Population-level effects of mass media campaigns on adult smoking

While there are older reviews38–42 of MMC effects on smoking behaviour, recent reviews on youth smoking effects43 44 and broader reviews of MMCs across health behaviours,1 4 we focused on the two most recent in-depth reviews of campaign effects on adult smoking.7 45 A Cochrane review examined a total of 11 MMCs, investigated in 58 research studies that employed controlled trial designs.45 The authors concluded that comprehensive tobacco control programmes that include MMCs can positively change smoking behaviour in adults, but noted studies were heterogeneous in their methodological design and quality.45 They further concluded that intensity and duration of MMCs may influence effectiveness, but length of follow-up and concurrent secular trends made this difficult for these reviewers to quantify.

A NCI review in 2008 was more comprehensive in its inclusion criteria, embracing controlled field experiments and population-based studies as part of comprehensive tobacco control programmes.7 This review therefore included more evaluations of the large-scale campaigns of more recent years. Although these interventions are ‘messy’ and difficult to evaluate, they represent a better test of the intended population-based scale of MMCs, with greater levels of exposure than campaigns of earlier years.2 Of 10 controlled field experiments promoting smoking cessation, 7 showed some evidence for an effect on reducing smoking behaviour. Of nine field experiments, eight showed positive benefits of campaigns on quitting. Of 11 population-based government-funded MMCs that occurred as part of state or national tobacco control programmes, all demonstrated positive effects on smoking prevalence and/or per capita tobacco consumption.7

Several more recent empirical studies of specific MMCs have shown mixed support for MMC effects on knowledge, beliefs or intentions (table 1). In a series of large cross-sectional surveys, confirmed awareness of at least one ‘truth’ television advert (aired at low to moderate intensity over 4 years) was associated with five of nine anti-smoking beliefs and stronger desire to quit completely.15 A longitudinal evaluation of a moderate intensity 5-week campaign emphasising the benefits of cessation found modest improvements in attitudes towards the adverse impacts of smoking among young adults, compared with those in a non-campaign area.11 A MMC aimed at Maori smokers and recent quitters in New Zealand found 54% reported the campaign made them more likely to quit, but no change in quit intentions was observed before versus after the campaign.12 A 21-day campaign in Brazil communicating the harm of secondhand smoke produced improvements in some, but not all, campaign-targeted beliefs.9 Although some positive effects were found in each of these studies, they only measured change in attitudes and intentions, rather than actual quitting associated with exposure to MMCs.

Recent evaluations that have examined quit attempts or sustained quitting, mostly provide further support for MMC effects (table 1). One pilot study and a large national evaluation both employed a cohort design to examine the effects of confirmed recall of the low to moderate intensity ‘EX’ campaign among smokers. The small pilot study found a trend for effects on quit attempts, but had limited power to detect effects on this behaviour. The larger national study found a significant increase in quit attempts and a trend towards higher quit rates at follow-up.16 17 A three-wave cohort study conducted before, during and after a 5-month national campaign encouraging and supporting cessation in The Netherlands found discussion about campaign adverts prompted discussions about quitting, which subsequently predicted quit attempts.19 A cohort study found confirmed recall of two specific advertising campaigns (keep trying to quit; dangers of secondhand smoke) aired at low to medium intensity was not associated with quit attempts or 1-year cessation, although it had limited power to detect effects on quitting behaviour.13 A cross-sectional study reported beneficial effects of a localised MMC for Arabic-speaking smokers on reducing Arabic smoking prevalence, compared with non-Arabic smoking prevalence.14

Three more comprehensive recent studies have attempted to overcome the limitations of lack of a control group and lack of power to detect quitting outcomes. These studies use natural experimental designs that exploit geographic and temporal variation in extent of cumulative population exposure to televised national and/or state MMCs, as measured by advertising rating points (gross ratings points (GRPs) and targeted ratings points (TRPs); GRPs are a standard advertising industry measure of campaign reach × frequency; for example, 1000 GRPs per quarter equates to, on average, 100% of those within a region exposed to 10 adverts, or 50% exposed to 20 adverts and so on). A time series analysis of monthly smoking prevalence over an 11-year period found greater population exposure to televised MMCs (increase of 390 GRPs 2 months earlier) was associated with acceleration in the decline in adult smoking prevalence (0.3% percentage point drop), after adjusting for variation in tobacco prices, smoke-free restaurant laws, tobacco marketing restrictions and availability of smoking cessation products.20 In a population cohort study, greater population exposure to televised MMCs (average 853 GRPs per month over 2 years) was associated with a higher likelihood of quitting at a 2-year follow-up (odds increased 11% for every 1000 GRPs).10 Another cohort study conducted over 6 years, found every 1000 GRPs in the 3-month period prior to follow-up, increased the odds of making a quit attempt by 11%.21 These studies of the effects of multiple campaigns over time indicate campaigns can generally expect small effect sizes (requiring large samples to detect), but because they involve mass numbers of individuals within a population, the degree of change is of high practical significance, helping to drive down population smoking prevalence by increasing successful quitting rates.

Since the last reviews of 2008 concluded that MMCs have beneficial effects, five of eight additional studies examining effects of MMCs on behaviour (quit attempts, sustained quitting or smoking prevalence) found beneficial effects, one found a mediated effect, one found a trend towards an effect and one did not find any effects. Of the additional four evaluations that only measured campaign-targeted knowledge, beliefs and/or quitting intentions, some positive findings were also observed for each campaign. Overall, these studies further strengthen the evidence that MMCs can improve attitudes and intentions, promote quitting and reduce adult smoking prevalence. Some mixed findings indicate not all campaigns are equally successful, and factors including the reach, intensity, duration and type of messages used are likely to determine overall impact.

Campaign decay, intensity and duration

An important aspect of determining optimal campaign investment is campaign decay, or the extent to which effects are detectable after the campaign broadcast ends.46 Although some consumer product advertising may be recalled long after broadcast ends, most advertising primarily has effects on product sales during broadcast and for a short time following.46 These effects have also been observed in tobacco control MMCs. A recent cohort study21 found quit attempts were associated with MMCs in the most recent 3 months, but not with exposure in earlier months. A time-series analysis showed the beneficial effect of MMCs on smoking prevalence lasted only up to 2 months after exposure.20 One reason that withdrawal of MMCs leads to decay of effects is because tobacco-promoting influences, such as tobacco marketing and tobacco addiction, gradually regain precedence. Sustained smoking behaviour change requires frequent longer-term campaign exposure.

Few studies have considered the intensity of campaign investment that can most efficiently achieve population changes. Hyland and colleagues47 found an average of 700 GRPs per quarter was only weakly associated with quitting. A more recent cohort study of adult smokers found a mean of 2560 GRPs per quarter was significantly associated with quitting.10 A time-series analysis found an average of 1200 GRPs each quarter was significantly associated with reduced smoking prevalence.20 Taken together, these findings suggest that, at least in high-income countries, an average of 1200 GRPs per quarter for a total of 4800 GRPs per year are needed to reduce adult smoking prevalence, although 2560 GRPs per quarter (total of 10 240 GRPs per year) could be expected to reap proportionally larger returns. Greater advertising exposure may be required to influence adult than youth smoking, due perhaps to the fact that most adult smokers are hampered by addiction. For comparison, a threshold for detecting beneficial effects for reducing smoking uptake among youth may exist around 300 teenage-targeted GRPs per quarter,48 with effects increasing linearly48–50 until potentially beginning to diminish above 1250 GRPs per quarter.49

Effects of MMCs on demographic subgroups

In contrast to producing general audience adverts that convey messages of relevance to most adult smokers, targeting tobacco control adverts to specific audience groups (eg, age, gender, education, ethnicity) may increase message relevance and persuasion. However, this strategy may also increase costs, as multiple adverts are required. Given finite resources, targeting messages may result in a lower proportion of funds available to broadcast these different adverts, resulting in lower rates of exposure. The extent of targeting and segmentation therefore needs to be weighed carefully against the importance of maximising campaign exposure. Research among adolescents reported in the NCI review7 found few differences between different demographic groups' responses to MMCs, that advert characteristics were more important than demographic characteristics, concluding that adverts that perform well do so among many population groups.

To examine these issues among adult smokers, we compared demographic subgroup differences in response to MMCs reported in previous reviews and more recent studies (online table 1). In the review by Bala and colleagues,45 two wide-reaching campaigns examined effects separately for different socioeconomic status (SES) groups (education and ethnicity) and campaign effectiveness did not differ by education or ethnicity. Another review of MMC effectiveness among socioeconomically disadvantaged populations,51 suggested that differences in the effectiveness of MMCs between SES groups may be due to differences in meaningful exposure, or motivational response, or opportunity to sustain cessation in the long term. Of the nine reviewed general population campaigns that were less successful among lower SES smokers, five suffered from low levels of exposure and promotion.52–57 Of the other four higher reach campaigns, two were less effective in generating quitline calls in lower SES,58 59 while the other two campaigns showed equal effects across SES in motivating quitline calls, but SES differences occurred in translating these attempts into success.60 61 In contrast, nine other reviewed general population campaigns showed equal62–67 or greater effects on motivational response and/or long-term cessation in lower SES groups, and all nine of these achieved sufficient exposure.68–70 This review51 also identified 13 MMCs that specifically targeted low SES populations. Underscoring the importance of adequate exposure and campaign reach, the five targeted campaigns that were not successful, were hampered by limited reach,71–73 or very low levels of recall.74 75 The other eight campaigns showed mixed findings.76–83

Three more recent population-level studies have reported effects separately by SES. An examination of high intensity televised MMCs found no significant interaction between total potential exposure to MMCs and quitting behaviour 2 years later across SES groups.10 A recent evaluation of a non-televised campaign,14 found a significant reduction in smoking prevalence in the highest SES group, but not in moderate or low SES groups. A cohort study of the ‘EX’ campaign found campaign awareness was associated with increased quit attempts among non-Hispanic black smokers, but not among Hispanics (or non-Hispanic white smokers), and among smokers with less than high school education, but also showed a positive trend (p=0.08) among those with most education.18 This campaign was aired at low intensity overall, but placement was targeted to media programmes and channels consumed by lower income and blue collar smokers of diverse race/ethnicity. The clear effects on lower SES smokers for the widely broadcast television-led campaigns10 45 51 versus the negative and mixed effects of low-reach general population campaigns and low-SES targeted campaigns,14 18 51 indicates that general population campaigns of at least moderate intensity and duration are effective for motivating quitting in lower SES groups. However, this strategy is likely to produce equivalent effects across SES groups, rather than greater effects in lower SES groups.

Examining differences by age and gender, Bala and colleagues45 found three studies that showed long-term MMC effects for men and one for women, while three found MMC effects among younger smokers and another three found MMC effects among older smokers. A recent non-televised campaign,14 found a significant reduction in prevalence in males but not females and among those aged 40 years or older, but not in those aged 18–39 years. Overall, given only one additional study addressing demographic subgroups, we defer to the conclusion of Bala and colleagues45 that campaign effectiveness does not consistently differ by gender and age.

Effectiveness of different channels of delivery of mass media campaigns

Few studies have examined the relative effectiveness of different advertising channels. The NCI review7 of 47 MMCs found 98% used television, 94% radio, 89% print and 87% billboards. Only two reviewed studies examined comparative effects of different media types, finding television adverts were recalled by twice as many respondents as radio adverts. This review highlighted another US national adult population survey which found television provides the greatest exposure among smokers, and that smokers are more likely to be heavier users of television and radio and less likely than non-smokers to be magazine or newspaper readers.84

Since the NCI review, two studies have examined the relative effectiveness of different MMC channels on quitline calls. Farrelly and colleagues85 found for each 10% increase in expenditure on television, radio and newspaper advertising, calls increased by 1.51%, 0.037% and 0.022% respectively, with the latter only a marginally significant association. Mosbaek and colleagues30 examined a range of different adverts aired on television and on radio and found the 10 most cost effective adverts were aired on television, while the most cost effective radio advert was ranked 11th overall. It is difficult to determine whether the reduced effectiveness of non-televised messages is due to the channel, to lower population reach, or to differences in message effectiveness.

One recent study found a MMC message broadcast on radio generated similar levels of concern about smoking and motivation to quit as a similar message shown on television.86 Although this study indicates radio messages can be effective, it was broadcast concurrently with a televised version of the campaign, and so part of its effectiveness may be due to smokers bringing to mind the images associated with the televised advert. Future research should examine the effects of a standalone radio campaign. Despite radio's lower costs its reduced population reach means that it is unlikely to be a good substitute for television in influencing population-wide smoking, and could be considered a reinforcing adjunct.

Effectiveness of different types of mass media messages

MMC messages differ in informational content (theme), purpose (why-to-quit vs how-to-quit), method, emotional tone and stylistic features.7 13 33 87 88 The NCI review identified four studies that explicitly addressed the effectiveness of different messages among adult smokers.7 One89 examined results from a collection of focus tests, concluding that anti-industry and secondhand smoke (SHS) messages were most effective; however, this study was criticised for lacking a transparent criteria for the definition of effectiveness.7 89 The other three studies demonstrated that messages which elicited negative emotions by describing the serious health consequences of smoking scored higher on memorability90 and ratings of perceived effectiveness91 and were more likely to be recalled by recent quitters who believed that anti-smoking advertisements had contributed to their quit attempt.62 Our search identified 6 new population studies, 3 quitline studies and 11 forced-exposure studies that compared the effectiveness of different message characteristics, including comparisons of message themes (negative health effects, how-to-quit, anti-industry, social norms, information about SHS regulations), methods (graphic depictions, medical simulations, testimonials, narratives, information), emotional tone (high vs low emotion) and inclusion of specific stylistic features (outcome extremity, message sensation value, argument strength, smoking cues) (table 2). In this review negative health effects (NHE) messages are those that emphasise the serious health effects of smoking for the individual and/or their family or friends, and include those previously described as ‘health effects’, ‘why-to-quit’, ‘reasons to quit’, ‘family testimonial’, ‘smoker testimonial’ and ‘dangers of SHS’.

Four studies examined the impact of inclusion of specific stylistic features, including outcome extremity, message sensation value (MSV), argument strength and presence or absence of smoking cues on campaign outcomes.27–29 36 One forced-exposure study found better performance on a recognition task and on psychophysiological measures of central cognitive processing after exposure to more extreme loss-framed messages than after exposure to less extreme and gain-framed messages.29 One study identified an advantage for low MSV on accuracy and response time,28 while another forced-exposure study found the impact of MSV on self-efficacy and beliefs depended on the sensation seeking level of participants.36 This study also found there was no effect of argument strength on beliefs and intentions.36 One other study found smoking cues increased urges to smoke and caused larger decreases in heart rate (potentially indicating an orienting response), but only when argument strength was low.27

Overall, two population,10 13 two quitline studies26 30 and four forced exposure studies reported in six articles22 31–35 compared the effectiveness of NHE messages to other message themes (social norms; anti-industry; how-to-quit). The two quitline studies found some specific NHE messages were as, or more effective than some specific how-to-quit messages, but other NHE messages were less effective. All four of the forced-exposure studies22 31–35 found NHE messages were more effective than how-to-quit, anti-industry and/or social norms themes at generating either increased knowledge, positive beliefs, perceived effectiveness ratings or motivation to quit. No studies that examined effects on quit intentions found differences between message themes31 32 34 35 however, these four studies included a majority of non-smokers in their samples and so were underpowered to detect effects on smokers' quitting intentions. There was also evidence from one large population study that high emotion or testimonial adverts (majority of which were NHE messages) were more effective than comparison adverts at increasing quitting rates.10 The other population study found neither a NHE-SHS message nor a keep-trying-to-quit message increased quit attempts or sustained quitting, however this study had limited power to detect changes in quitting behaviour.13

Overall these findings provide good support for the use of NHE messages, comparatively less support for how-to-quit and anti-industry messages, and an indication that more work is required to determine the importance of stylistic features, such as MSV and the inclusion of smoking cues. As discussed in the NCI review, untangling the effective elements of NHE messages remains difficult because of the tendency for particular elements (for example, the NHE information, using a graphic and/or testimonial format, and high levels of negative emotions) to co-occur.7

A number of recent studies have attempted to disentangle the key effective elements of NHE adverts. Examining different types of NHE messages separately, one study found those messages that included neither graphic images nor negative emotion (typically how-to-quit adverts) and those with graphic images only were associated with increased quitline calls, while there was a trend for those including graphic images and negative emotion to be associated (p=0.089), but those with negative emotion only did not increase calls.26 However, unlike other research which has used smokers to rate the level of emotion in adverts10 25 91 this study had coders rate adverts in terms of the strength of emotion they expected would be elicited from smokers. In contrast, another study25 using an independent smoker rating of advert emotion level, found that NHE messages that generated high levels of emotion and/or had narrative elements drove greater quitline calls, while there was no effect of airing low emotion non-narrative NHE messages. Similarly, a large population cohort study found those adverts that did not generate high emotion or include personal stories were not associated with sustained quitting.10 Further research on the relative effectiveness of the level of emotion and narrativity of NHE messages is required, and should also explore whether the use of emotion and story-based messages might explain the more successful how-to-quit and anti-industry messages.

Effectiveness of different message types by demographic subgroups

A number of population and forced-exposure studies specifically explored message characteristics that might be most effective for lower SES smokers (online table 2). A population study found baseline recall of ‘keep-trying-to-quit’ adverts was associated with lower probability of making a quit attempt among those with high school or less education, but with a greater probability among higher educated smokers.13 In contrast, there was no interaction between education level and baseline recall of an NHE-SHS advert. Another study10 found exposure to high emotion and/or personal testimonial adverts was associated with quitting at 2-year follow-up for mid, and to a lesser extent for low SES smokers, but not for high SES smokers. A large forced-exposure study33 found differences in perceived effectiveness between NHE adverts and how-to-quit adverts were more pronounced among lower educated smokers compared to higher educated smokers. Greater increases in quitline calls from lower SES groups were also observed in one study25 when high emotion testimonial NHE messages were aired, and this advantage was evident but not as strong among higher SES groups.

Two studies have explored whether males and females and smokers of different age groups respond in similar ways to different message characteristics. A large forced-exposure study22 found few age and gender differences in responses to different adverts (online table 2). Another forced-exposure study found no gender effects on message-targeted knowledge or beliefs.32

The literature examining differences in the effectiveness of various message types among different SES groups indicates that widespread broadcast of NHE messages using testimonials or graphic depictions may work well across population groups and may contribute to reductions in disparities between high SES smokers and smokers with some degree of disadvantage. Research should examine whether these types of messages can reduce disparities between smokers with greater degrees of disadvantage. There is some evidence that some messages may increase SES disparities in smoking (how-to-quit; keep trying to quit),13 33 however more research is required. Examinations of the effectiveness of different message types across age and gender revealed no consistent patterns of effects.