1. Introduction
Obesity is a significant risk factor for a range of noncommunicable diseases, from cardiovascular diseases to diabetes and even some cancers [
1,
2,
3]. The global obesity burden is growing, not just in high-income countries but also in low- and middle-income countries such as South Africa, [
1] where almost 40% of women and 11% of men are obese [
4].
Unhealthy diets have been found to be a chief contributor to rising obesity rates [
5,
6,
7]. In particular, excess sugar consumption, often in the form of sugary drinks, is a major cause of obesity. It also increases the risk of diabetes, liver and kidney damage, heart disease, and some cancers [
8]. Nevertheless, consumption of sugary drinks remains high globally [
9], including in South Africa, where the rates of consumption have grown in both urban and rural areas [
10,
11,
12,
13].
Taxes on sugar-sweetened beverages (SSBs) that increase the price of the beverage and thereby reduce its accessibility, have emerged as an effective policy solution to counter increased consumption [
14,
15,
16]. This has been demonstrated not only in modeling studies but also in real-world evaluations [
17,
18,
19,
20,
21,
22]. Even so, SSB taxes continue to face significant opposition from the food and beverage industry [
23], potentially slowing their global adoption as an obesity prevention strategy [
24,
25]. Hence, as the South African government prepared to introduce a sugary drinks tax (also described as a Health Promotion Levy) in 2016, industry pressure against the tax was anticipated, and civil society organizations prepared to counter this pressure [
26,
27,
28,
29].
2. Materials and Methods
2.1. South Africa’s “Are You Drinking Yourself Sick?” Campaign
In 2016, the Healthy Living Alliance (HEALA), a coalition of civil society organizations in South Africa committed to supporting policies to improve health and nutrition, with support from Vital Strategies, a global health organization, launched a campaign on the health harms of sugary drink consumption. Targeted at South African adults ages 18 to 45, this campaign sought to build the public agenda for obesity prevention by (a) shifting attitudes toward sugary drinks, (b) building personal risk perceptions of the health harms of consuming sugary drinks, and (c) building public support for strong government action, in particular a tax on sugary drinks.
With support from Vital Strategies, HEALA pretested public service announcement (PSA) concepts with the target audience, including several PSAs from the United States (U.S.) and Mexico, to ensure their adaptability, cultural appropriateness, and potential effectiveness. In particular, noting the potentially stigmatizing nature of messages on obesity prevention [
49], special attention was paid to ensuring that the PSAs did not stigmatize overweight or obese individuals. After pretesting, “Are You Drinking Yourself Sick?” was selected as the campaign’s core message and title. The campaign consisted of two 30-s PSAs that use family- and child-focused messaging. The first PSA, launched in October 2016, features a father and daughter drinking a sugary drink. As they swallow, we see via animation how sugar is dumped into the bloodstream with every sip and leads to fat buildup in and around vital organs bringing on obesity, heart disease, and type 2 diabetes. The second ad, “Journey,” launched in May 2017, follows a woman throughout her day as she drinks several sugary drinks. The PSA reinforces that the amount of sugar she consumed over one day—50 teaspoons—increases the risk of diseases that can take people away from their families and loved ones.
From October 2016 to June 2017 (with the exception of late December to early January), the campaign PSAs ran on multiple mass media outlets, including television and radio; outdoor (billboards, posters) and print (newspaper) media advertisements followed. The television and radio PSAs were broadcast in English and indigenous languages, isiZulu and isiXhosa, on major television and radio channels. Outdoor advertisements consisted of billboards and posters with an image of a child holding her hand up to stop someone from giving her a sugary drink that included the key messages “You wouldn’t give your child 10 spoons of sugar, would you?” and “Type 2 diabetes? No thanks” and provided additional information on the health risks of sugary drinks, such as “Sugary drinks are dangerous” and “Sugary drinks lead to fat build-up in and around vital organs, bringing on obesity, type 2 diabetes, and heart disease” (
Figure 1 and
Figure 2). In addition, another set of outdoor media paired the image of the child with tax-focused messaging that called for support for the sugary drinks tax, including “Support the sugary drinks tax and support our kid’s health” (
Figure 3). The outdoor advertisements were placed mainly in urban areas in three provinces and key political locations, including the path between the national Parliament in Cape Town and the capital, Pretoria. Newspaper advertisements included key messages “Whose side are you on?” and “10 teaspoons of sugar,” encouraging policymakers to support the tax in order to support children’s health, address industry opposition to the tax, and urgently address rising diabetes rates.
The media campaign was a significant part of a broader effort, which also included evidence-based research, a strategic advocacy campaign, and stakeholder and community engagement, to build political and public support for a tax on sugary beverages in South Africa [
50]. The evidence-based research was conducted by a research-to-policy unit at the University of the Witwatersrand, named PRICELESS (or “Priority Cost Effective Lessons for System Strengthening South Africa”). Backed by this research, PRICELESS, the Public Health Association of South Africa (PHASA), HEALA, and other advocates of the tax invested in a strategic advocacy campaign, which included stakeholder and community engagement, to garner support for a sugary drinks tax from government officials, civil society and academics. An open letter in support of the sugary drinks tax was published in the Sunday Times, and multiple public hearings with presentations in support of taxing sugary beverages were held. As with the mass media campaign, these efforts sought to encourage national conversations about the health harms of sugary beverages and build political and public support for the sugary drinks tax to address obesity.
A campaign evaluation was subsequently undertaken to measure the campaign’s effectiveness in increasing knowledge, shifting attitudes, and building support for the SSB tax. Based on the literature, behavioral changes were not anticipated from a campaign of relatively short duration, but they were nonetheless assessed. Campaign impact was measured in representative cross-sectional face-to-face household surveys of adults ages 18 to 56 to measure campaign-related changes in knowledge, attitudes, and behaviors toward sugary drinks and support for government action, particularly a sugary drinks tax to curb the rising obesity epidemic in the country. The surveys were conducted by Genesis Analytics just before the launch of the campaign (the “pre-campaign” survey) from October 7 to 10, 2016, and immediately following its conclusion (the “post-campaign” survey) from July 12 to 21, 2017.
2.2. Sample and Data Collection
The pre- and post-campaign surveys were conducted face-to-face in cities in three provinces: Gauteng, Kwa-Zulu Natal, and Western Cape. These provinces were chosen purposively in areas where the campaign was confirmed to have aired according to the media planners’ reports. Within the provinces, a multistage probability sampling procedure was implemented. Metro areas and cities within the survey provinces were first chosen. Households in each survey site were then selected from the Nielsen GeoFrame, a database of 6 million addresses arranged alphabetically by suburb and within suburb by street name. At each household, a screener questionnaire was implemented, and a Politz grid was used to select one respondent from multiple eligible respondents. For unavailable households, three callbacks on different days were made per household; thereafter, substitution was allowed from households in the immediate vicinity. Respondent eligibility criteria included ages between 18 and 56 years and resident in that location for over six months. The final samples in both the pre- and post-campaign surveys included 1,000 respondents.
2.3. Questionnaire and Measures
The questionnaire was administered face-to-face in English, Afrikaans, Zulu, Sotho, Xhosa, and Tswana via computer-assisted personal interviewing (CAPI). The following were the key measures in the questionnaire (the order of presentation in the questionnaire differed from what is presented here in order to minimize order effects).
Recall of messages on the harms of sugary drinks was measured in two ways: first, respondents were asked to recall any messages they had seen in the prior three months about the health harms of sugary drinks. They were also asked to indicate where they had come across those messages and what main messages they recalled. Later in the interview, campaign ad recognition was measured by showing participants images from the “Are You Drinking Yourself Sick?” campaign, including images from the TV PSA.
Reactions to the campaign were then assessed by asking campaign-aware respondents how strongly they agreed or disagreed with a series of statements about the campaign, including that it was (1) believable; (2) relevant; (3) taught something new; (4) created concern about the harms of sugary drinks; (5) increased interpersonal communication about the harms of sugary drinks; and (6) increased support for government action. Each of these statements was evaluated separately and strength in agreement or disagreement for each statement was measured using a five-point Likert scale.
Knowledge about obesity, and sugary drinks as a risk factor for obesity, was assessed through a series of yes or no questions that probed respondents’ knowledge of the main contributors of obesity and of the illnesses that might be caused by obesity or sugary drinks. Attitudes toward obesity and sugary drinks were assessed by measuring respondents’ agreement (on five-point Likert scales) with a series of statements on obesity and sugary drinks. Specifically, the questions assessed respondents’ attitudes on: the extent to which they considered obesity/sugary drinks to be a serious public health problem in South Africa; their attitudes toward sugary drinks as a cause of obesity; and the role of advertising and food labels on levels of sugary drinks consumption among adults and children.
A series of questions was asked to assess support toward government action to reduce obesity, including a sugary drinks tax. Respondents were asked how strongly they agreed or disagreed with a series of statements about government action to reduce obesity. Strength in agreement or disagreement for each statement was measured using a five-point Likert scale. Support for a sugary drinks tax itself was assessed in two ways: first, participants were asked how strongly they support or oppose a tax on sugary drinks if the money collected was to be invested in public programs. Later, they were specifically asked about the South African proposal as follows: “As part of its plan to address obesity in South Africa, the Department of Health recommends increasing the tax on sugary drinks, and the Department of Finance/Treasury has proposed a tax on sugar content that amounts to a roughly 20% tax on sugary drinks. Do you support or oppose the government’s proposal to tax sugary drinks?”
Behaviors pertaining to sugary drinks and intentions regarding their consumption in the future were measured.
Sociodemographic information was also measured, including gender, age, socioeconomic status (low, medium, and high), and parent/primary caregiver status (are you a parent or primary caregiver to children under the age of 16 who reside with you?) in the screener questionnaire. In the main survey, additional sociodemographic information was collected, including frequency of vigorous physical activity (vigorous physical activity three days or more a week—defined as activities that make people breathe much harder than normal and may include heavy lifting, digging, aerobics, or fast bicycling and only those that they did for at least 10 min at a time), fruit and vegetable intake in the last seven days (consumed fruits and vegetables over three times a week), and frequency of television watching (watched more than four hours weekly).
2.4. Data Analysis
Data were weighted to adjust for oversampling or any mismatch between the sample profile and the estimated universe. The source of weighting was the All Media and Products Study (AMPS) (Jan 2015–Dec 2015). The data were then analyzed using IBM SPSS Version 25.
Two sets of comparisons were made to assess campaign impact. First, pre-campaign data were compared with post-campaign data to detect changes over time. These comparisons between proportions were made using chi-square tests. For continuous variables, a t-test was used for the comparison. Second, in order to identify campaign-attributable impacts, the respondents were categorized according to whether they were “campaign aware” or “campaign unaware.” All respondents who recalled either of the ads from TV, which was the predominant campaign channel, were categorized into the “campaign aware” group, while all others were categorized as campaign unaware. Hence, those that recalled the campaign only via one of the complimentary newspaper advertisements were categorized as “campaign unaware.” The comparisons between the “campaign aware” group and the “campaign unaware” group were first made using chi-square test for categorical variable or using a t-test for continuous variables. Then, campaign awareness was regressed onto dichotomized measures of knowledge, attitudes, government policy support, and behavioral items. Covariates for the logistic regression analysis included age, gender, socioeconomic status, and frequency of watching television. The significance level for all tests was set to p < 0.05.
4. Discussion
The study findings demonstrated that the South African “Are You Drinking Yourself Sick?” campaign performed as intended. More than half the population surveyed recalled the campaign, and it was well received. Campaign-aware respondents accurately recalled the campaign’s message, they found it to be believable and relevant, and it increased their concern about the harms of sugary drinks. Over half of campaign-aware respondents discussed its key message with others, and 81% of campaign-aware respondents said that the campaign made them more supportive of government action to reduce sugary drink consumption.
There were significant improvements in knowledge, attitudes, and behaviors between the pre- and post-campaign periods. There was a significant increase in recognition of the problem of childhood obesity, increased knowledge of the harms of sugary drinks on the health of adults and children, and increased knowledge that sugary drinks contribute to the obesity problem in South Africa.
While the changes from the pre- to the post-campaign period may arguably have been the result of concurrent activities beyond the campaign, the comparison of campaign-aware and campaign-unaware respondents in the post-campaign data suggests the independent impact of the campaign on these outcomes. Even after potential confounders were controlled for, the data showed that campaign-aware respondents were significantly more likely than unaware respondents to demonstrate increased knowledge about the serious health risks of overweight/obesity and to express increased knowledge and concern about the harms of sugary drinks. In this regard, campaign awareness also played an important role in reversing the mistaken belief that exercise can protect against the harm of sugary drinks. Campaign awareness in the post-campaign period was associated with a decreased tendency for people to believe that exercise would have this protective effect.
Most importantly, this study found significant increases in support for government action, the primary objective of the campaign. These increases in support were observed from the pre- to the post-campaign period: for instance, there was a significant increase in South Africans’ support for government actions that discourage the consumption of sugary drinks and junk foods and in support for a tax on sugary drinks if the money collected was invested in public programs.
Participants were also asked more specifically about the proposal by the Department of Finance. While the expression of support to this more pointed question may be expected to be lower since conceivably many participants would not believe themselves to be fully informed of the proposal, the majority expressed support for this proposal and the proportion who expressed support grew significantly from the pre- to the post-campaign period. Furthermore, campaign awareness was found to have an independent association with support for government action. Within the post-campaign period, even after controlling for potential confounders, campaign-aware respondents were significantly more likely than unaware respondents to express increased support for strong government action, and for the sugary drinks tax, including the Department of Finance proposal.
Finally, while there were significant improvements in behavior between the pre- and the post-campaign periods, there was no independent association between campaign awareness and behavioral changes in the post-campaign data alone, suggesting that the behavioral changes may have been the result of the confluence of other activities surrounding the sugary drinks tax proposal and the increased interpersonal communication that may have been generated by the campaign and these other activities. In fact, increased interpersonal communication has been established as an important outcome and co-benefit of media campaigns.
That said, campaign awareness was independently associated with one item—increased water consumption—which could potentially be explained by the presence of an unrelated but extensive social media campaign promoting water at around the same time. These findings are consistent with the agenda-setting objective of the campaign, and they suggest the important benefits that might accrue from the increased interpersonal communication and narrative shifts that occur during media campaigns.
This study extends the literature on the important role of media campaigns in addressing behavioral risk factors [
30,
31,
32,
33,
34]. It replicates findings from high-income countries, and as in other public health applications, it shows that media campaigns for obesity prevention can play a crucial role in improving awareness, and changing knowledge, attitudes, and social conversations around obesity and its behavioral risk factors [
30,
31,
32,
33,
34]. Additionally, this study extends the literature by suggesting the important agenda-setting function that can be served by media campaigns for obesity prevention. Fiscal policies, such as the sugary drinks tax, have typically met with significant opposition and expensive public relations campaigns from the food and beverage industry that have sought to turn public opinion against these public health fiscal measures [
23,
24,
25]. This study demonstrates that public health practitioners can successfully use evidence-based media campaigns that present the public health case—in this case, the harms of sugary drinks and the need to reduce its consumption in South Africa—to generate social conversations, build public engagement, and thereby generate support for such policies. Indeed, as described above, the independent association between campaign awareness and increased support for government action highlights the important need for policy proposals, such as sugary drinks taxes, to be paired with complementary, evidence-based media campaigns that highlight the public health case for such measures [
30,
44,
45,
46,
47].
There were a few limitations to this study that are important to consider. First, despite the application of similar methods, the study sample in the post-campaign period varied in demographic profile compared to the sample in the pre-campaign survey. These variations were to an extent controlled for in the regression analysis that examined the association between campaign awareness and study outcomes. More importantly, the nature of the variation between the samples would have made the campaign impact harder—not easier—to detect. In fact, a common critique of media campaigns has been that their impact tends to be greater among higher-educated groups with greater socioeconomic status. Thus, this study’s evidence of improvement, and the association between campaign awareness and study outcomes, is particularly noteworthy. Second, the definition of “campaign awareness” in this study was restricted to those who recalled the TV ad alone. This was done to reflect the main channel of the communication campaign, but conceivably respondents who recalled the newspaper ads—and those who were exposed to the campaign but did not recall it in response to the survey question—may have been included in the campaign-unaware group, rendering this group a mix of those who were truly unaware of the campaign and those with some recall of the campaign. That said, it is a frequent limitation of campaign evaluations that rely on survey recall questions that those with campaign exposure but no recall cannot be distinguished from those with no exposure to the campaign. From the perspective of the assessment of campaign impact, however, this limitation would make it harder to detect campaign impact. Hence, the inherent bias is toward weakening the detection of campaign impact and not its converse, making the observation of campaign impacts particularly noteworthy. Finally, this campaign evaluation, as others, may be prone to the “selective attention” bias—the tendency of greater recall among those sympathetic toward or prone toward the campaign’s cause. However, the consistency in findings between the pre- and post-campaign comparisons and the regression analysis on most measures, and the fact that the post-campaign sample had more unhealthy dietary habits and poorer self-reported health status, and were thus less—not more—inclined toward the campaign, lends support for the effectiveness of the campaign.