with an eye for what to show, and also how to display information – a complex communication challenge. Placing chemicals near health harms increased usability and adoption intentions by allowing viewers to identify relationships. Displaying chemical information with a visual risk indicator increased website usability by activating an accessible red-danger heuristic to process risk. Together, these visualization strategies can optimize chemical disclosures to increase use of educational resources among the public.

Summary.

What was already known

• Communicating about chemicals in tobacco products is a complex challenge due to limited public understanding.

• Chemical disclosures that are easy to interpret are preferred by the public.

What this study added to our knowledge

• Website layout can illustrate the connection of chemicals and health harms.

• Color-coded risk indicators provide meaning for chemical information.

• Layout and color cues increase the usability of online chemical disclosures.

• Heuristics provide insights for public interpretation of chemical disclosure.

Appendix A.

Effects of Design Cues on Intentions Mediated by Heuristic and Usability Outcomes

Note: Path values are unstandardized beta coefficients (b). *p < .05, **p < .01, ***p < .001.