Introduction

The prevalence of overweight and obesity in Australia, after rising steadily since 1990 [1] remains at a worryingly high level [2] placing a large proportion of the Australian adult population at risk of chronic disease and ill health. In 2011–2012, 70.3% of Australian men and 56.2% of women were classified as overweight or obese.[3]

Modifiable lifestyle-related risk factors for obesity such as PA and diet are the target of many interventions conducted at the individual, institutional and community level.[4–6] Social marketing campaigns incorporating mass media have also targeted lifestyle behaviours in order to tackle the obesity problem and its health consequences [7]. Evidence suggests that mass media campaigns can have a positive, if modest, effect on health knowledge, beliefs, intentions and behaviours in health,[8] although studies evaluating mass media obesity prevention campaigns appear infrequently in the scientific literature.[9, 10]

Measure-Up, an obesity prevention campaign run in Australia between 2008 and 2010, focused on waist circumference as an indicator of an unhealthy lifestyle and chronic disease risk (www.measureup.gov.au). Formative qualitative research indicated broad appeal of information about a “healthy waist circumference” as a compelling, credible and easy to understand goal and the salience of the consequences for the Measure-Up characters regarding their progressive weight gain over time.[11] An evaluation of Measure-Up in one Australian state (New South Wales (NSW)) demonstrated it communicated new information about waist circumference and risk of chronic disease but population-level lifestyle-related behaviours were little changed following the campaign.[12]

“Campaign awareness” is the first step in the change process induced by social marketing and mass media campaigns and can therefore act as a proximal indicator of impact.[7, 10] Further, whilst attributing change in health behaviours to a single influence is problematic given their multi-determined nature, on average, if a campaign is effective, those aware of a campaign should report more short-term change in the desired direction compared with those who are not (the need to undertake change notwithstanding). Therefore, incorporating campaign awareness into impact evaluation analyses can indicate whether subsequent population-level change (or lack of it) could be due to campaign penetration (low or high reach) or the campaign content itself. That is, if behavioural outcomes remain unchanged post-campaign, in the context of low campaign awareness this could indicate poor targeting; in the context of high campaign awareness, absence of behaviour change could indicate ineffective content.

Campaign awareness in obesity campaigns has been estimated at around 60% (at first wave for continuing campaigns).[9, 13–15] Wammes et al (2007) showed that unprompted recall increased over the three years of their evolving weight gain prevention campaign (from baseline <40% to 72% at wave 6). This was less than Measure-Up which achieved 89% prompted recall [12] for wave 2. However, different campaigns have reported differing recall measures. For example, Van der Feen de Lille et al (“Fat Watch” campaign, 1998),[16] Van Wechem (“Fat Watch”, 1998)[13] and Croker et al, 2012 (“Change4Life” campaign)[17] reported recall as “recognition of the campaign name or tagline”, whereas Wammes et al (2007)[14] and Verheijden et al (2012)[18] for the “Don’t Get Fat” campaign, Morley et al (2009) (“Piece of String”),[9] King et al (2013) (“Measure-Up”)[12] and Wardle et al (“Fighting Fat Fighting Fit”, 2001)[15] reported both “unprompted” and “prompted” recall of campaign content. James et al (2011)[19] reported outcomes for those “aware of the Draw the Line campaign” (p e24) but did not describe prevalence of campaign awareness or how awareness was determined (prompted or unprompted).

Three obesity-prevention-mass-media campaigns have analysed the knowledge, perceptions and/or behaviours by campaign recall by combining those who recognised and those who recalled the campaigns and comparing them with those who were not exposed.[9, 14, 18] Wardle et al (2001)[15] presented demographic and weight data separately for those who recognised the campaign and those who recalled four campaign messages but did not formally compare them. Thus studies reporting the impact of mass media obesity campaigns vary in their measures of campaign recall, making cross-campaign comparison problematic.

To date there have been no comparisons of those who recalled an obesity campaign unprompted, compared with those recalling after prompting and those with no campaign recall. Using data from the Measure-Up campaign, the current analysis examines whether prompted and unprompted recallers are in fact distinct groups demographically and in terms of subsequent campaign effects.[20, 21] Understanding the associations between campaign recall and subsequent outcomes enhances evaluation of obesity campaigns by signalling whether these two types of recall identify different levels of message uptake.[22, 23] Further, we tested whether risk profile differs between levels of obesity campaign recall, addressing whether those reached by the campaign were those most in need.

Specifically, the following analysis investigates: 1) the demographic profile of the groups with different levels of recall of the Measure-Up campaign; 2) whether there is a dose-response effect of campaign recall on knowledge, attitude and behaviour uptake, 3) whether those who did not report recalling the campaign had a higher risk profile than those who recall or recognise the campaign.

Method

Results

The response rate, as a proportion of the 7769 eligible households able to be contacted (those completed plus those who refused), was 36% (n = 2812).[39] No information was available on the eligibility or demographic make-up of those who refused or household who could not be contacted. The demographic characteristics of the post-campaign sample are shown in Table 1. Half were male, over one-third reported a university education, two-thirds were in paid employment and half of the sample self-reported as overweight or obese (53%). Consistent with the quota sampling approach outlined above, the sample compared well to population distributions for sex (within 0.1%) and age (within 1.2%) [38].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Unweighted demographic characteristics and (weighted) distribution of BMI in post-campaign sample.

https://doi.org/10.1371/journal.pone.0121387.t001

Without prompting, the Measure-Up campaign was recalled by approximately 41% of respondents (n = 1154) (Unprompted Recall), recalled after prompting by a further 46% (n = 1284) (Prompted Recall) while 13% (n = 374) reported neither recalling nor recognising the campaign (No Recall). Table 2 shows the demographic profile for each of the campaign recall groups.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Demographic profile for each level of Measure-Up campaign awareness (n = 2812).

https://doi.org/10.1371/journal.pone.0121387.t002

There were overall significant differences among the campaign recall groups for all demographic characteristics with the exception of employment status and income (Table 2). Contrasts showed that specifically the Unprompted Recall group were more likely to be university educated and female than the Prompted Recall group but not the No Recall group, and were less likely to be located in an metropolitan area than the No Recall group although not the Prompted Recallers. Unprompted Recallers were more likely than the other two groups to be living with their children, speak English at home, to be in the primary target group (25–45 year old group with children), but less likely to belong to the secondary target group (45–65 year-old group). Employment and income did not vary significantly by recall.

Table 3 shows the knowledge, past behaviour and future intentions regarding diet and PA patterns across the three campaign recall groups, with a test for trend across level of recall. The prevalence ratio (PR) shows the percentage increase in prevalence for moving from the No Recall group to the Prompted Recall group to the Unprompted Recall Group. In general the Unprompted Recallers had better knowledge than both the No Recall group and those that only recognised the campaign after prompting and demonstrated a dose-response effect of between 5% and 73%. For example, the prevalence of correct responses for recommended daily intake of vegetables was on average 51% higher for each increment in campaign recall.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Measure-Up campaign-related knowledge, behaviours, and intentions for each campaign recall group, with tests for trend.

https://doi.org/10.1371/journal.pone.0121387.t003

With respect to behaviour change, proportionately fewer respondents in the No Recall and Prompted Recall groups reported trying to improve their fast food intake or PA in the last six months compared than Unprompted Recallers. The No Recall group was also significantly less likely to report attempting to improve their vegetable intake in the last six months compared to the Unprompted Recallers, but Prompted Recallers and Unprompted Recallers were comparable across the other behaviours (Table 4). Supplementary analyses comparing the two recall groups only on behaviours respondents reported changing “as a result of seeing this campaign” (data not shown) confirm this with no significant differences for trying to increase fruit or vegetable consumption, and PA and one small but significant difference for waist measuring (Unprompted Recall: 18.5% vs Prompted Recall: 13.9%, p = .013). Intentions in the next six months showed a pattern where Unprompted Recallers were more likely to intend to change their behaviours towards a healthier pattern (e.g., higher vegetable intake) than No Recall but not Prompted Recall, with the exception that a higher proportion of the Prompted Recallers than Unprompted Recallers intended to increase their fruit intake (33% vs 28% respectively, p = .004). There were significant trend effects for trying to increase waist measuring, decreasing fast food, increasing vegetables and PA in the next six months with between 12–14 percent more respondents reporting intending to improve per increment in recall of the Measure-Up campaign. However, the lack of significant differences for the contrasts and small effect sizes for Unprompted versus Prompted Recallers suggest that the significant results for trend were likely due to the No Recall group’s result rather than a true dose-response effect.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Risk profile and risk perceptions for each level of Measure-Up campaign awareness.

https://doi.org/10.1371/journal.pone.0121387.t004

To examine whether those in the No Recall group are, and/or perceive themselves to be, at greater or lesser risk of obesity-related chronic disease than those who recalled (prompted or unprompted) the Measure-Up campaign, the risk profile and perceptions of risk for the three campaign recall groups were analysed (Table 4). The No Recall group had a significantly lower prevalence of being overweight/obese (46%) compared with Prompted Recallers (55%) and Unprompted Recallers (54%). A lower proportion of the No Recall group (37%) were eating fast food more than once a week compared with the other two categories (both 44%). Significantly more Unprompted Recallers than No Recallers reported being concerned that they would develop a chronic disease and perceived they were overweight (43% vs. 35%, respectively), despite comparability on perceived chance of developing a chronic disease (Table 4). The No Recall group reported consuming the recommended amount of vegetables than Unprompted Recallers, but was not different behaviourally than the other two groups on fruit consumption or meeting PA guidelines.

Discussion

Confirming earlier research, the Measure-Up campaign reached approximately half of the Australian adult population with campaign relevant messages with this innovative approach to frame obesity-related chronic disease risk.[12] Both prompted and unprompted campaign recall were very high (around 87% combined). Analysis by level of recall showed different associations with demographic characteristics, risk profile and campaign-related knowledge, behaviour and intentions by degree of campaign recall. Further, the risk profile for chronic disease of those respondents unaware of the campaign was no poorer, and in some respects more favourable, than those aware of the campaign. From a methodological perspective, the analyses suggest combining prompted and unprompted recallers may mask informative differences between these groups irrespective of the sources of those differences, be they pre-existing receptivity to health messages, campaign exposure and/or message absorption.

Demographic heterogeneity may be found even within those designated as ‘aware’ of the campaign. The Unprompted Recall group had the highest proportion of the primary target group (adults aged 25–49 years with children), but the lowest proportion of the secondary target group (45–65years). That the Measure-Up campaign reached and was recalled by its target audience confirms the value of campaign investment resulting in high media message exposure.[8, 41] Recall rates were similar by gender, unusual for health campaigns as women are usually more likely to attend to health messages.[42, 43] In the case of Measure-Up, this finding may stem from greater salience for men of the male-modelled Measure-up advertisement. (http://www.measureup.gov.au/internet/abhi/publishing.nsf/Content/television). Further analyses found that men were more likely to recall only after prompting, whereas women were more likely to recall unprompted, once again suggesting closer attention to health campaign messages by women. Unprompted Recallers were more likely to have had a university education than Prompted Recallers but were comparable to those not recalling the campaign. Proportionately fewer people from non-English speaking backgrounds were among the Unprompted Recallers, perhaps reflecting the role of language proficiency in health literacy[44], lower identification with the (Caucasian) message role-models, and/or lower exposure to mainstream Australian television.[34, 45]

A key target group for health campaigns are people in non-metropolitan areas as they can less easily access health services.[46] Measure-Up reached a disproportionate number of residents from non-metropolitan locations (equal for prompted and unprompted recall) underscoring the important role of television advertising in social marketing in reaching rural populations. Indeed the free telephone lifestyle coaching service (“Get Healthy Service” (GHS)) advertised in tandem with Measure-Up in and GHS specific advertising for NSW showed the reach to non-metropolitan communities was reflected in uptake of obesity prevention support, a central step towards behaviour change.[47] Thus the current analysis revealed subtle differences in the exposure and/or attendance by different demographic subgroups that would not be apparent in a simple aware/unaware dichotomy.

The significance of differential recall among demographic subpopulations is found in the dose-response relationships for knowledge regarding the key messages from Measure-Up. Better campaign recall was consistently associated with correct responses for fruit, vegetable and PA guidelines and understanding the link between lifestyle and chronic disease. However, the corresponding self-reported behaviours followed a slightly different pattern, with Unprompted Recallers more likely to try to change their behaviours towards healthier practices compared to No Recallers, confirming findings from evaluations of other weight related campaigns.[9, 14, 18) For example, in the Dutch national “Don’t Get Fat” campaign’s second phase, low SES respondents reporting exposure to the campaign showed an increase in their attention to food choices compared to those reporting no campaign exposure; in the longer term, BMI increases were less likely among respondents from non-Dutch backgrounds with some exposure to the campaign relative those reporting no message exposure.[18]

Despite greater recall, with only a few exceptions, Unprompted Recallers and Prompted Recallers were comparable on past attempts and future intentions to change their health behaviours, a result further confirmed when behaviour change was narrowed specifically to that which the respondents said was “as a result of seeing this campaign”. Hence, while reporting exposure to the advertisement was correlated with reported or intended behaviour change, despite their greater knowledge, the Unprompted Recall group did not report greater efforts to change than the Prompted Recallers.

From a health promotion perspective, the lower fruit consumption reported by the Prompted Recall group was matched by an intention to increase fruit intake in the next six months. Of greater concern was the small proportion across all groups that met vegetable consumption guidelines, yet less than half (≈40%) reported intending to increase their vegetable intake. Adequate vegetable intake remains one of the more difficult lifestyle factors to change[48] despite knowledge of recommended intake and positive intentions. Thus the priming steps generated by mass media campaigns such as Measure-Up require further translation to behaviour change with community and environmental supportive programs and policies, or access to community-wide programs (e.g. GHS);[47] unfortunately these data did not include whether the respondent contacted a support service such as GHS.

The No Recall group did not appear to be at any greater risk of chronic disease than their campaign-aware counterparts. Indeed, for BMI, vegetable (vs Unprompted Recall only) and fast food consumption they fare significantly better than the two recall groups and the population in general (at least for BMI).[2] Thus, it would appear that those unaware of the Measure-Up campaign were not the intended target group of the campaign, and may have not attended to the campaign (if they had seen it) because of a perceived and actual lower than average prevalence of overweight/obesity (although still 46%). Conversely, the primary target group as identified by pre-campaign formative research, who constitute those in the Australian population most at risk of lifestyle related chronic disease, was better represented among those who could recalled the campaign than those who did not; 87% of this group recalled the campaign (unprompted + prompted, data not shown).

Conclusion

Mass media campaigns are an important first step in communicating to the community about complex public health problems such as obesity. Measure Up was the initial step in a multi-year preventive approach to obesity in Australia. In this analysis, better recall of the Measure-Up campaign was associated with better knowledge, about campaign messages; behaviour change and intentions to change were stronger for the two recall groups compared with the No Recall group but not different between them. Moreover, the demographic profiles of the two recall groups suggest that campaign messages have differential penetration with exposure. From an analytical perspective, differentiating between prompted and unprompted recallers in impact analysis and profiling campaign reach may reveal useful insights into message uptake among different subgroups, although behaviour change did not necessarily flow from greater message uptake. Further, examining whether those unexposed or at least not recalling a campaign are at any greater risk than their exposed counterparts is a key element in evaluating the reach. While numbers may prevent some campaign evaluations from partitioning those recalling a campaign into prompted and unprompted, investigating demographic profiles and impacts in this fashion may afford greater insights into the relationships between exposure, recall, uptake of campaign messages and behavioural change.

Acknowledgments

The authors would like to acknowledge and thank the Australian National Preventive Health Agency for their contribution to this evaluation, the Social Research Centre and GfK Blue Moon for conducting the tracking survey for the mass media campaign evaluation; and Amy Zhong, Prevention Research Collaboration, University of Sydney and Liz King, Cancer Australia for their assistance with coding the free text data variables.