Investigating Single- Versus Multiple-Source Approaches to Communicating Health Messages Via an Online Simulation

Abstract

Purpose:

To assess whether exposing drinkers to information about the alcohol–cancer link via multiple and diverse sources in an online simulation produces larger improvements in attitudes and intentions relative to exposure to a single source of information.

Design:

Experimental; unequal randomization with respondents allocated to either the single-source (20%) or multiple-source condition (80%). Alcohol-related behavioral intentions were assessed preexposure and postexposure.

Setting:

Australia.

Participants:

A total of 2087 drinkers consuming alcohol at least twice per month.

Measures:

Scales were used to assess attitudes toward the messages (believability, convincingness, and personal relevance) and behavioral intentions (extent to which participants believed that they should and would reduce their alcohol consumption and their intention to consume 5 or more drinks in a single session).

Analysis:

Hierarchical linear regression.

Results:

Source condition was significantly associated with all 3 attitudinal variables (P < .001). Those exposed to an alcohol warning statement from multiple sources found the message more believable, convincing, and personally relevant compared to those exposed to a warning statement via a single source. They also reported significantly greater change preexposure to postexposure on the 2 behavioral beliefs that they should (ΔM = 0.25 vs ΔM = 0.09) and would (ΔM = 0.23 vs ΔM = 0.00) reduce their current alcohol consumption (P < .001). Further, those in the multiple-source condition reported reduced intentions to consume 5 or more standard drinks in a single sitting (ΔM = 0.21 vs ΔM = 0.14; P < .001).

Conclusion:

Findings from the online simulation provide support for the suggestion that integrated approaches involving the combination of multiple sources to deliver a message produce superior outcomes compared to relying on a single source (eg, warning labels on alcoholic beverages).

Keywords

alcohol messages online simulation cancer harm reduction health promotion

Purpose

Alcohol is recognized as one of the primary modifiable factors contributing to noncommunicable diseases globally.¹ The costs associated with alcohol-related harm are now comparable to those of tobacco.^2,3 The significant gains made in tobacco control provide evidence that a comprehensive and multifaceted approach to public education that includes on-product warnings, school-based education programs, and publicity campaigns can be effective in increasing awareness and changing health-related attitudes and behaviors.^4

–8 Smoking prevalence in the United States has decreased from about 43% in 1965 to 18% in 2014.⁹ In Australia, the context of the present study, smoking prevalence has decreased from 43% in 1964 to 13% in 2013.^10,11

Although the tobacco experience holds key learnings for alcohol control, convincing the drinking public that alcohol consumption is harmful is likely to be an especially challenging task. In Western society, alcohol remains highly culturally accepted and is considered part of the social fabric.^12

–15 This is reflected in large majorities of adults being classified as current drinkers (eg, 64.9% of American adults¹⁶ and 80.7% of Australian adults¹⁰ report consuming alcohol in the past year).

Evidence relating to the harms associated with even moderate levels of alcohol consumption is relatively recent and follows a widely publicized interpretation that alcohol in moderation can provide positive benefits for health, which has since been disputed.^17

–20 In this context, health communications attempting to convince drinkers that their consumption behaviors could be harmful (ie, messages that use fear appeals) may be potentially greeted with counterproductive psychological reactance in the form of resistance, counterarguments, and noncompliance.^21
–23 It has therefore been proposed that a reduction in alcohol-related harm will require a broad-based and sustained approach involving comprehensive strategies to inform individuals of the risks associated with alcohol consumption and encourage changes in drinking behaviors.^1,24
–26 In particular, there have been calls for warning statements to be placed on alcoholic beverages as an important method of informing individuals of the risks associated with alcohol consumption to enable them to make informed choices.^27

–30

This recommendation for warning labels on products to be just 1 component of a broader public education campaign is supported by the disappointing results of evaluations of the warning labels that have been mandatory on alcoholic beverages in the United States since November 1989. While there were some awareness and attitudinal changes, significant behavioral change did not eventuate.^31

–36 Although this may be at least partially because of the low level of noticeability of the warning labels as implemented by the alcohol industry,³⁷ it is acknowledged that such labels are unlikely to be effective on their own and need to be part of a comprehensive approach to optimize their effects.^28,29

Alcohol and Cancer

Over 25 years ago, the International Agency for Research on Cancer classified alcoholic beverages as a group 1 carcinogen, stating that a variety of malignant tumors are “causally related to the consumption of alcoholic beverages.”³⁸ This classification was confirmed in 2007,³⁹ and recent empirical research has linked even light to moderate alcohol consumption with increased risk of cancer.^40,41 For example, in their study of over 135 000 US adults, Cao et al⁴² found that the risk of alcohol-related cancers in men increases by 6% with 0.5 to 1.5 standard drinks consumed per day and by 26% with 1.5 to 3 standard drinks per day. For women, the figures are 13% and 24%, respectively. Risk is highest for breast cancer, which accounts for all of the increase at the 0.5 to 1.5 standard drinks per day level for women.

Although it is estimated that 4.2% of all deaths caused by cancer are attributable to alcohol consumption,⁴³ there is a lack of public awareness of alcohol-related cancer risk.^{15,44

–48} Drinkers are entitled to information about the health risks associated with alcohol consumption,³⁶ making it important to develop and implement effective methods of educating consumers about the cancer risk associated with alcohol use. There is evidence that media campaigns communicating the link between alcohol consumption and cancer risk can be effective in enhancing intentions to reduce consumption⁴⁹; however, such publicly-funded campaigns need to compete with pervasive and creative alcohol advertising that emphasizes the positive social outcomes of drinking.^50
–52

Health Information Dissemination

Consistent with the tobacco control experience reported above, the behavior change literature emphasizes that a multipronged approach is needed to deliver and reinforce consistent health messages across multiple platforms to enhance comprehension and recall, and encourage attitudinal and behavioral change.^53
–55 Utilizing a combination of multiple sources to deliver a message has been found to produce additive effects on knowledge relative to when relying on a single source.^56,57 Message source refers to the entity delivering the message, which can range from organizations, celebrities, unknown individuals, cartoon characters, or unspecified others.^58
–60 In the context of alcohol warnings, messages can be attributed to entities such as health agencies or government departments³¹ or they can be positioned as if being communicated by an unidentified source.³⁶

However, policy makers need to be assured of the effectiveness of such comprehensive approaches due to the level of resources required. There is little direct evidence of this efficacy in the alcohol domain, an important exception being Kaskutas and Graves’⁶¹ study of the effects of exposure to multiple sources of health information pertaining to alcohol-related risk of birth defects. A linear dose–response relationship was found between the number of message sources to which people reported being exposed and awareness and behavior outcomes. The greater the number of different sources to which respondents were exposed, the higher the likelihood of (1) being aware of the risk of birth defects, (2) having a conversation about drinking during pregnancy, and (3) reducing alcohol consumption due to health concerns. Kaskutas and Graves⁶¹ called for further research into the additive effects that may be created by exposure to a variety of messages disseminated by multiple and diverse sources, such as print media and conversations with health physicians. Although research in the advertising literature also points to the persuasive power of multiple sources compared to single source,⁶² Kaskutas and Graves’ call for further research in the health information dissemination area remains unanswered to date.

The dose–response relationship found by Kaskutas and Graves⁶¹ may be at least partially attributable to information delivered by multiple sources being perceived by recipients as reflecting “different perspectives and independent pools of knowledge” and therefore being “more worthy of diligent consideration” than the same information presented by a single source.⁶³ The reorientation to a new stimulus may capture the attention of message recipients, encourage greater message processing, and be more persuasive.^62,63 However, in a series of student experiments by Harkins and Petty⁶⁴ on the topic of varying examination formats, there was no significant attitudinal change among recipients exposed to multiple sources presenting a single message relative to recipients exposed to a single source presenting a single message. Only recipients exposed to 3 different sources presenting 3 different arguments reported significant attitudinal change.

Present Study

The objective of this study was to compare outcomes from exposure to warning messages relating to the alcohol–cancer link delivered by a single source versus multiple sources to assess whether the latter approach has the potential to generate larger improvements in drinkers’ attitudes and intentions. As the messages do not already exist in reality, it was not possible to replicate Kaskutas and Graves’⁶¹ approach of asking individuals to report the different sources from which they have received relevant messages. To overcome this problem, an online survey (described in detail below) was created that had 2 conditions. One condition involved respondents being exposed to an alcohol warning statement delivered by multiple sources via an online simulation, whereas the other condition exposed respondents to an alcohol warning statement delivered by a single source. Changes in respondents’ alcohol-related attitudes and consumption intentions preexposure and postexposure were assessed. Given previous research indicating the greater efficacy of multiple sources relative to a single source,^61,62 it was hypothesized that respondents exposed to a message relating to the alcohol–cancer link delivered by multiple sources would exhibit greater changes in attitudes and intentions compared to respondents exposed to a message delivered by a single source.

Methods

Design

The present study constitutes the third stage of a multistage study aimed at developing and testing warning statements that can be used to advise drinkers of the cancer risk associated with alcohol consumption.^65,66 In stage 1, 12 warning statements that could be effective in encouraging drinkers to reduce their intake of alcoholic beverages were suggested by participants attending 6 focus groups (n = 48). In stage 2, these statements were assessed on their believability, convincingness, and personal relevance in an online survey of over 2000 Australian drinkers.⁶⁵ The statements differed according to the type of cancer mentioned (eg, breast or bowel cancer vs reference to cancer in general) and the wording used to convey risk (eg, “alcohol can cause cancer” vs “alcohol increases your risk of cancer”). Each respondent was exposed to a randomly allocated series of 3 messages. The results indicated that all the tested statements were generally considered to be believable, convincing, and relevant by both moderate and heavy drinkers.

On the basis of the results of stage 2, a subset of 6 statements (Alcohol increases your risk of bowel cancer; Alcohol increases your risk of breast, bowel, throat, and mouth cancer; Alcohol increases your risk of breast cancer; Warning: Alcohol increases your risk of cancer; Alcohol increases your risk of cancer; Reduce your drinking to reduce your risk of cancer) was selected for further testing in stage 3 to assess behavioral intention outcomes.⁶⁶ In this phase, respondents were only exposed to a single message to enable identification of any changes in drinking intentions resulting from exposure to individual messages. A further objective of this stage was to examine the effectiveness of these messages when delivered by single versus multiple sources, which is the focus of the present study.

Sample

A large web panel provider (PureProfile) was used to recruit 2087 adult Australian drinkers for stage 3 (see Table 1 for sample profile). Panel members either responded to an e-mail inviting them to participate in a study on alcohol or accessed a link to the study that was presented on the panel’s website among other studies available for panel members to complete. Ethics clearance for the study was obtained from a university research ethics committee. Eligibility criteria were being 18 to 65 years of age and average alcohol consumption frequency of at least twice per month. In addition, potential respondents could not have completed the stage 2 online survey. Quotas were used to achieve an even gender split across 3 main age categories (<31 years, 31-45 years, and 46-65 years).

Table 1.

Sample Profile.

	Condition		Total (n = 2087)	Significance
	Multiple Source (n = 1680)	Single Source (n = 407)	Total (n = 2087)	Significance
Gender (%)				χ²(1) = 0.11, P = .739, Φ = .01
Male	50.1	49.1	49.9
Female	49.9	50.9	50.1
Age (mean), years	35.98 (12.74)	36.33 (12.39)	36.05 (12.67)	t(2085) = −0.50, P = .620, d = −0.03
<30 years (%)	49.5	49.4	49.4
31-41 years (%)	25.2	25.3	25.3
46-65 years (%)	25.3	25.3	25.3
Tertiary education (%)				χ²(1) = 0.88, P = .348, Φ = .02
Yes	46.1	43.5	45.6
No	53.9	56.5	54.4
SES (%)^a				χ²(2) = 4.12, P = .127, Φ = .05
Low	28.4	33.4	29.4
Mid	43.3	39.6	42.6
High	28.4	27.0	28.1
Drinks per week (mean)	9.27 (13.60)	12.22 (18.01)	9.85 (14.60)	t(523.43) = −3.09, P = .002, d = −0.19

Abbreviations: Φ, Phi coefficient (measure of effect size); SES, socioeconomic status.

^aMissing values (n = 19; 1.1%) treated listwise.

To ensure that outcome variable differences between the single- and multiple-source conditions were due to the type of condition and not due to other differences in respondents assigned to these groups, a series of checks were conducted comparing the demographic profile of the groups. Table 1 shows that those assigned to each condition did not differ significantly on gender, age, tertiary education, and socioeconomic status (SES). However, a significant difference was noted for the drinks consumed per week variable. Respondents in the simulation condition reported consuming an average of 9.27 drinks per week compared to 12.22 drinks per week among those in the control condition. This difference was controlled for in subsequent regression analyses.

Measures

The survey commenced with respondents being asked about their alcohol consumption using the items included in periodic national alcohol intake surveys.^10,67 Respondents reported on the frequency with which they consumed alcohol in the previous 12 months, the number of standard drinks consumed on a usual drinking occasion, and the largest number of standard drinks consumed during 1 sitting in the previous 12 months. The first 2 items were used to create an average number of drinks consumed per week. Demographic questions (other than those used for initial screening) were also asked in relation to respondent education level and postcode (used to derive SES as per the Australian Bureau of Statistics’ Socio-Economic Indexes for Areas classification⁶⁸).

At the beginning and end of the survey, alcohol consumption intentions were measured with the following items (as per Johnston and White⁶⁹; Sherman et al⁷⁰): To what extent do you think that you, personally, should reduce the amount of alcohol you consume? To what extent do you think that you, personally, will actually reduce the amount of alcohol you consume? and Do you intend to drink 5 or more standard alcoholic beverages in a single session in the next 2 weeks? The extent to which respondents believed they should and will reduce the amount of alcohol they consume was assessed on a 5-point scale of 1 (not at all) to 5 (to a very great extent). Respondents’ intention to consume 5 or more drinks in a single sitting within the following 2 weeks was assessed on a 5-point scale of 1 (definitely intend not to) to 5 (definitely intend to).

Consistent with previous studies examining attitudes to messages in terms of perceived attributes of the messages,^71,72 and particularly those in the tobacco control field,^73,74 the attitudinal outcome variables of believability, convincingness, and personal relevance were assessed. Postexposure, respondents in both conditions were asked about their perceptions of the believability (How believable/unbelievable did you find this message?), convincingness (How convincing/unconvincing did you find this message?), and personal relevance (How much do you feel this message applies to you?) of the warning statement to which they were exposed. Perceived believability and convincingness were measured on a 5-point scale of 1 (not at all believable/convincing) to 5 (very believable/convincing), and personal relevance was measured on a 5-point scale of 1 (it does not apply to me at all) to 5 (it directly applies to me).

Intervention

An unequal randomization (1:4) approach was taken, with respondents allocated to either the single-source (20%) or multiple-source (simulation) condition (80%). This allocation maximized the sample size of the multiple-source condition which, as part of the broader study, was the subject of analyses requiring additional power.⁶⁶ Respondents in both conditions were equally likely to receive each of the 6 warning statements, with respondents randomized to view just one of the statements. In the single-source condition, respondents were randomly exposed to 1 of the 6 statements in plain black text on a white background. To ensure equal exposure to the message between the 2 conditions, the same message was shown on-screen 5 times in succession, with respondents needing to click on the message each time to progress to the next exposure. In the multiple-source condition, respondents entered an online simulation featuring 3 locations: a home living room, a doctor’s office, and a road-side bus stop. Respondents were randomly allocated to commence their simulation experience in either the living room or the doctor’s office and invited to navigate through the 3 locations from this starting point.

While within the simulation, respondents could click on various items in each of the encountered locations. Some of these items were filler items, whereas others produced the statement to which the respondent was randomly allocated. Respondents could not proceed to the next location in the simulation until all items generating the warning statement were selected, thereby ensuring all participants were exposed to the message 5 times from 5 different sources. In the living room location, they were exposed to the statements via an advertisement in a newspaper, a warning on an alcohol product, and a comment made by a child about information learned during a health class at school. The type of alcoholic beverage used to convey the warning statement (beer, wine, spirits/liquor) was determined by the respondent’s previously stated preference to increase the relevance of the specific source. In the bus stop location, respondents were exposed to the statement on a billboard. At the doctor’s office, respondents were exposed to the statement via a message given by a doctor (as suggested by Kaskutas and Graves⁶¹). Across these sources within the simulation, 2 were identifiable (the doctor and the child) and the remainder were presented as originating from an unidentified source. In both conditions, after 5 exposures to the message, respondents were returned to the survey to complete additional items.

Analysis

Hierarchical linear regression analyses were conducted to assess the unique contribution of source condition to each of the attitudinal and behavioral intention outcomes. Given the significant difference between conditions on drinks consumed per week, this variable was entered in block 1 and therefore controlled for in analyses. Where both preexposure and postexposure scores were obtained, preexposure scores were also entered in block 1 of the regression analysis. Condition (1 = single source, 2 = multiple source) was then entered in block 2.

Results

The results relating to the effectiveness of the messages relative to each other have been reported elsewhere.⁶⁶ Overall, the Alcohol increases your risk of bowel cancer message was found to be most effective for both the single- and multiple-source conditions.

Attitudinal Outcomes

Table 2 presents means and standard deviations for the outcome variables of believability, convincingness, and personal relevance across conditions. As per the previous phase of the larger study, the messages had average scores above the midpoint across both conditions, suggesting that respondents found them to be generally appropriate. Those in the single-source condition demonstrated the lowest levels of perceived believability, convincingness, and personal relevance.

Table 2.

Believability, Convincingness, and Personal Relevance for Total Sample and by Condition.

Variable	Total Mean (SD), N = 2087	Multiple Source Mean (SD), n = 1680	Single source Mean (SD), n = 407
Believability^a	3.76 (1.03)	3.80 (1.00)^b	3.59 (1.11)
Convincingness^a	3.52 (1.09)	3.58 (1.06)^b	3.29 (1.15)
Personal relevance^c	3.01 (1.20)	3.06 (1.19)^b	2.83 (1.24)

Abbreviation: SD, standard deviation.

^aFive-point scales of 1 (not at all believable/convincing) to 5 (very believable/convincing).

^bSignificantly different from single-source condition at P < .001.

^cFive-point scale of 1 (it does not apply to me at all) to 5 (it directly applies to me).

Table 3 presents block 2 results from the hierarchical linear regression analyses conducted to determine whether a significant difference arose on each of the attitudinal variables as a result of source condition. Controlling for preexposure differences in the amount of alcohol consumed per week, source condition was significantly and positively associated with all 3 attitudinal variables. Those exposed to the messages via multiple sources found the messages more believable, convincing, and personally relevant compared to those exposed to the message via a single source.

Table 3.

Hierarchical Linear Regression Analysis Determining Influence of Condition on Attitudes.^a,b

Outcome	Predictor	b	SE	95% CI for b	β	P	Part r²
Believability (ΔR2 = .006)	Drinks per week	.00	0.00	0.00-0.00	−.01	.694	−.01
Believability (ΔR2 = .006)	Condition	.21	0.06	0.10-0.32	.08	<.001	.08
Convincingness (ΔR2 = .011)	Drinks per week	.00	0.00	0.00-0.00	.00	.844	.00
Convincingness (ΔR2 = .011)	Condition	.29	0.06	0.18-0.41	.11	<.001	.11
Personal relevance (ΔR2 = .008)	Drinks per week	.02	0.00	0.01-0.02	.20	<.001	.20
Personal relevance (ΔR2 = .008)	Condition	.27	0.07	0.15-0.40	.09	<.001	.09

Abbreviations: b, unstandardized regression coefficient; β, standardized regression coefficient; CI, confidence interval; Part r2, individual variance explained; R², total variance explained; SE, standard error.

^aBlock 1 is not presented in this table.

^bCondition dummy coding: 1 = single source, 2 = multiple source.

Behavioral Intention Outcomes

Changes in behavioral intentions as a result of exposure to the warning statements were examined overall and by condition. Table 4 presents preexposure and postexposure results for respondents’ beliefs relating to the extent to which they should and would reduce their alcohol consumption and their intentions to consume 5 or more drinks in a single sitting. There were significant improvements in pre–post scores for all outcome measures, with the exception of respondents in the single-source condition who reported no change in the extent to which they believed they would reduce their consumption.

Table 4.

Behavioral Intentions Premessage and Postmessage Exposure for Total Sample and by Condition.

Variable	Total (N = 2087)				Multiple Source (n = 1680)				Single Source (n = 407)
Variable	Pre-Mean (SD)	Post-Mean (SD)	Δ	d	Pre-Mean (SD)	Post-Mean (SD)	Δ	d	Pre-Mean (SD)	Post-Mean (SD)	Δ	d
Believe should reduce^a	2.68 (1.31)	2.93 (1.33)	0.25	−0.28^b	2.66 (1.29)	2.96 (1.32)	0.30	−0.34^b	2.75 (1.39)	2.84 (1.35)	0.09	−0.11^c
Believe will reduce^a	2.49 (1.18)	2.72 (1.22)	0.23	−0.27^b	2.48 (1.17)	2.76 (1.21)	0.28	−0.32^b	2.56 (1.23)	2.56 (1.21)	0.00	0.00
Intent to consume 5+ drinks^d	2.70 (1.50)	2.49 (1.40)	0.21	0.23^b	2.64 (1.48)	2.41 (1.37)	0.23	0.25^b	2.93 (1.56)	2.79 (1.49)	0.14	0.18^b

Abbreviations: SD, standard deviation; d, effect size; Δ, pre-mean to post-mean change.

^aFive-point scale of 1 (not at all) to 5 (to a very great extent).

^bP < .001.

^cP < .05.

^dFive-point scale of 1 (definitely intend not to) to 5 (definitely intend to).

^eP < .01.

Table 5 presents results from the hierarchical linear regression analyses conducted to determine whether a significant difference in behavioral intentions arose as a result of source condition. Those in the multiple-source condition reported significantly greater change preexposure to postexposure, including when controlling for preexposure differences in the amount of alcohol consumed per week between those assigned to each of the conditions.

Table 5.

Hierarchical Linear Regression Analysis Determining Influence of Condition on Behavioral Intentions.^a,b

Outcome	Predictor	b	SE	95% CI for b	β	P	Part r²
Post-believe should reduce (ΔR2 = .004)	Pre-believe should reduce	0.77	0.02	0.74 to 0.80	.76	<.001	.71
	Drinks per week	0.00	0.00	0.00 to 0.01	.05	.001	.04
	Condition	0.21	0.05	0.12 to 0.30	.06	<.001	.06
Post-expect will reduce (ΔR2 = .007)	Pre-expect will reduce	0.77	0.02	0.74 to 0.80	.74	<.001	.74
	Drinks per week	0.00	0.00	0.00 to 0.00	.02	.309	.02
	Condition	0.27	0.05	0.18 to 0.35	.09	<.001	.09
Post-intent to consume 5+ drinks (ΔR2 = .002)	Pre-intent to consume 5+ drinks	0.73	0.01	0.70 to 0.75	.78	<.001	.71
	Drinks per week	0.01	0.00	0.00 to 0.01	.06	<.001	.06
	Condition	−0.16	0.05	−0.25 to −0.07	−.05	<.001	−.05

Abbreviations: b, unstandardized regression coefficient; β, standardized regression coefficient; CI, confidence interval; Part r², individual variance explained; R², total variance explained; SE, standard error.

^aBlock 1 is not presented in this table.

^bCondition dummy coding: 1 = single source, 2 = multiple source.

Discussion

In light of calls for comprehensive strategies to increase public awareness of alcohol-related harms and encourage behavioral change,^75,76 the present study compared alcohol-related beliefs and intentions after exposure to alcohol warning messages delivered by a single source versus multiple sources. Given previous research,^61,62 it was hypothesized that respondents exposed to a warning message delivered by multiple sources would exhibit greater changes in attitudes and behavioral intentions compared to respondents exposed to the same message delivered by a single source.

Supporting this hypothesis, respondents exposed to an alcohol warning statement presented by multiple sources found the statement more believable, convincing, and personally relevant compared to those exposed to the statement via a single source. Similarly, those in the multiple-source condition were more likely to report that they should and would reduce their current alcohol consumption and had reduced intentions to consume 5 or more standard drinks in a single sitting. This stands in contrast to Harkins and Petty’s⁶⁴ results where it was found that significant attitudinal change in message recipients was only evident in those exposed to 3 different sources presenting 3 different arguments. Unlike the present study, they did not find significant change in those exposed to a single message presented by multiple sources. However, Harkins and Petty⁶⁴ exposed respondents to a video tape featuring different individuals presenting non–health-related information, whereas the current study presented health-related information to respondents via a range of information delivery mechanisms that more closely resemble the range of mediums available to policy makers. These different information topics and presentation methods may account for the varied results. Of note is that the present study was not intended to be a replication of Harkins and Petty⁶⁴ and, as such, the 2 additional conditions assessed by Harkins and Petty (multiple messages from a single source and multiple messages from multiple sources) were not included. Given the value of rotating alcohol warnings highlighted in the literature,³⁶ extending the present methodology to assess the outcomes of individuals’ exposure to multiple messages would be a valuable next step in progressing this research.

The present findings provide support for the suggestion that comprehensive approaches involving the use of multiple sources to deliver messages may produce additive effects.^56,57 As noted in the Introduction, such effects appear to have occurred in the context of tobacco control, where the use of consistent messages delivered via multiple pathways has resulted in significant attitudinal and behavioral change.^{4

–8,50,56,57} Similarly, efforts to increase awareness of the alcohol–cancer link need to be part of a comprehensive program that educates consumers about the cancer-related harms associated with alcohol use through multiple information pathways.³⁰

Although overall effect sizes for the influence of condition were small, this is consistent with findings from previous advertising research that emphasizes the importance of repetition in producing behavioral change,⁷⁷ and hence the likelihood of small effects resulting from low levels of exposure.⁷⁸ However, at a population level, these small effects have the potential to aggregate to meaningful changes in overall levels of alcohol consumption and alcohol-related harm.

This study has several limitations. First, the online simulation used in the multiple-source condition only partially reflects real life and it therefore cannot be assumed that the results are directly translatable. This simulation required respondents to interact with the stimuli in a way that is unlikely to occur in the real world (eg, by clicking on stimuli to generate an alcohol warning). This may have produced greater effects than would occur due to more ambient exposure in the real world because of the greater attentional allocation. However, differences in the level of interaction between conditions were minimized by having respondents in both the single-source and multiple-source conditions interact with the content by clicking.

A second limitation concerns the use of a web panel provider to recruit the sample of drinkers and the ability of respondents to self-select to participate in the study via the PureProfile portal. Although quota sampling was employed, population representativeness cannot be assumed and it was not possible to estimate response rates due to the range of strategies used by the panel provider to recruit respondents. However, the large sample size and sample diversity assist in minimizing these limitations. Third, the sample was confined to those consuming alcohol at least twice per month. Future research could consider including lighter drinkers (ie, fewer than 2 drinks per month) and nondrinkers to examine whether the delivery of warning statements across multiple sources also contributes to the reinforcement of nonconsumption. Finally, the present study reports on findings from an Australian sample of drinkers. Additional work is required to assess whether the results translate to other Western societies where alcohol consumption is similarly culturally embedded.

In conclusion, the results of the present study provide preliminary support for the utility of a comprehensive approach to warning message dissemination rather than reliance on a single mechanism of communication. This finding is important in the context of alcohol harm information being typically relegated to a limited number of communication methods compared to the well-resourced, creative, and ubiquitous advertising messages the alcohol industry uses to endorse alcohol consumption. Respondents assigned to the multiple-source condition in an online simulation found messages about the alcohol–cancer risk significantly more believable, convincing, and personally relevant compared to respondents assigned to a single-source condition. Increases in the extent to which respondents believed they should and will reduce the amount they drink and decreases in their intention to consume 5 or more drinks in a single sitting were found postexposure in both conditions, with significantly greater change demonstrated in the multiple-source condition. These results may be of value to policy makers and health agencies in developing effective interventions and to advocacy groups in their efforts to encourage greater action to address alcohol-related harms.

SO WHAT?

What is already known on this topic?

Utilizing a combination of sources to deliver a message has been found to produce additive effects on knowledge relative to when relying on a single source.

What does this article add?

Respondents assigned to the multiple-source condition found messages about the cancer risks associated with alcohol consumption to be significantly more believable, convincing, and personally relevant compared to respondents assigned to a single-source condition. They also reported significantly greater change in behavioral intentions preexposure to postexposure.

What are the implications for health promotion practice or research?

Efforts to increase awareness of the alcohol–cancer link need to be part of a comprehensive program that involves multiple information pathways. These results may be of value to policy makers and health agencies in developing effective multicomponent interventions to address alcohol-related harms.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by the Western Australian Health Promotion Foundation (Healthway), research grant #20338.

References

World Health Organization. Global Strategy to Reduce the Harmful use of Alcohol. Geneva, Switzerland: WHO; 2010.

Bouchery

Harwood

Sacks

Simon

Brewer

. Economic costs of excessive alcohol consumption in the US, 2006. Am J Prev Med. 2011;41(5):516–524.

Scarborough

Bhatnagar

Wickramasinghe

Allender

Foster

Rayner

. The economic burden of ill health due to diet, physical inactivity, smoking, alcohol and obesity in the UK: an update to 2006-07 NHS costs. J Public Health (Oxf). 2011;33(4):527–535.

Levy

Chaloupka

Gitchell

. The effects of tobacco control policies on smoking rates: a tobacco control scorecard. J Public Health Manag Pract. 2004;10(4):338–353.

National Cancer Institute. The Role of the Media in Promoting and Reducing Tobacco Use (Tobacco Control Monograph No. 19). Bethesda, MD: U.S. Department of Health and Human Services NIoH, National Cancer Institute; 2008.

Siegel

. The effectiveness of state-level tobacco control interventions: a review of program implementation and behavioral outcomes. Annu Rev Public Health. 2002;23(1):45–71.

Wakefield

Durkin

Spittal

. Impact of tobacco control policies and mass media campaigns on monthly adult smoking prevalence. Am J Public Health. 2008;98(8):1443–1450.

Wakefield

Loken

Hornik

. Use of mass media campaigns to change health behaviour. Lancet. 2010;376(9748):1261–1271.

US Department of Health and Human Services. The Health Consequences of Smoking-50 Years of Progress. A Report of the Surgeon General. Atlanta, GA: US Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health; 2014.

10.

Australian Institute of Health and Welfare. National Drug Strategy Household Survey Detailed Report: 2013. Drug Statistics Series no 28. Cat. no. PHE 183. Canberra, Australia: Australian Institute of Health and Welfare; 2014.

11.

Woodward

. Trends in cigarette consumption in Australia. Aust NZ J Med. 1984;14(4):405–407.

12.

Australian Institute of Health and Welfare. Australia’s Health 2012. Canberra, Australia: Australian Institute of Health and Welfare; 2012.

13.

Jones-Webb

Nelson

McKee

Toomey

. An implementation model to increase the effectiveness of alcohol control policies. Am J Health Promot. 2014;28(5):328–335.

14.

Payne

Watkins

Jones

. Midwives’ knowledge, attitudes and practice about alcohol exposure and the risk of fetal alcohol spectrum disorder. BMC Pregnancy Childbirth. 2014;14(1):377.

15.

Rehm

Lachenmeier

Room

. Why does society accept a higher risk for alcohol than for other voluntary or involuntary risks? BMC Med. 2014;12:189.

16.

Schoenborn

Adams

Peregoy

. Health Behaviors of Adults: United States, 2008–2010. U.S. Department of Health and Human Services, Centre for Disease Control and Prevention, National Center for Health Statistics; 2013.

17.

Chikritzhs

Stockwell

Naimi

Andreasson

Dangardt

Liang

. Has the leaning tower of presumed health benefits from ‘moderate’ alcohol use finally collapsed? Addiction. 2015;110(5):726–727.

18.

Fernández-Solà

. Cardiovascular risks and benefits of moderate and heavy alcohol consumption. Nat Rev Cardiol. 2015;12(10):576–587.

19.

Fillmore

Kerr

Stockwell

Chikritzhs

Bostrom

. Moderate alcohol use and reduced mortality risk: systematic error in prospective studies. Addict Res Theory. 2006;14(2):101–132.

20.

Liang

Chikritzhs

. Observational research on alcohol use and chronic disease outcome: new approaches to counter biases. Sci World J. 2013;2013:860915.

21.

Dillard

Shen

. On the nature of reactance and its role in persuasive health communication. Commun Monogr. 2005;72(2):144–168.

22.

Kingsbury

Gibbons

Gerrard

. The effects of social and health consequence framing on heavy drinking intentions among college students. Brit J Health Psych. 2015;20(1):212–220.

23.

Ringold

. Boomerang effects in response to public health interventions: some unintended consequences in the alcoholic beverage market. J Consum Policy. 2002;25(1):27–63.

24.

Kaskutas

Greenfield

. First effects of warning labels on alcoholic beverage containers. Drug Alcohol Depen. 1992;31(1):1–14.

25.

Stockwell

. A review of research into the impacts of alcohol warning labels on attitudes and behaviour. Centre for Addictions Research of BC, University of Victoria, British Columbia, Canada. http://www.uvic.ca/research/centres/carbc/assets/docs/report-impacts-alcohol-warning-labels.pdf 2006. Accessed September 18, 2015.

26.

World Health Organization. WHO expert committee on problems related to alcohol consumption—second report. 2007.

27.

Al-hamdani

. The case for stringent alcohol warning labels: lessons from the tobacco control experience. J Public Health Pol. 2014;35(1):65–74.

28.

Blewett

Goddard

Pettigrew

Reynolds

Yeatman

. Labelling Logic. Commonwealth of Australia; 2011.

29.

Martin-Moreno

Harris

Breda

Møller

Alfonso-Sanchez

Gorgojo

. Enhanced labelling on alcoholic drinks: reviewing the evidence to guide alcohol policy. Eur J Public Health. 2013;23(6):1082–1087.

30.

Pettigrew

Jongenelis

Chikritzhs

. Developing cancer warning statements for alcoholic beverages. BMC Public Health. 2014;14:786–795.

31.

Greenfield

Graves

Kaskutas

. Long-term effects of alcohol warning labels: findings from a comparison of the United States and Ontario, Canada. Psychol Market. 1999;16(3):261–282.

32.

Hilton

. An overview of recent findings on alcoholic beverage warning labels. J Public Pol Mark. 1993;12(1):1–9.

33.

MacKinnon

Nohre

Cheong

Stacy

Pentz

. Longitudinal relationship between the alcohol warning label and alcohol consumption. J Stud Alcohol. 2001;62(2):221–227.

34.

MacKinnon

Lapin

. Effects of alcohol warnings and advertisements: a test of the boomerang hypothesis. Psychol Mark. 1998;15(7):707–736.

35.

Scammon

Mayer

Smith

. Alcohol warnings: how do you know when you have had one too many? J Public Pol Mark. 1991;10(1):214–228.

36.

Wilkinson

Room

. Warnings on alcohol containers and advertisements: international experience and evidence on effects. Drug Alcohol Rev. 2009;28(4):426–435.

37.

Laughery

Young

Vaubel

Brelsford

Jr . The noticeability of warnings on alcoholic beverage containers. J Public Pol Mark. 1993;12(1):38–56.

38.

International Agency for Research on Cancer. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans: Alcohol Drinking. Lyon, France: International Agency for Research on Cancer; 1988.

39.

Baan

Straif

Grosse

. Carcinogenicity of alcoholic beverages. Lancet Oncol. 2007;8(4):292–293.

40.

Bagnardi

Rota

Botteri

. Alcohol consumption and site-specific cancer risk: a comprehensive dose-response meta-analysis. Br J Cancer. 2015;112(3):580–593.

41.

Klatsky

Udaltsova

Baer

Tran

Friedman

. Moderate alcohol intake and cancer: the role of underreporting. Cancer Causes Control. 2014;25(6):693–699.

42.

Cao

Willett

Rimm

Stampfer

Giovannucci

. Light to moderate intake of alcohol, drinking patterns, and risk of cancer: results from two prospective US cohort studies. BMJ. 2015;351:h4238.

43.

World Health Organization. World Cancer Report 2014. http://www.iarc.fr/en/publications/books/wcr/wcr-order.php. Accessed September 10, 2015.

44.

Bowden

Delfabbro

Room

Miller

Wilson

. Alcohol consumption and NHMRC guidelines: has the message got out, are people conforming and are they aware that alcohol causes cancer? Aust N Z J Public Health. 2014;38(1):66–72.

45.

Hawkins

Berkowitz

Peipins

. What does the public know about preventing cancer? Results from the Health Information National Trends Survey (HINTS). Health Educ Behav. 2010;37(4):490–503.

46.

Redeker

Wardle

Wilder

Hiom

Miles

. The launch of Cancer Research UK’s ‘Reduce the Risk’ campaign: baseline measurements of public awareness of cancer risk factors in 2004. Eur J Cancer. 2009;45(5):827–836.

47.

Sanderson

Waller

Jarvis

Humphries

Wardle

. Awareness of lifestyle risk factors for cancer and heart disease among adults in the UK. Patient Educ Couns. 2009;74(2):221–227.

48.

Thomson

Vandenberg

Fitzgerald

. An exploratory study of drinkers views of health information and warning labels on alcohol containers. Drug Alcohol Rev. 2012;31(2):240–247.

49.

Dixon

Pratt

Scully

. Using a mass media campaign to raise women’s awareness of the link between alcohol and cancer: cross-sectional pre-intervention and post-intervention evaluation surveys. BMJ Open. 2015;5(3):e006511

50.

Agostinelli

Grube

. Alcohol counter-advertising and the media. Alcohol Res Health. 2002;26(1):15–21.

51.

Jernigan

. The global alcohol industry: an overview. Addiction. 2009;104(suppl 1):6–12.

52.

Pettigrew

Roberts

Pescud

Chapman

Quester

Miller

. The extent and nature of alcohol advertising on Australian television. Drug Alcohol Rev. 2012;31(6):797–802.

53.

Andreasen

. Marketing social marketing in the social change marketplace. J Public Pol Mark. 2002;21(1):3–13.

54.

Grier

Bryant

. Social marketing in public health. Annu Rev Public Health. 2005;26:319–339.

55.

Pechmann

Reibling

. Anti-smoking advertising campaigns targeting youth: case studies from USA and Canada. Tob Control. 2000;9(suppl 2): ii18–ii31.

56.

Brennan

Durkin

Cotter

Harper

Wakefield

. Mass media campaigns designed to support new pictorial health warnings on cigarette packets: evidence of a complementary relationship. Tob Control. 2011;20(6):412–418.

57.

Thrasher

Murukutla

Pérez-Hernández

. Linking mass media campaigns to pictorial warning labels on cigarette packages: a cross-sectional study to evaluate effects among Mexican smokers. Tob Control. 2012;22(e1):e57–e65.

58.

Kang

Herr

. Beauty and the beholder: toward an integrative model of communication source effects. J Consum Res. 2006;33(1):123–130.

59.

Quester

Pettigrew

Hawkins

. Consumer Behaviour: Implications for Marketing Strategy. Sydney, Australia: McGraw-Hill; 2011.

60.

Smith

Bauman

McKenzie

Thomas

. Awareness of message source and its association with the impacts of sun protection campaigns in Australia. Health Educ. 2005;105(1):42–52.

61.

Kaskutas

Graves

. Relationship between cumulative exposure to health messages and awareness and behavior-related drinking during pregnancy. Am J Health Promot. 1994;9(2):115–124.

62.

Pettit-O’Malley

Bozman

. Friendly-contention ads using multiple endorsers: assessing source effects on attribute perceptions and brand attitudes. J Appl Bus Res. 2011;18(3):33–44.

63.

Harkins

Petty

. Information utility and the multiple source effect. J Pers Soc Psychol. 1987;52(2):260–268.

64.

Harkins

Petty

. Effects of source magnification of cognitive effort on attitudes: an information-processing view. J Pers Soc Psychol. 1981;40(3):401–413.

65.

Pettigrew

Jongenelis

Chikritzhs

. Developing cancer warning statements for alcoholic beverages. BMC Public Health. 2014;14:786–795.

66.

Pettigrew

Jongenelis

Glance

. The effect of cancer warning statements on alcohol consumption intentions. Health Educ Res. 2016;31(1):60–69.

67.

Australian Institute of Health and Welfare. 2010 National Drug Strategy Household Survey Report Drug Statistics Series No. 25. Canberra, Australia: AIHW; 2011.

68.

Australian Bureau of Statistics. SEIFA Census of Population and Housing: Socio-Economic Indexes for Areas (SEIFA), Australia, 2011. Canberra, Australia: Australian Bureau of Statistics; 2011.

69.

Johnston

White

. Binge-drinking: a test of the role of group norms in the theory of planned behaviour. Psychol Health. 2003;18(1):63–77.

70.

Sherman

Nelson

Steele

. Do messages about health risks threaten the self? Increasing the acceptance of threatening health messages via self-affirmation. Pers Social Psychol B. 2000;26(9):1046–1058.

71.

Drolet

Williams

Lau-Gesk

. Age-related differences in responses to affective vs. rational ads for hedonic vs. utilitarian products. Mark Lett. 2007;18(4):211–221.

72.

Shao

Grace

Ross

. Self-regulatory focus and advertising effectiveness. Mark Intel Plan. 2015;33(4):612–632.

73.

Beltramini

. Perceived believability of warning label information presented in cigarette advertising. J Advertising. 1988;17(2):26–32.

74.

Hwang

. Selective exposure and selective perception of anti-tobacco campaign messages: the impacts of campaign exposure on selective perception. Health Commun. 2010;25(2):182–190.

75.

Henley

Kanny

Roland

. Alcohol control efforts in comprehensive cancer control plans and alcohol use among adults in the USA. Alcohol Alcoholism. 2014;49(6):661–667.

76.

Jones

. Using social marketing to create communities for our children and adolescents that do not model and encourage drinking. Health Place. 2014;30:260–269.

77.

D’Souza

Rao

. Can repeating an advertisement more frequently than the competition affect brand preference in a mature market? J Mark. 1995;59(2):32–42.

78.

Pettigrew

Tarabashkina

Roberts

Quester

Chapman

Miller

. The effects of television and internet food advertising on parents and children. Public Health Nutr. 2013;16(12):2205–2212.