Mass media to communicate public health messages in six health topic areas: a systematic review and other reviews of the evidence

Review of reviews

Reviews of reviews are becoming an established component in the repertoire of evidence-informed policy and practice. 12–14 They allow key findings from a range of studies to be easily accessed, while also identifying research gaps. We reviewed and synthesised evidence from English-language systematic reviews published between January 2000 and January 2016 on the effectiveness of mass media campaigns across six health topics that represent the main preventable risk factors for disease morbidity and mortality in developed countries:15 alcohol use, illicit substance use, diet, physical activity, sexual and reproductive health and tobacco use. We registered this review of reviews (review A) with PROSPERO (CRD42013004170)16 (see Chapters 2 and 5).

Reviews of primary studies

No systematic reviews addressing alcohol use or diet met our inclusion criteria for the review of reviews described in the previous section. As a result of this, and as anticipated in our protocol,16 we conducted two reviews of primary studies. The first (review B), a systematic review of English-language primary studies (published by July 2016), was conducted to assess the effectiveness of mass media public health campaigns to reduce alcohol consumption and related harms. Studies examining drink driving mass media interventions and college campus campaigns were excluded. We registered this review with PROSPERO (CRD42017054999)17 (see Chapter 3).

The second (review D) was a systematic review of English-language primary studies of mass media campaigns targeting the same six health topics, conducted in the UK and published between January 2011 and September 2016. The focus of the review was on evidence concerning the characteristics of UK mass media campaigns associated with effectiveness, rather than on the effectiveness of those campaigns per se (see Chapter 5).

Other reviews

We conducted a rapid review of reviews describing the cost-effectiveness of mass media campaigns (review C). We reviewed reviews and systematic reviews, published between January 2000 and January 2017, which assessed economic studies that evaluated both the costs and benefits of mass media campaigns for any of our six health topics of interest (see Chapter 4).

As described previously, no systematic reviews addressing diet met our inclusion criteria for the review of reviews. A scoping search for English-language primary studies (published by August 2016) was conducted for studies of mass media public health campaigns aiming to improve dietary behaviours. The modified search strategy (diet terms and mass media terms) was tested in one database (MEDLINE) and identified > 16,500 hits. A full review and synthesis was too great to conduct within the time and resources of the current project. Project resources were instead directed towards the review of recent UK primary studies (published between January 2011 and September 2016), referred to previously (review D). We focused on UK studies to complement the review of reviews (review A) and enhance the relevance for UK practitioners, policy-makers and commissioners (see Chapter 5).

The utility of logic models in systematic reviews

In a broad systematic review such as this, a range of different types of interventions in different contexts are compared and contrasted. Critical to this process is an understanding of how the different interventions are thought (or intended) to work; this provides a conceptual framework to structure the analysis. Based on the idea of programme theory from the evaluation literature, this framework is often described as a ‘logic model’, which is a diagrammatic representation of the key intervention inputs, the activities undertaken in the intervention, and the causal pathway that is triggered by the intervention, resulting in the desired (or not desired) outcome(s). 18 Thinking critically about the causal pathway is important in public health interventions, as there are often long chains of outcomes between the intervention and the ultimate health outcome. For example, in this review, a given mass media campaign might be designed to have a given message to raise awareness about the consequences of a given behaviour. It may adopt a given strategy or intervention theory in order to raise awareness, but merely raising awareness does not necessarily result in improved health. The raised awareness needs to result in a decision to change behaviour; the initial behaviour change, and, ultimately, sustained healthier behaviours, may lead to an improvement in population health.

Many systematic reviews develop a logic model a priori, as this can then drive many of the decisions that need to be made during the systematic review process. First, systematic reviewers need to make consistent decisions about which studies are in the scope of the review and which are not. The logic model can be used to develop inclusion/exclusion criteria in order to delineate the scope of the review. Once the studies for the review have been identified, the logic model can be used to determine what data need to be extracted about studies in a standardised way, in order to structure the comparative analysis. The logic model then helps to structure the analysis, enabling reviewers to identify commonalities and differences in interventions that may help to explain variance between their results. However, although the existence of an a priori logic model can be useful for these reasons, it should be considered provisional and subject to change once the studies have been examined. This is important because once reviewers have seen the range of studies in their review they may find that the logic model does not contain sufficient nuance to capture significant differences in intervention approach, content or in the contextual factors that might influence intervention implementation – or the long causal chain between intervention and health outcome. For this reason, this review contains two logic models: the first, which informed the early stages of the review, helped reviewers to determine what was relevant and irrelevant, and what data should be extracted; the second is based on the first, and also summarised the reviewers’ understanding of the research contained within the review.

Development of the initial logic model for mass media interventions

The initial logic model owes much of its overarching structure to the work of Chen. 19 We split the model into two major components: the action model (comprising the intervention inputs, activities and media outcomes) and the change model (made up of proximal, intermediate and distal outcomes). Although this may appear to be rather linear, and not cognisant of relevant theorising about complex interventions (e.g. feedback loops, phase changes and emergent outcomes; see Rogers20), we consider mass media interventions as operating in different ways to other public health interventions, and it is possible to conceptualise the intervention as a coherent entity that is implemented, and then the outcomes that result from it in a linear way (i.e. there may be feedback loops and other manifestations of complexity within the change model, but these can be understood as operating downstream of the mass media intervention, and not interacting with it).

We first developed our initial logic model for mass media interventions separately for each of the public health areas of the review before synthesising these into a common logic model. As well as demonstrating how mass media interventions may work, the resulting logic model was used to guide the evidence synthesis through helping to define inclusion and exclusion criteria, identifying moderators (and potentially subgroup analyses if meta-analyses had been possible at a later stage), and identifying mediating factors and guiding the search for evidence. 21 Our initial model represents a synthesis of logic models developed independently of mass media interventions of smoking cessation and mass media interventions of healthy eating/physical activity. In common with the development of logic models more broadly, both logic models were developed through working backwards across an outcome and action chain starting from the distal outcome.

Beginning with smoking cessation, we first located the small number of systematic reviews of mass media interventions for smoking that included a logic model, and used the model included in the review by Niederdeppe et al. 22 as a starting point. This included detail on the change part of a logic model in particular, but was enhanced with further details that helped to disaggregate some of the intermediate outcomes around behaviour change; this corresponded with other models of ‘stages of change’ in health promotion. The action part of the model was enhanced through examining logic models that were developed in other studies of mass media interventions of public health but that were not necessarily specific to smoking cessation (e.g. Huhman et al. 23), as well as significant components that were identified in reviews of mass media smoking interventions, but that were not conceptualised in a logic model (e.g. Durkin et al. 24). Finally, further stages of change of smoking cessation were identified through examining the logic models that were included in reviews of public health and policy interventions for smoking cessation, but that did not necessarily involve mass media. 25 A similar process was employed to develop the logic model for healthy eating/physical activity. To synthesise the models, common components were identified and the language harmonised; for example, both the physical activity and smoking cessation logic models included common stages of change around the attempts at adopting healthier behaviours and reduction in unhealthy behaviours as precursors to successful behaviour change, although these were originally expressed in language specific to each health topic. Even though the two health topics included here were chosen because they were conceptually relatively different and could affect very different populations (making them suitable candidates to pilot this approach), their synthesis was relatively straightforward as both involved synthesising logic models of mass media interventions to stimulate behavioural change for lifestyle behaviours. However, as we expected that some of the health topics that the review would consider may be more complex, we expected that our process of synthesising logic models and developing an overall logic model might result in topic-specific pathways being depicted within the final model; for example, mass media interventions for some health topics might also attempt to change behaviour through an intermediary party, and this might need to be depicted in the logic model. Thus, our initial logic model is presented here (Figure 1), and it was continually challenged and refined throughout the process of the review.

FIGURE 1.

The National Institute for Health Research mass media review: logic model.

Identification of reviews

We combined terms concerning mass media campaigns, such as media, ‘mass communication’, ‘social marketing’ and broadcast, with terms denoting systematic reviews and meta-analyses (see Appendix 1). We searched the Database of Promoting Health Effectiveness Reviews, the Cochrane Database of Systematic Reviews, the Database of Abstracts of Reviews of Effects, the Campbell Collaboration Library of Systematic Reviews, the Health Technology Assessment database hosted by the Centre for Reviews and Dissemination, EMBASE, PubMed, the Cumulative Index to Nursing and Allied Health Literature, MEDLINE and Web of Science between 10 December 2015 and 5 January 2016. We did not systematically search the grey literature, which was a departure from our protocol; however, systematic reviews published as reports, rather than in peer-reviewed journals, were still identified by the strategy described above. To check the quality of the searches, we searched the results to find systematic reviews already known to the team. The reference lists of any relevant reviews of reviews were also searched. Results were uploaded to an Evidence for Policy and Practice Information and Co-ordinating Centre (EPPI)-Reviewer 4 database and deduplicated (Figure 2). 26

FIGURE 2.

The PRISMA flow diagram of the identification and selection of reviews (review A).

Review selection

Records were screened against the inclusion criteria listed in Box 1. Reviews were screened on title and abstract by three reviewers. We carried out comparison coding as an inter-rater reliability test, and, when we agreed on the included and excluded reviews at a 90% rate, the reviewers continued individually. Full-text reviews were then retrieved, and individual expert teams assessed the papers in the different health topic categories to reach the final list of reviews. Two researchers from the wider team adjudicated if there was uncertainty about whether or not to include a review. See Appendix 2 for a list of reviews excluded by full-text assessment.

Data extraction

Data from reviews identified as meeting the inclusion criteria were extracted into a standardised data extraction form. Data extracted included review characteristics, participant characteristics, types of study design, types of synthesis and outcome data (particularly social cognitive and behavioural outcomes). For each topic, one reviewer extracted the data, and a sample (≥ 25%) was checked by a second reviewer to ensure the consistency of the extraction.

Quality appraisal and relevance assessment

We used the Risk Of Bias In Systematic reviews (ROBIS) tool to assess the risk of bias of included systematic reviews. 28 Included reviews were assessed by one researcher, and a second researcher checked all their assessments against the full-text review and ROBIS guidelines, with any disagreements discussed between the two researchers. We rated the relevance of the included reviews to our aims (high or low relevance), based on two dimensions: its relevance to an Organisation for Economic Co-operation and Development (OECD) context (included studies conducted in OECD countries versus all studies in the review that were conducted in non-OECD countries)29 and whether or not the review’s main focus was on mass media interventions. We also extracted information on the quality of the included studies in each review as assessed by the review authors: good, medium or low quality or not stated.

Data synthesis

Given the highly heterogeneous nature of the interventions and reviews, we did not attempt to conduct meta-analysis, and a narrative synthesis approach was pursued. 30 We initially tabulated all available data according to topic and tried to identify duplicate results. We then created tabular summaries of the full data, with information on potential bias within the included evidence base retained. We investigated patterns in the available results, making comparisons across topics, outcomes and population subgroups [based on the PROGRESS (Place of residence; Race/ethnicity/culture/language; Occupation; Gender/sex; Religion; Education; Socioeconomic status; Social capital) characteristics],31 with due attention paid to contradictory data. Analysis proceeded iteratively, with the whole team regularly meeting to discuss findings. To summarise the results for the outcomes of interest (behaviours, intentions, awareness/knowledge and attitudes), a symbol was applied to indicate how good the evidence was for a positive or negative effect. 32 This incorporated the risk of bias of the relevant reviews and reported effect sizes/directions. Inconsistency statistics were extracted from relevant meta-analyses.

To make conclusions based on the available evidence, we developed a systematic and transparent approach, building on principles of the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) approach. 33,34 In addition to risk of bias, we also assessed the domains of inconsistency, indirectness and imprecision for each behaviour. Inconsistency assessed whether or not the reported effects for a behaviour differed between assessments of behaviour change and its determinants, as well as whether or not high statistical heterogeneity was observed within meta-analysis. Directness referred to how directly the evidence related to the health topics examined in this review within the UK context. Evidence on behavioural outcomes was considered ‘direct’ whereas evidence regarding awareness/knowledge, attitudes or intentions only was considered ‘indirect’. Similarly, if available evidence was primarily drawn from non-OECD countries then this was considered indirect. Imprecision was assessed on the basis of the precision of the effect estimate [e.g. did the 95% confidence interval (CI) exclude no effect?]. ‘Overall effect’ was assessed by taking into account the direction of effect for behaviour with consideration of the indirect outcomes and the risk of bias in the evidence available. When there was evidence at a low risk of bias that was directly observed for the behaviour of interest, with little imprecision and inconsistency, we considered this to have a high level of certainty. We downgraded to moderate, low or very low certainty if there was a high risk of bias (by two levels), indirect evidence (by two levels), inconsistency (by one level) and imprecision (by one level).

Reducing harmful behaviours

Fourteen reviews reported evidence on whether or not mass media campaigns reduced harmful behaviours, as outlined in Table 2. 27,35,38,40,41,43–45,55,62,63,65,66,68 Eleven focused on a specific health topic and three examined mixed health topics. 65,66,68 All 14 reviews included studies based in OECD countries, and seven included studies conducted in the UK. 27,41,43–45,55,68 Ten of the reviews were rated as having a low risk of bias27,35,38,40,43–45,55,62,68 and four were considered to have a high risk of bias. 41,63,65,66 Eleven reviews focused on a specific health topic and three examined mixed health topics. 65,66,68 Three reviews used meta-analysis,55,62,66 with the remainder presenting results in a narrative synthesis.

Effects on sedentary behaviour were examined in one review. A meta-analysis of studies based in OECD countries on the effect of mass media campaigns on physical activity in adults found evidence of mass media campaigns reducing sedentary behaviour [relative risk (RR) 1.15, 95% CI 1.03 to 1.30], with moderate heterogeneity observed (I² = 63%; p = 0.018). 55

Two reviews reported on whether or not mass media campaigns had an impact on illicit drug use. A meta-analysis of RCT studies of campaigns targeting young people (aged < 26 years) found no effect (standardised mean difference –0.02, 95% CI –0.15 to 0.12), but did find evidence of reductions in the use of illicit drugs in an analysis of non-RCT studies. 62 The other, a review of the effects of anti-illicit drug public service announcements on youth (no definition by age specified) found very small positive reductions in illicit drug use, with considerable inconsistency; however, it should be noted that this review had a high risk of bias. 63

Nine reviews (eight specifically focusing on tobacco27,35,38,40,41,43–45 and one examining a range of health topics68) examined the impact of mass media campaigns on tobacco use. All included OECD-based studies and five included UK studies. 27,41,43–45 Four reviews, all considered to have a low risk of bias, examined the impact on preventing smoking uptake in young people. The review by Richardson et al. ,44 which included one UK study, reported positive results for smoking prevention: the narrative synthesis found evidence to suggest that mass media campaigns can prevent the uptake of smoking in young people (evidence from one review and two studies) and that industry-sponsored studies are less effective (evidence from one study). The other three reviews – Wilson et al. 45 (which included one UK study), de Kleijn et al. 38 and Brinn et al. 35 – all reported mixed results.

Five reviews examined smoking cessation or quit rates. Four reviews with a low risk of bias that included UK or OECD studies reported mixed results. 27,40,43,45 The fifth review reported no effect on quit attempts; the review had a high risk of bias and the evidence was from one study conducted in the UK. 41 Finally, a review that examined a range of health topics reported evidence of mass media campaigns having a positive effect on the combined outcomes of smoking prevention and cessation. 68

The impact of mass media on the use of a combination of substances (alcohol, illicit drugs and alcohol) was examined by two mixed health topic reviews. 65,66 Although both of these reviews reported positive effects, both reviews were rated as having a high risk of bias.

Increasing healthy behaviours

Twelve reviews reported evidence on whether or not mass media campaigns can increase healthy behaviours (Table 3). Ten focused on specific health topics (either physical activity55–61 or sexual health52–54) and two examined a range of topics. 68,69 None of the included reviews focused exclusively on diet/healthy eating, but one of the mixed-topic reviews included evidence on diet-related behaviours. 69 Nine of the reviews included studies conducted in OECD countries,55–61,68,69 four included studies conducted in the UK55,57,58,60 and two did not report the countries. 53,56 Two of the reviews, focusing on sexual health interventions, comprised studies conducted in low- and middle-income countries. 52,54

Eight of the reviews examined whether or not there was evidence that mass media campaigns could increase physical activity. A range of physical activity outcomes were reported, including walking, overall levels of physical activity, and using the stairs. In reviews that examined impact on stair use, the mass media campaigns typically comprised ‘point-of-decision prompts’, such as posters in locations with high footfall (e.g. public transport hubs and workplaces), encouraging people to use the stairs rather than the lift or escalator.

Two reviews with a low risk of bias reported evidence that mass media campaigns increased walking behaviour. In a meta-analysis of four studies, Abioye et al. 55 found evidence that mass media campaigns could produce an increase in brisk walking (RR 1.53, CI 1.25 to 1.87), whereas Ogilvie et al. 61 found evidence from two studies that mass media campaigns increased the time spent walking. Two reviews with a low risk of bias, one focusing specifically on physical activity60 and one examining a range of topics,68 found that stair use was increased by mass media campaigns comprising point-of-decision prompts (e.g. signs and banners to encourage using stairs). A third review, with a high risk of bias, also reported evidence that mass media campaigns could increase stair use. 58

However, reviews that examined overall levels of physical activity or time spent in physical activity reported generally mixed evidence. A meta-analysis of four studies in one review with a low risk of bias found no clear impact on overall physical activity (RR 1.02, 95% CI 0.91 to 1.14; I² = 72%). 55 In contrast, a mixed-topic review at low risk of bias found a positive effect on increases in overall physical activity. 68 The evidence from three reviews with a high risk of bias that examined changes in physical activity behaviours was generally mixed. 56,57,59

Four reviews provided evidence on whether or not mass media campaigns could increase healthy sexual health behaviours. Three reviews with a low risk of bias examined the impact of mass media on condom use: two of these reviews conducted meta-analysis and found that media campaigns had a positive effect on condom use, with inconsistency in the effect estimates [odds ratio (OR) 1.39, 95% CI –1.06 to –1.83;53 and OR 2.01, 95% CI 1.42 to 2.84, OR 2.10, 95% CI 1.51 to 2.9154]. The third review, which was of mixed health behaviour topics, also reported positive effects on condom use. 69 The fourth review reported mixed results of the effect of mass media on sexual health behaviours;52 this review was found to have a high risk of bias.

Finally, a mixed-topic review with a low risk of bias reported that mass media campaigns could have a positive effect on the consumption of healthy food. 68

Treatment-seeking

Ten reviews provided information on treatment-seeking: six focused on treatment-seeking in relation to sexual health46–50,52 and four focused on treatment-seeking in relation to tobacco use (Table 4). 27,37,42,43 Seven of the reviews included studies conducted in OECD countries,27,37,42,43,47–49 and all seven included studies conducted in the UK. Six were at a low risk of bias27,43,46–49 and four were at a high risk of bias. 37,42,50,52

Of four reviews examining the impact of media campaigns on the use of sexual health services or clinics, one found a positive effect47 and one reported mixed results. 49 Positive results were reported in two further reviews,50,52 but results were from only one study in each review and both reviews were at high risk of bias. The effects of mass media campaigns on the uptake of HIV testing or HIV services was examined in two reviews with a low risk of bias, both reporting mixed evidence. 46,48

There was evidence of mass media campaigns having a positive effect on calls to smoking quitlines from two reviews with a low risk of bias,27,43 although this was based on only one study in one of the reviews. 27 Mixed evidence was reported for the impact of mass media campaigns on smoking quitlines in two reviews with a high risk of bias. 37,42

Intentions

Seven reviews examined whether or not there was evidence that mass media campaigns had an impact on intentions to change behaviour (Table 5). 35,39,47,59,62,63,69 All of the reviews included studies from OECD countries but none included studies from the UK. Statistical methods were used in two reviews to assess the impact of mass media campaigns on illicit drug use intentions. 62,63 The remaining five reviews used narrative synthesis. Most of the reviews were of good quality (at a low risk of bias).

Three reviews with a low risk of bias examined tobacco use: two focused solely on tobacco35,39 and one mixed-topic review included tobacco. 69 Positive results for intentions to quit or to smoke were reported in two of the reviews,39,69 whereas one review that focused on reducing smoking prevalence in young people reported largely mixed results for the intention to start smoking. 35 The quality of the included studies was assessed by the reviews themselves as medium to low.

Statistical pooling in two reviews, one at low risk of bias62 and one at high risk of bias,63 found a mixed impact of mass media campaigns on illicit drug use intentions (including not to use, to reduce use or to stop use), with no clear indication of either a positive or negative overall effect.

One sexual health review with a low risk of bias reported largely mixed results for intentions to use contraception,47 whereas a physical activity review reported largely positive results for intentions to be more active,59 but the review had a high risk of bias.

Awareness and knowledge

Fifteen reviews reported on whether or not mass media campaigns had an impact on awareness and knowledge (Table 6). 27,35,39,44,46,47,50–52,57,62,65–68 The reviews had varying levels of relevance to the UK context: three reviews included non-OECD country research only, five reviews included one or two UK studies and the rest were reviews of studies from mainly OECD countries. Two reviews presented statistical results, with the remaining reviews presenting only narrative results. 51,66

There was evidence that mass media campaigns increased knowledge and awareness in relation to sexual health (including knowledge of HIV prevention and transmission, contraception and services). One low-risk-of-bias meta-analysis of 54 studies found consistent positive results for the improvement in knowledge of HIV transmission (d+ = 0.30, 95% CI 0.18 to 0.41, based on 47 reports) and prevention (d+ = 0.39, 95% CI 0.25 to 0.52, based on 65 reports). 51 Positive results regarding sexual health awareness and knowledge outcomes were also reported in four reviews using narrative synthesis,46,47,50,52 but three of these46,50,52 did not include any studies from the UK or other OECD countries, one review had a high risk of bias,52 and both Speizer et al. 52 and Kesterton and Cabral de Mello50 based their results on only one study. The review by Bertrand and Anhang46 also reported some mixed results.

Four reviews with a low risk of bias found mixed evidence that mass media campaigns could improve awareness and knowledge regarding tobacco. Two reviews, which both included studies from the UK,27,44 reported mixed results, whereas the third review reported positive results39 and the fourth review reported negative results. 35 A mixed-topic review with a low risk of bias that examined the effects on knowledge of smoking cessation helplines reported positive results. 67

The effects on knowledge of illicit drugs were examined in one illicit drugs review with a low risk of bias, which reported mixed results. 62 In addition, two mixed-topic reviews65,66 examined the effects on tobacco, alcohol and illicit drugs knowledge. A meta-analysis of the effects on drugs knowledge reported positive results [Δ = 0.05 standard deviation (SD); p < 0.05]66 and a narrative review also reported positive results;65 however, both of these reviews had a high risk of bias.

There was weak evidence that mass media campaigns could have an impact on awareness and knowledge regarding physical activity. Overall positive results, including from UK studies, were reported in one mixed-topic review with a low risk of bias that examined this outcome,68 whereas positive results were also reported by Finlay and Faulkner,57 but the review had a high risk of bias.

Finally, two of the mixed-topic reviews examined evidence of the impact on diet-related awareness and knowledge, both reporting positive results;67,68 the review by Mozaffarian et al. 68 included UK studies.

Attitudes, beliefs, norms and self-efficacy

Ten reviews reported on whether or not mass media campaigns had an impact on attitudes, beliefs, norms and self-efficacy (Table 7). 27,35,44,46,47,52,62,65,66,68 One review conducted a meta-analysis;66 however, only narrative results were presented in the other nine reviews. Most of the reviews were of good quality (low risk of bias), but their relevance to the UK varied.

For illicit drugs, the evidence was mixed. A mixed-topic meta-analysis that examined effects on drug use attitudes reported overall positive results (Δ = 0.02 SD; p < 0.05),66 but the review was at high risk of bias, whereas mixed evidence of impact on attitudes to illicit drug use and perceived peer norms was reported in a review with a low risk of bias62 and in a review with a high risk of bias. 65

For sexual health, overall positive results regarding beliefs about risk of pregnancy were reported in one review with a low risk of bias. 47 Positive results regarding other attitude changes were reported in two other sexual health reviews,47,52 but Speizer et al. 52 included only low-income countries and in both cases the reported results were from only one study. Mixed results were reported for the impact on self-efficacy, and negative results were reported for the impact on beliefs by Bertrand and Anhang,46 but this review was limited to low-income country studies and, therefore, it is of less relevance.

The evidence was mixed for tobacco. Three reviews, two including UK studies, reported overall mixed results for the impact on attitudes,27,35,44 and Brinn et al. 35 also reported overall negative results for the impact on self-efficacy. 35 However, a mixed-topic review including UK studies that examined the impact on attitudes to smoking reported overall positive results. 68 The same review also reported overall positive results for attitudes to physical activity.

Summary of the approach to subpopulations in reviews

The majority of the included reviews provide evidence about whether or not the effects of mass media campaigns were comparable across one or more subpopulations. Reviews differ in the extent to which the identification and synthesis of subpopulation differences formed a primary objective. In several reviews, all focusing on tobacco control campaigns, the main aim was to determine the equity or inequity of effects of campaigns across socioeconomic groups. 37,40,42 Some reviews dedicated part of their synthesis to looking at effects in specific subgroups,43–45,54,62 or to looking more generally for factors that moderate sizes of effect51,55 or described results separately for subgroups when this was shown in the original papers. 27,58,61,63 Most reviews provide a narrative synthesis of results for different subpopulations as described by the original studies; very few have conducted a formal statistical subgroup analysis. Some reviews that have included a meta-analysis have examined the factors that cause heterogeneity in study findings,51,55 or analyse in subgroups when available from the original studies. 54,62 A few reviews simply highlighted the subgroups in which statistically significant effects had been found in the original studies; if this was not part of a more formal subgroup analysis, these results have not been included.

When reviews focused on effects of mass media campaigns in a particular target population, those effects have been described earlier according to the relevant outcomes.

The majority of reviews concentrated on behaviour change outcomes, either reducing harmful behaviour or increasing health behaviour, rather than proximal outcomes, when describing and synthesising effects in subpopulations.

Description of the subpopulations that have been considered

The subpopulations considered differ markedly according to health behaviour, with sex27,43,44,51,54,55,58 and age27,44,45,51,55,62 being the only common factors across a number of reviews in different areas. Differences have also been examined according to ethnicity for several health behaviours. 27,39,44,58 Consideration of socioeconomic factors and the equity of effect across socioeconomic groups has been exclusively a feature of reviews of the tobacco control literature, in line with the strong socioeconomic differential in the pattern of smoking and smoking-related morbidity in many developed countries. 27,37,40,42,44 Other subpopulations have been defined according to the pre-campaign level of behaviour, for example by the level of initial physical activity or obesity for campaigns aimed at improving physical activity,58,61 by prior sexual health behaviour for a review of campaigns relating to sexual health,51 and a review of campaigns relating to illicit drugs examined effects according to sensation-seeking behaviour. 63

Effects by subpopulations

Alcohol consumption

There was little evidence of reductions in alcohol consumption associated with exposure to campaigns based on 13 studies reporting this outcome. 79–82,84,88,91–93,95,97,100,102 Six of these studies compared exposed and non-exposed groups, or exposed groups over time; five reported no statistically significant differences in consumption. 79,81,82,84,95 One study found that participants in a group exposed to a television and mailed letter campaign decreased their consumption on a typical day by 47%, contrasting with increases in the groups receiving either the television or letter components or neither. 88 Other studies examined associations between campaign viewing or awareness and consumption: one reported that increases in awareness in older adolescence, but not younger adolescence, was associated with decreases in binge drinking,80 one reported that campaign viewing significantly predicted drinks consumed102 and two found no significant difference in consumption. 91,97

Treatment- and information-seeking

There was some evidence, from four weak-quality studies, that campaigns generated increases in treatment-seeking or information-seeking. 85,87,92,97 For example, referrals for alcoholism were reported to have increased by 65% following a campaign. 97

Intentions and motivation

Three studies reported intentions to reduce alcohol consumption. In one study, those who reported that they had seen a campaign reported an increase in intentions to decrease alcohol use whereas others decreased their intentions. 102 Another study found an increase in the proportion of people reporting that they were likely to reduce their consumption from pre test to post test. 91 In the other study, the impact of the campaign on intentions was not well described. 84 In a single study reporting motivation to reduce consumption, half of drinkers recognising the campaign reported that it made them feel either very or somewhat motivated to reduce their consumption. 91

Beliefs and attitudes

Five studies reported alcohol-related beliefs or attitudes with mixed findings. 81,84,88,90,97 For example, one study reported an increase in support for some policies aimed at limiting consumption but not for others. 81

Knowledge

There was evidence that alcohol-related knowledge increased, based on eight studies. These included knowledge of unit consumption guidelines,91,92 cancer risk,91 the risks of alcohol use during pregnancy,83,89,93 sources of support for problem drinkers97 and general knowledge about alcohol. 84,96 For example, there was a significant improvement in knowledge of the risks of alcohol use during pregnancy in an exposed group compared with a control group. 83

Other outcomes

Other studies reported self-efficacy,79,96 social norms,102 campaign interaction or discussion,82,83,86,98,99 recall or recognition of campaigns82,83,85,88,89,91,92,94,95,97–100 and attitudinal or emotional responses. 79,92 The evidence was strongest for recall, with the proportions of people remembering campaigns or their messages being generally high.

Literature searches for reviews of economic evaluations

We combined relevant reviews identified in the original search (see Chapter 2) with additional reviews and systematic reviews identified using a literature search that specifically sought to identify reviews of economic evaluations published between January 2000 and January 2017. We conducted a rapid literature search using Web of Science and Google Scholar. The search strategy is summarised in Appendix 5.

Inclusion and exclusion criteria

We included reviews or systematic reviews describing the cost-effectiveness of paid mass media interventions (including multicomponent interventions in which a major component was mass media) for the relevant health topics. The cost-effectiveness findings had to be described in a results section, even if the findings were that there were no relevant studies. Included reviews were required to assess economic studies that evaluated both costs and benefits of mass media campaigns (i.e. full economic evaluations, not just intervention costs or cost savings). Systematic reviews were defined as those including both a systematic literature search and quality assessment of included papers.

We excluded reviews of reviews and papers that assessed industry-funded mass media campaigns.

Synthesis of findings

Of the 20 reviews that met the inclusion criteria, 13 were systematic reviews27,35,48,61,103–111 and seven were non-systematic reviews. 24,112–117 The characteristics of the reviews are summarised in Tables 9 and 10.

Eight reviews were on the topic of smoking,24,27,35,104,110,111,113,114 four were on sexual health,48,103,107,109 two were on physical activity61,106 and two were on diet. 116,117 Four reviews covered multiple behaviours. 105,108,112,115

Four of the included reviews had mass media campaigns as the sole focus and reviewing economic evaluations as a specific aim of the review. 48,104,105,111 Four of the reviews had mass media campaigns as the sole focus but did not have reviewing economic evaluations as a specific aim of the review. 24,27,35,109 Ten of the included reviews did not have mass media campaigns as the sole focus but did have reviewing economic evaluations as a specific aim of the review. 103,106–108,110,112,113,115–117 Two of the reviews did not have mass media campaigns as the sole focus or reviewing economic evaluations as a specific aim of the review, but reported on economic evaluations in the results section of the review. 61,114

Eight of the included reviews included no relevant studies. 24,27,48,61,103,105,107,109 Nine reviews included one or two relevant studies. 35,106,108,110,113–117 Three reviews included three or more relevant studies, with a maximum of 11. 104,111,112 Of these, two reviews were on the topic of smoking104,111 and one was a mixed-topic review but all of the relevant included studies were on smoking. 112

A total of 15 individual primary studies were reported in the reviews. The characteristics of the primary studies are summarised in Table 11, based on the information extracted from the reviews. Eleven primary studies were on smoking,118–128 two were on physical activity129,130 and two were on diet. 131,132 The majority of the primary studies were from the UK and the USA.

Taken together, the reviews and the findings of the primary studies within the reviews provide moderate evidence that tobacco control mass media campaigns can be cost-effective. There is weak evidence in relation to diet (although this is restricted to reductions in salt intake) and physical activity. There is no evidence in relation to the cost-effectiveness of sexual health campaigns, despite efforts to identify such evidence in systematic reviews.

Theoretical frameworks

Given that interventions developed with an explicit theoretical framework are considered more likely to be effective than those lacking any theoretical base,133 we considered that it was important to identify any theories used in the development, implementation and evaluation of campaigns in relation to their effectiveness and whether or not they had used multiple theories to achieve multiple goals. We returned to all the original included review papers to search the full texts for theory in detail. The earlier data extraction did not always include mentions of theories within introductions and discussions of the papers. We defined a theoretical framework in broad terms as a set of interrelated concepts, definitions and propositions that explain or predict events or situations by specifying relations among variables. From our reading of the reviews, we anticipated that theoretical frameworks would encompass two broad areas – communication theories and behaviour change theories – and that these theories would offer different explanatory insights. For instance, communications theories might help us to understand communication processes in the context of campaigns, such as theories underpinning how best to create and target messages, whereas behavioural change theories might help explain how different people might respond to such messages. By being inclusive, we considered that this would provide a comprehensive analysis of the conceptual thinking underpinning any mass media interventions, and identify the important features that might move people through pathways from increased awareness to eventual behaviour change.

Mass media channels used and interactivity

Mass media campaigns can be run via traditional media channels, such as television, radio, cinema, newspapers, magazines and billboards, or via new digital media including websites, pop-up and banner advertisements, QR codes, viral marketing and social media. New media often feature an element of interactivity (e.g. liking, sharing or commenting on content and downloading campaign apps). This ability to actively engage with a campaign may be associated with the increased effectiveness of a public health campaign. We have assessed the evidence relating to the different media channels used in campaigns, including the types of channels used and the use of single versus multiple channels, and on the use of new interactive media.

Intervention duration and intensity/exposure

Duration refers to the overall length of time that a campaign is in operation. Intensity and exposure refer to the overall ‘strength’ or ‘dose’ of a campaign, and can be measured in a number of different ways: the number of different advertisements or media materials produced, the number of times an advertisement is shown, the amount of media spend, the size of the audience reached by the campaign or the percentage of the audience in a given region or area that had a potential opportunity to see the campaign. Duration and intensity/exposure are often closely related, in that the longer a campaign runs, the more likely the audience is to be exposed to it. The two are therefore discussed together in this chapter.

Intervention content: messages, targeting and source

We have examined the evidence on the impact of different types of campaign content. We defined ‘message’ as referring to the types of messages used in campaigns (e.g. ‘positive’ vs. ‘negative’ messages, or messages containing information on the effects of the behaviour or advice on how to change behaviour). ‘Targeting’ referred to whether the campaign as a whole or specific campaign elements were intentionally targeted at specific population groups. Because the strategic decision regarding targeting was often strongly bound up with choice of messages, these two elements are discussed together. ‘Source’ refers to the apparent source of the message, and includes the person delivering the message (e.g. the use of actors or testimonials from real people) and the campaign funder (e.g. government, charities or pharmaceutical companies). As the source is likely to be associated with the type of message used and the intended target group, these elements are discussed together.

Intervention scale: local, regional and national campaigns

Even in the context of evidence that mass media campaigns featuring particular characteristics are likely to be effective, decision-makers need to consider how the scale of a campaign may influence its impact. We review the published evidence that considers the differential impact of local, regional and national campaigns.

Use of theory: findings from the review of reviews

Table 12 summarises the number of included reviews of mass media campaigns that examined various aspects of theory, and the various behaviour change theories mentioned in the reviews.

The majority of these reviews did not include a focus on theories or conceptual frameworks or models and, therefore, there was little evidence on the role that theory plays in the effectiveness of the intervention. Eight reviews incorporated theory into the process of their work. Three used logic models to frame their research questions and to identify the outcomes relating to their research. 47,56,58 These models were not adjusted in the light of the results from the review. One review used the presence of a theoretical framework as one of its inclusion criteria, but did not refer to the theories in the primary studies in the results section. 65 Leavy et al. 59 used the hierarchy of effects (HOE) framework to specify levels of impact to typify outcomes. Matson-Koffman et al. 60 and Robinson et al. 69 used theoretical framings to provide the rationale for their reviews. Finally, Werb et al. 63 used theory to explain the findings of their review.

Seven reviews listed the theories they found in their primary studies. Five of these listed behaviour change studies27,35,48,57,59 and four mentioned communication theories, notably social marketing. 27,48,57,59 LaCroix et al. 51 mentioned that 45% of the studies in their review were theory based.

In terms of relating theory to effectiveness, in their analysis of the effects of mass media campaigns on drug use, Ferri et al. 62 discovered that two studies based on social learning theory and the social ecological framework produced better results, whereas the study based on the social influence approach favoured the control group.

Mass media channels used and interactivity: findings from the review of reviews

Table 13 summarises the numbers of included reviews that examined media channels or interactivity as potential mediators of effectiveness.

Although most of the reviews included information on the media channels used in the individual included studies, most simply listed them when summarising the characteristics of the included studies, and did not examine media channels or degree of interactivity as mediators of campaign outcomes, or discuss the potential relationship between channels and interactivity and effectiveness. The lack of attention paid to interactivity is unsurprising in the older reviews whose included studies would have predated the development of new interactive media. Findings from the five reviews that reported findings specifically relating to channels are summarised below.

Derzon and Lipsey66 conducted a meta-analysis comprising 72 studies of mass media substance use campaigns, with media channels used (radio, television, video or print) as one of the sets of variables in the analysis. Overall, the analysis found that reduced substance use behaviour was associated with exposure to all the different media channels, but that radio was associated with the greatest relative effects (Δ = 0.10) and print with the least (Δ = 0.04). For attitude outcomes, the relative effects were greater for those exposed to video and print rather than other channels. Video was associated with particularly large relative effects on substance use knowledge. In conclusion, the review authors stated that with regard to media channel, messages communicated via video were associated with greater improvements in knowledge, attitudes and behaviour than campaigns that did not use video. It is likely that most of the video-based campaigns would have been implemented in school settings and would have involved discussion and other informational and educational input, in other words in a quite different context from general population mass media campaigns on broadcast media. The age of the Derzon and Lipsey66 review (published in 2002) means that some of the studies are likely to be quite old, which again limits the relevance of the findings to the present UK context.

A review of mass media campaigns on HIV/acquired immunodeficiency syndrome (AIDS)-related behaviour among young people in developing countries specifically examined the strength of the evidence for the effects of three types of mass media interventions (radio only, radio with supporting media or radio and television with supporting media) and assessed whether or not these interventions reach the threshold of evidence needed to recommend widespread implementation. 46 One of the studies included in the review was radio only. Six of the studies evaluated interventions using radio with other supporting media (e.g. written materials, videos, posters, theatre performances and school workshops). The remaining eight interventions involved television and radio with other supporting media.

The review reported that the one radio-only campaign showed mixed results regarding awareness of a helpline but no improvements in social norms, interpersonal communication about HIV/AIDS or various measures of condom use. 46 Of the six studies using radio with other supporting media, all reported some measure of knowledge gain, although results tended not to be significant, and most studies showed some positive effects on skills, knowledge, interpersonal communication and social norms. Findings relating to various sexual behaviours were mixed, although the weight of the evidence across studies reporting condom use was strongly positive.

Eight studies examined the effect of campaigns using television and radio with other supporting media. Evaluations of this type of intervention generally showed improvements in knowledge and skills related to HIV/AIDS, knowledge about health services, interpersonal communication regarding HIV/AIDS and social norms. Data on different measures of condom use showed positive effects in the majority of studies, although evidence regarding other behaviour change (e.g. reduced age of sexual behaviour, number of partners and abstinence) varied but leaned towards having no effect. The review authors concluded that ‘Campaigns that include television require the highest threshold of evidence, yet they also yield the strongest evidence of effects’. 46 The findings have limited relevance for the UK and OECD countries.

In their review of mass media interventions targeting young people smoking, Richardson et al. 44 reported evidence from a single study on mass media channels, one that examined the effects on youths of antismoking cinema advertisements played during a film. Effects included increased disapproval of smoking in films and, among smokers, a desire to quit. In a wide-ranging review of mixed health behaviours, Mozaffarian et al. 68 noted that mass media and education campaigns using ‘multiple modes’ (described as including ‘print, radio, Internet, television, social networking, other promotional materials’) demonstrated effectiveness for increased consumption of specific healthy foods, reduced consumption of less healthful foods and drinks and the promotion of physical activity.

In a review that aimed to synthesise evidence evaluating the effectiveness of mass media interventions on helping people to quit smoking/tobacco use and/or to prevent relapse, Jepson et al. 43 presented data in a format that organised the results by channel of mass media. Interventions that drew on two or more forms of the media (such as newspaper, television and billboard advertising) were referred to as ‘multichannel’ mass media. A particular emphasis was placed on evaluating relevance to the UK setting. Interventions were examined in terms of both the effectiveness of the channel of communication and the effectiveness of message content (see Intervention content: findings from the review of reviews). The review noted that many of the studies used multiple types of media combined with other interventions, which made it difficult to evaluate which particular component was effective or ineffective. It reported some good-quality evidence that the use of technology such as mobile phones could be effective, and offered the potential to deliver culturally specific materials to targeted groups. It also reported evidence that the internet could be an effective way of delivering interventions and may be a particularly appealing channel of communication for young people. Three studies, probably relevant to the UK, found an effect of multichannel mass media on smoking cessation, but there was no evidence about which of the mass media components of the interventions were most effective (or most ineffective), and the review authors cautioned that the results of these studies should be interpreted with caution owing to their poor methodological quality.

One review examined interactivity as a potential moderator of effectiveness. In a meta-analysis of the effects of new-media interventions on sexual health behaviours by Swanton et al. ,53 the interventions examined included websites, SMS (short message service) messages, social media and online video. Components of interventions were coded as interactive (e.g. personalised e-mails) or static (e.g. watching an online video). Moderation analyses found a variation in the effect of new-media interventions on condom use depending on the interactivity of the new-media channel used:53

Interventions using interactive components yielded significant effects (OR 1.79, 95% CI 1.15 to 2.77; p = 0.01); however, this effect was significantly heterogeneous [Q (6) = 28.03; p < 0.01, I² = 78.59%], suggesting that further factors may influence the size of the effect. Interventions using static content did not yield significant effects, and the effect size was homogeneous.

Swanton et al. 53

In conclusion, the review found that interventions that used interactive components were shown to be more effective than static components in improving condom use.

Intervention duration and intensity/exposure: findings from the review of reviews

Table 14 summarises the number of included reviews of mass media campaigns that examined intervention duration or intensity/exposure.

Six of the 36 reviews examined or commented on the relationship between intervention duration or intensity/exposure and effectiveness. Information on duration or intensity/exposure in individual studies was extracted in other reviews, but no attempt was made in the review analysis or synthesis to draw conclusions about the relative effectiveness of different levels of duration or intensity/exposure.

Two of the reviews, both examining sexual health campaigns, conducted statistical analysis to examine whether or not intervention effectiveness was moderated by intervention duration. 51,53 Both found some evidence that effectiveness increases with greater duration and intensity/exposure, although for only some of the outcomes examined. In the meta-analysis of the effects of new-media interventions on sexual health behaviours by Swanton et al. ,53 the interventions examined included websites, SMS messages, social media and online video. They were coded into three duration categories (a single session, ≤ 6 months or > 6 months in duration) and the relationship between duration and effectiveness was examined in moderator analysis. Intervention duration was not related to effectiveness in interventions targeting condom use, but was related to effectiveness when the outcome was participation in testing for sexually transmitted infections (STIs): single-session interventions were less effective than longer interventions (up to 6 months in duration, OR 1.64, 95% CI 1.36 to 1.98; p < 0.01; and > 6 months in duration, OR 2.17, 95% CI 1.36 to 3.47; p < 0.01). A meta-analysis of the impact of sexual health mass media interventions on condom use and HIV-related knowledge51 examined whether or not effectiveness was moderated by various intervention characteristics, including duration and proportion of target population exposed to the campaign. The relationship between campaign characteristics and magnitude of effects was examined using a modified weighted least squares regression analysis. Greater increases in condom use were seen following longer campaigns (k = 51, β = 0.48, p < 0.001).

Three reviews [one National Institute for Health and Care Excellence (NICE) rapid review44 and two Cochrane Database Systematic Reviews27,35] provide some evidence to suggest that intervention duration and intensity/exposure are associated with effectiveness in mass media campaigns targeting tobacco use, but do not examine the relationship statistically. A NICE rapid review of the effectiveness of mass media interventions designed to prevent the uptake of smoking in children and young people examined whether or not the intensity of interventions influenced effectiveness or duration of effect. 44 Evidence statements in the review stated that:

The duration of a mass media intervention influences its effect. Increased exposure to anti-tobacco messages over time decreases intent to smoke and smoking initiation, meanwhile increasing negative attitudes towards the tobacco industry.

© NICE 2008. Interventions to Prevent the Uptake of Smoking in Children and Young People. 44 Available from www.nice.org.uk/guidance/ph14/evidence/review-of-effectiveness-pdf-369941869. All rights reserved. Subject to Notice of rights. NICE guidance is prepared for the National Health Service in England. All NICE guidance is subject to regular review and may be updated or withdrawn. NICE accepts no responsibility for the use of its content in this product/publication

Underpinning support for the evidence statement came from a Cochrane Database Systematic Review (Sowden 1998 as cited by Richardson et al. 44), suggesting that intervention duration was an important influence on behaviours, and from three cross-sectional studies, all conducted in the USA, which found that increased exposure to antismoking advertisements over time resulted in a decrease in young people smoking in the past 30 days (compared with those in markets with no exposure to state-sponsored anti-tobacco laws), intent to smoke, initiation of smoking, enhanced perception of risk and negative attitudes about smoking. The same NICE rapid review44 also examined facilitators of and barriers to implementation, and noted that lack of exposure and longevity are barriers to effective mass media interventions, noting that one-off campaigns are not likely to induce behaviour change and that:

Messages must appear frequently enough for audiences to notice them and internalise them.

© NICE 2008. Interventions to Prevent the Uptake of Smoking in Children and Young People. 44 Available from www.nice.org.uk/guidance/ph14/evidence/review-of-effectiveness-pdf-369941869. All rights reserved. Subject to Notice of rights. NICE guidance is prepared for the National Health Service in England. All NICE guidance is subject to regular review and may be updated or withdrawn. NICE accepts no responsibility for the use of its content in this product/publication

Bala et al. 27 examined the effectiveness of mass media interventions for smoking cessation in adults in a Cochrane Database Systematic Review. From reviewing 11 campaigns, the review concluded that the intensity and duration of campaigns may influence effectiveness, but length of follow-up and concurrent secular trends and events could make it difficult to quantify the effects. The review cited a UK study (McVey 2000 as cited by Bala et al. 27) that compared the impact of single- and double-weight television campaigns on quit rates and found no significant differences at 6 months, with the single-weight region (Granada) at 6.3% and the double-weight region (Tyne Tees, Yorkshire) at 6.6%, yielding an adjusted OR of 1.02 (p = 0.94). It also discussed studies in other countries that appeared to support the use of longer running or more intense campaigns, but noted that effects may have been confounded by the concurrent presence of other intervention elements such as face-to-face counselling. The authors concluded that ‘The duration and intensity of an intervention may affect its impact on smoking behaviour, but evaluations need to last long enough to detect lasting changes, and to allow for confounders and for secular trends’. 27

Another Cochrane Database Systematic Review examined the effectiveness of mass media interventions for preventing smoking in young people, with seven studies, all using a controlled design, meeting all of the inclusion criteria. 35 The three effective campaigns were all ‘of reasonable intensity over extensive periods of time’; in contrast, the campaigns that did not report positive findings, much more heterogeneous in duration and generally shorter. The three effective campaigns comprised (1) 190 television, 350 cable television and 350 radio spots purchased in each of the 4 years during which the campaign was running (Flynn 1995 as cited by Brinn et al. 35); (2) 167 television and cinema spots in three annual campaigns (Hafstad 1997 as cited by Brinn et al. 35); and (3) television (local, cable and network), radio, websites, magazines, movie theatres and several other media used over 6.5 years (the National Youth Anti-Drug Media Campaign) (Longshore 2006 as cited by Brinn et al. 35).

Finally, a wide-ranging mixed-topic review examining the effectiveness of a range of population approaches reported ‘strong evidence’ that sustained (i.e. longer duration) campaigns are important, particularly for reducing smoking. 68 The review also noted that some of the campaigns found to be effective in improving diet knowledge and behaviours were implemented for many years, suggesting that ‘behavioural changes are sustainable when media and educational campaigns are continued’. However, the effects of mass media campaigns were not distinguished from those of campaigns using multiple approaches including mass media, and the review conducted no statistical analysis of the moderating role of campaign duration. It recommended that media and education campaigns to improve diet, physical activity and smoking should be sustained and should use multiple modes/channels.

Intervention content: findings from the review of reviews

Table 15 summarises the reviews that assess the impact of message type, targeting and source.

Although the majority of reviews provide some detail on the content of mass media interventions, often including some detail about the message and target population for individual studies and campaigns, only 12 provided some degree of analysis or synthesis of the influence of message type on behavioural or other outcomes. A further 10 provided analysis or synthesis of the effectiveness of the targeting of campaigns. The main focus of three reviews was the effectiveness of targeting campaigns towards specific groups (girls, indigenous populations and disadvantaged groups), whereas none focused solely on the characteristics of effective messages. There is a clear overlap between reviews that provide information on effect of message type and effect of targeting, which is reflected in the summary below.

Intervention scale: findings from the review of reviews

None of the reviews examined scale of campaign as a potential mediator of effectiveness. Although the reviews noted whether their included studies were implemented at local, regional or national levels, most simply listed the information and none of them provided any evidence or commentary regarding the relative effectiveness of campaigns implemented at these different levels or discussed factors that might be associated with effectiveness when campaigns are implemented at the different levels.

Identification of studies

Studies were eligible if the paper was published in or after 2011 and the study was conducted in the UK. Multicountry studies were eligible if findings for the UK were reported separately. The campaigns had to address one of our six health topics: alcohol use, illicit substance use, diet, physical activity, sexual and reproductive health and smoking cessation and prevention. Apart from targeted health topics, the study eligibility criteria for primary research studies evaluating the effectiveness of mass media interventions were the same as those used in the related review of mass media campaigns to reduce alcohol consumption. 17 Box 3 describes the mass media campaign intervention study eligibility criteria. Studies could report data or outcomes of any format for the following outcomes: behavioural or social cognitive outcomes, media outcomes (e.g. recall, understanding and onward transmission of the campaign) and distal (e.g. societal or policy changes) or process (e.g. cost-effectiveness) outcomes. Studies of multicomponent interventions were eligible if they assessed the specific effects of a mass media component, and published conference abstracts were eligible provided there was a description of the campaign and outcome data were reported.

The search strategy was a revision of that used for the project’s review of mass media campaigns to reduce alcohol consumption, combining terms for mass media and health communication campaigns and for the target health topics with a UK studies search filter (see Appendix 6 for an example strategy). Sixteen academic databases were searched on 7–9 September 2016 and the results were uploaded to an EPPI-Reviewer 4 database and deduplicated (Figure 4):26

• Applied Social Sciences Index and Abstracts

• EMBASE

• Epub Ahead of Print, In-Process & Other Non-Indexed Citations, Ovid MEDLINE(R) Daily and Ovid MEDLINE(R) – 1946 to present

• Social Sciences Citation Index – 1900 to present, Arts and Humanities Citation Index – 1975 to present, Conference Proceedings Citation Index – Social Science & Humanities – 1990 to present, Book Citation Index – Social Sciences and Humanities – 2005 to present and Emerging Sources Citation Index – 2015 to present

• PubMed, via EPPI-Reviewer 4 PubMed direct import

• EBSCOhost Research Databases Database – Communication Source, Business Source Complete, ERIC

• The Cochrane Library, Cochrane Central Register of Controlled Trials, issue 8 of 12, August 2016; NHS Economic Evaluation Database; Health Technology Assessment Database, issue 3 of 4, July 2016; Cochrane Methodology Register, issue 3 of 4, July 2012.

FIGURE 4.

The PRISMA flow diagram of the identification and selection of primary studies (review D).

Records (n = 10,520) were screened for initial topic relevance, date and country by two reviewers (KH and KA). One per cent (n = 105) of randomly selected records were double-coded to pilot the screening process and check for inter-rater reliability. Once agreement was > 90% on included and excluded categories, then a proportion of the rest was allocated to each reviewer for single coding. The resulting 240 were retrieved as full texts and all double-coded by the reviewers (KH and KA) for inclusion; any final decision disagreements were resolved by a third reviewer (LB or MS). Twenty-seven papers covering 25 recent evaluations of UK mass media campaigns were included for analysis (see Appendix 7). 119,120,134–158 A list of UK primary studies excluded by full-text assessment is provided in Appendix 8.

The data extraction form was a revision of that used for the project’s review of reviews and designed to capture the study methods and outcome data as well as details of the mass media campaigns (the intervention) towards answering the project’s objectives. Studies were double data extracted independently by three reviewers [KA, KH and Joanne Morrison (JM)], who held telephone meetings in pairs to agree and finalise the data for each study. A pilot data extraction exercise, using two studies independently coded by all three reviewers, familiarised them with the process to ensure consistent interpretation of the headings. No study authors were contacted in order to obtain or confirm data.

In a departure from the review project’s protocol, two additional risk-of-bias appraisal tools were added after study selection to correspond to the included studies’ design. The proposed EPHPP Quality Assessment Tool for Quantitative Studies was used to assess experimental studies,78 with the additional US National Institutes of Health’s assessment tool for observational cohort and cross-sectional studies159 and the Critical Appraisal Skills Programme (CASP) tool for qualitative studies. 160 Studies were appraised for quality and relevance (at a study level) independently in duplicate by three reviewers (KA, KH and JM), who held telephone meetings in pairs to agree and finalise the risk-of-bias assessment for each study. The pilot exercise described above also included an independent appraisal of quality and relevance for the same two studies by all three reviewers.

This review of recent UK primary studies was designed to capture data or outcomes of any format for the following outcomes: (1) behavioural, social cognitive or media outcomes and (2) distal (e.g. societal or policy changes) or process (e.g. cost-effectiveness) outcomes. Thus, from its inclusive nature, we did not expect to be able to carry out a meta-analysis of the data.

Narrative synthesis

Evidence from the UK primary studies was reviewed and synthesised around the following characteristics of mass media campaigns: theoretical frameworks, intervention duration and intensity, mass media channels used and intervention content (messages, targeting and source) and intervention scale. How these characteristics were defined and analysed is described in Methods: systematic review of reviews.

Use of theory: findings from the review of UK primary studies

Table 16 summarises the numbers of included UK primary studies that examined various aspects of theory, and the various behaviour change theories mentioned in the studies.

Out of 23 primary studies, 17 did not include grand theories, conceptual frameworks or models to inform their mass media interventions. Of the six studies that did incorporate theory, three grand theories were identified: ‘social contagion theory’ to amplify a campaign by normalising a behaviour and turning it into a movement;120,136 ‘prime theory’ (plans, responses, impulses/inhibitory forces, motives and evaluations) to alter people’s desires to succeed in changing maintaining patterns of behaviour through increasing motivational inputs;120,136,152,153 and ‘theory of planned behaviour’ to alter behaviour through influencing people’s intentions to perform such behaviours. 139,140 One further study mentioned using theoretical ideas rather than identifying a specific grand theory, by drawing on ideas from social marketing and behaviour change, but in this study it was unclear how these ideas were used. 138 Across the studies, there was no mention of using other theoretical frameworks or logic models to inform their research questions or to identify the outcomes relating to their research.

In terms of how theory was used in the basis of the programme design, in the study by Brown et al. ,120 social contagion theory was used to piggyback onto an annual ‘no smoking day’ that aimed to help smokers stop by providing a nationally supportive environment and drawing attention to available treatments; they also developed a national cessation campaign (Stoptober) to generate a burst of activity around key time points when quit rates had reduced. In a later study, Brown et al. 120 used ‘prime theory’ to refine film content on the basis of focus group testing with a diverse group of smokers. 136 Using ‘theory of planned behaviour’ to alter behaviour through influencing people’s intentions, Eves et al. 139 tested the success of two different messages in encouraging more physical activity in members of the public.

Mass media channels used and interactivity: findings from the review of UK primary studies

Table 17 summarises the different mass media channels examined in the review of UK primary studies.

As can be seen, only one of the studies compared different media channels within the same study. 143 This was a tobacco control social media campaign that aimed to raise awareness about the health risks of waterpipe tobacco smoking. The number of Facebook (Facebook, Inc., Menlo Park, CA, USA; www.facebook.com) subscribers, Twitter (Twitter, Inc., San Francisco, CA, USA; www.twitter.com) followers and YouTube (YouTube, LLC, San Bruno, CA, USA; www.youtube.com) views all increased over time (measures taken at 3, 6 and 9 months). In a descriptive comparison of interactions with the social media channels, the authors concluded that Twitter provided the most organisation-based contact (e.g. 8% of tweets were retweeted, and nearly two-thirds were on health effects of waterpipe tobacco and other Twitter users interacted with @shishaware 70 times), whereas Facebook was the most interactive channel (e.g. 69% of status updates had more than one ‘like’, 23% had more than one comment from users). The YouTube channel had also been ‘favourited’, had ‘likes’ and ‘dislikes’ and had a rate of 112.2 comments per 10,000 views.

The remainder of the UK primary studies reported the channel(s) used in the campaigns, but did not provide evidence comparing or commenting on the effectiveness of different channels.

Intervention duration and intensity/exposure: findings from the review of UK primary studies

Table 18 summarises the number of UK primary studies that examined intervention duration or intensity/exposure.

Eleven of the 25 recent UK primary studies commented on or examined the relationship between mass media intervention duration or intensity/exposure and effectiveness.

A series of six related statistical studies evaluated televised tobacco control campaigns broadcast in England (2002–10) by examining existing data sources and indicators of smoking behaviour against detailed population-level campaign exposure data. 144,148,152–155 Exposure data used the standard advertising industry measures of a campaign’s reach [gross ratings points (GRPs)] and its reach and frequency [television ratings (TVRs)]. The standard advertising industry measures of GRPs are equivalent to the summed ratings of individual advertisements across multiple campaigns, giving a per capita measure of advertising exposure. TVRs are defined as the percentage of a particular audience that has seen a commercial break.

An analysis using 4 years of longitudinal panel data from the International Tobacco Control UK Survey found that increased exposure (as measured by GRPs) was associated with higher recall at 6 months, although participants were recalling any advertising or information that talked about the dangers of smoking or encouraged quitting on television – not specifically tobacco control campaigns. 153 The effect was significant only for recent exposure (OR 1.51, 95% CI 1.14 to 2.01); exposure in the 4–6 months before the survey had no impact on recall.

An analysis of monthly Opinions and Lifestyle Survey data found a small association between increased exposure (by 400 GRPs) to television adverts, equivalent to all adults in the population seeing four tobacco control television adverts per month, and a 3% lower odds of smoking (OR 0.97, 95% CI 0.95 to 0.999; p = 0.044) a couple of months later, and a 1.80% (95% CI 0.47 to 3.11; p < 0.01) decline in average daily cigarette consumption 1 month later. 155 Further analyses of exposure by emotive characteristics of the messages in the advertisements are reported in Intervention content: findings from the UK primary studies. 154

In an analysis of time series data of quitline calls and NRT over-the-counter sales and prescribing data against exposure to anti-tobacco mass media advertising and smoking cessation medication advertising,144 there was some evidence that effects on those information- and treatment-seeking behaviours tended to peak in periods of greater campaign exposure (TVRs were higher in January and were highest in January 2005 and 2010). This suggests that exposure is associated with effectiveness. More specifically, a 1% increase in tobacco control TVRs led to a 0.129% increase in quitline calls in the same month (or a 0.085% increase in the seasonally adjusted model, p = 0.007). For NRT, in most years there was a clear peak in prescribing in the first 3 months of the year and a much smaller peak in October and in January to March each year there were generally peaks in over-the-counter NRT sales. A 1% increase in pharmaceutical company TVRs led to a 0.05% increase in sales in the same month; however, this finding was not statistically significant in the adjusted model. A further study evaluated the impact of different types of message in national campaigns on calls to the English NHS Stop Smoking telephone helpline. 152 Exposure to all types of campaign (as measured by GRPs) was associated with increased calls to the helpline (see Intervention content: findings from the UK primary studies for findings for message types).

Looking at the impact of the campaign on smoke-free homes, Lewis et al. 148 found that level of exposure to all tobacco control campaigns (measured by GRPs) was not associated with increased likelihood that a home would be smoke-free, but level of exposure to second-hand smoking campaigns was associated with an increased likelihood that a home would be smoke-free, with increased exposure being associated with increased odds (by 7% for each additional 100 GRPs) at a 1-month lag after the campaigns (OR 1.07, 95% CI 1.01 to 1.13; p = 0.033). There was no significant association during the same month of the campaign (OR 0.99, 95% CI 0.93 to 1.05; p = 0.740) or 2 months after (OR 0.98, 95% CI 0.92 to 1.04; p = 0.490), suggesting that the impact of the second-hand smoking campaign did not happen immediately and that it also fades quickly.

A series of naturalistic studies evaluating local point-of-choice prompts in public places (e.g. multistorey workplaces and railway stations) to encourage stair use for increasing physical activity varied the numbers of posters and signage shown throughout the experimental period. 140,145,147 This variation in intensity of exposure was tested alongside variations in message types, and the findings are reported in Intervention content: findings from the UK primary studies.

One study evaluated a regional 10-month sexual health poster, leaflet and online campaign (2009–10) in the west of Scotland aimed at men who have sex with men. 141 Survey results were analysed by the extent of respondents’ exposure to the campaign (their recall and recognition of the campaign). At the end of the campaign, greater exposure was associated with positive changes for some but not all outcomes. For example, those with no exposure were significantly less likely to report always using appropriate lubricant than those with low exposure (adjusted OR 0.42, 95% CI 0.23 to 0.77; p = 0.005); however, those with high exposure to the campaign did not differ significantly from those with low exposure.

Finally, one study evaluating a 2003–6 national mass media campaign to improve diets reported behavioural outcomes annually throughout the radio, television and online campaign. 137 The campaign to encourage consumption of five portions of fruit and vegetables per day used ‘mini campaign relaunches’ every 6 months; although the overall aim was the same, each relaunch targeted a different subpopulation (by income, age or sex) and used slightly different methods and channels. Significant campaign effects for increased fruit and vegetable intake did not happen until the third year of the campaign and the effect was stronger for those in lower income groups. The fact that significant impacts did not emerge until the third year of the campaign may suggest that campaigns need to last a certain length of time before effects are felt, although the authors suggest that the delay in impact may have been partly attributable to the nature of the activities in different phases of the campaign, with earlier phases involving a lower investment in targeted advertising.

Intervention content: findings from the UK primary studies

Table 19 summarises the UK primary studies that assess the impact of message content.

The UK evidence on different types of message and campaign content is based on nine studies on two types of behaviour: physical activity (four studies) and smoking (five studies).

The UK has been exposed to very diverse tobacco control campaigns and is therefore a good setting in which to investigate the effect of different types of messages. Several studies on tobacco campaigns investigate the differential impact of ‘positive’ (eliciting happiness, satisfaction or hope) and ‘negative’ (eliciting fear, sadness, guilt, anger or disgust) messages152–154 and find that both are effective. Unlike negative messages, positive messages have not been found to have an effect on recall,153 but both positive and negative messages have been shown to influence information-seeking152 and smoking behaviour,154 with positive campaigns having a bigger effect. This indicates that recall does not necessarily translate into changes in behaviour, which should be taken into account in studies assessing the ‘effectiveness’ of campaigns by measuring recall.

There is evidence from one study that the source of the campaign is important,144 with government- and charity-funded campaigns being more effective at triggering quitting behaviour than pharmaceutical company-funded campaigns. None of the primary studies compared the effects of targeting on specific groups; however, one primary study reports that targeted campaigns about the risks of second-hand smoke exposure increased the odds of a home being smoke-free more than tobacco control campaigns overall. 148

The physical activity studies are based on poster campaigns to increase stair use. They suggest that multicomponent messages (e.g. providing information on calories burned as well as posters directing people to the stairs) delivered at the point of choice (or ‘volitional’ messages) are more effective than those that have a single component and/or are ‘motivational’. 140,145,146 One of the physical activity studies suggests that simple messaging may be more effective in busy settings. 147

Intervention scale: findings from the review of UK primary studies

None of the UK primary studies examined whether or not scale (i.e. whether a campaign was implemented at a local, regional or national level) was a potential mediator of effectiveness.

End-of-study event

On 28 September 2017, an end-of-study event was hosted in the conference centre at the Iris Murdoch Building, University of Stirling. A series of presentations on the project’s background, methodology and results was given by the research team, which represented all the academic teams involved (University College London and the Universities of Nottingham, Glasgow and Stirling). In addition to sharing the findings of the review, the purpose of the event was to discuss the findings with interested stakeholders and experts to identify strengths and limitations of the evidence, to identify implications for commissioning/using mass media campaigns and to identify recommendations for future research. The event was also intended to inform the final report for the study.

An open invitation (see flyer in Appendix 9) was issued to relevant policy and practice networks in the UK. Personal invitations were also sent to the study advisory group and key organisations that design and deliver mass media campaigns for public health, such as Public Health England and the Scottish Government. Forty-nine delegates attended the event. These included representatives from local and national charities and advertising agencies, NHS staff, civil servants, clinical academics, public health and marketing researchers, research students and the PPI representative for the study. Delegates had a range of experience, from a deputy director of a Scottish Government department to master’s students. There was also representation from across the UK: attendees had travelled from Northern Ireland, Orkney, Shetland, the north-east and north-west of England and London, in addition to those from other UK regions.

The event included a series of presentations from the research team on all the elements of the study. The PI opened the day and began by showing a series of videos of past mass media public health campaigns (Box 4). Presentations of study findings were then followed by question-and-answer sessions and small-group discussions with feedback to the room (see the event agenda in Appendix 10). Both the event’s presenters and attendees were encouraged to share slides and views via social media.

Key points raised among those attending related to a range of themes. Some of these focused on the limitations of the evidence our study identified. Overall, delegates had expected the evidence on the effectiveness of mass media campaigns to be more conclusive. There was a general perception, particularly among those involved in planning or delivering services or promoting public health interventions, that tobacco control mass media in general was highly effective and cost-effective and that this might be the case for mass media on other public health topics. Our review did find that tobacco control campaigns can be effective but this varied based on the type, duration and content of the campaign. Evidence for campaigns on other behaviours was also mixed and in some cases very limited. This challenged the views of some delegates and meant that some of the questions and discussion at the event were not straightforward.

Delegates had a range of questions and useful feedback on the limitations of our reviews and the evidence in general; for example, they discussed the parameters of our review of systematic reviews, the changing media landscape and how this might affect the available evidence, the time lag between campaigns and evaluations being published and the fact that the results of many campaigns, particularly at a regional level, are not captured in the peer-reviewed literature. Attendees also reflected on their own experience (in both research and practice) compared with the review evidence. This included issues to do with fitting mass media campaigns into wider political messages, the fact that diet and nutrition campaigns can come from commercial or industry sources (in contrast to the tobacco industry, which is not permitted to conduct marketing in the UK) and also that commercial advertising involves large budgets, which it is difficult for public health campaigns to counteract (alcohol advertising was highlighted as a particular example). There are also limited available resources for public health campaign evaluations and, when decisions have to be made, resources will focus on designing and delivering the campaign rather than research around it. Limited resources also result in difficult choices being made around the medium of communication, with a more recent focus just on social marketing or online campaigns. The literature identified by the study had limited evidence on campaigns run via digital media only, and evaluating these should be a priority for the future.

Discussions also focused on why the evidence differs between health topics; for example, for some health behaviours, it can be more difficult to define what the ‘message’ should be. Those planning campaigns may favour positive rather than fear-based messaging, but questions remain about which is more effective. With limited budgets, how should campaigns be targeted? The evidence is unclear for campaigns aiming to reach particular population groups. Some of the evidence in our review suggests that a whole-population approach may be desirable, but again this has cost implications. Finally, delegates discussed how indicative evidence identified in our reviews suggests that campaigns focused on increasing healthy behaviours may be more appealing or effective than those focused on reducing unhealthy behaviours. This may be influenced by some of the unhealthy behaviours involving dependence-forming products including tobacco, alcohol and illegal drugs and using mass media to change behaviour around these may be particularly challenging. The insights and suggestions raised by delegates have helped inform how we have approached key sections of this final report and have directly informed how we describe the context and implications of our findings.

Engagement with young people

Two members of the research team (Professor Linda Bauld and Dr Srinivasa Vittal Katikireddi) met with 15 young people on 29 September 2017 to discuss the findings of the study and the logic model that had been developed to inform the research. The meeting took place in a City of Edinburgh Council leisure facility: the Royal Commonwealth Pool, Edinburgh. The young people were attending a youth club that provides access to gym and pool facilities at the venue along with snacks, board games and art materials; it is held every Friday evening during term. The club is known as ‘Open All Hours’ and is funded by City of Edinburgh Council in partnership with Edinburgh Leisure. Research team attendance at the club was made possible by Laurene Edgar, a youth work organiser, whom Dr Katikireddi contacted after obtaining her contact details from a member of the study advisory group.

The researchers met with two separate groups of young people between the ages of 11 and 15 years. In the first group there were six boys and two girls, and in the second group there were seven girls. The researchers showed some examples of public health mass media campaigns on a television screen connected to a laptop. These ranged from television campaigns on physical activity, alcohol and chronic disease prevention from the early 1990s to slightly more recent NHS campaigns on smoking cessation and second-hand smoke from around 2005–7 and concluded with a current online video of the ‘This Girl Can’ physical activity campaign funded by the National Lottery. Young people had mixed views about the advertisements, perhaps preferring more contemporary examples, and just one or two of them recognised the ‘This Girl Can’ film.

The researchers then asked the young people to note down on paper what they thought constituted a ‘good’ or ‘appealing’ advertisement. Some of the main suggestions are included in Box 5.

With the first group (which spent slightly longer discussing the campaigns than did the second group) the researchers also distributed copies of the study logic model with the details of the model removed and just the headings showing. The researchers explained the purpose of the model and the young people then wrote down some of the elements they thought might connect a public health advertisement with short-term, medium-term and longer term outcomes.

They were able to describe in some detail the key ‘inputs’ needed for a mass media campaign, including funding, staffing, equipment, ideas and production. They were also adept at identifying key activities after inputs were secured, including where advertisements or other forms of marketing should be placed (television, social media, YouTube, etc.) and how the public and others should be informed about the release of a campaign. They were less sure about the pathway after that between people viewing an advertisement or other campaign output and longer term outcomes. They articulated very clearly what the ideal behavioural outcomes would be in terms of reducing harmful behaviours (stopping smoking and being less sedentary) or increasing healthy behaviours (eating a more balanced diet and being more physically active). They also understood that these behavioural outcomes would be linked to longer life or reducing the risk of chronic diseases such as cancer. However, unsurprisingly, the mechanisms for change between viewing a mass media campaign and possible behaviour change were much more difficult for them to describe.

Despite this, the concept of the logic model and how inputs and activities could be linked to longer term outcomes was something that appeared to resonate with them and supports the concept of the logic model approach. Their notes also illustrate how an unappealing or perhaps not well-sustained campaign could be ignored and might make little difference. They also talked about how young people could view a campaign (e.g. on second-hand smoke) and discuss behaviour change with parents or carers. One girl in the second group gave examples of a friend who had seen campaigns about the harms of second-hand smoke exposure and talked to both her parents about their smoking and smoking cessation.

The discussions with these young people were informative at the stage when the team was drafting the final report for the study and their views have been useful in assisting us to write the discussion section of the report, the further description of the logic model and our summary of findings.

Strengths

This review adds value to the current literature on mass media interventions by bringing together a large amount of evidence for a variety of health topics and enabled a comparison between them. It combined the breadth that is offered when looking across review-level evidence with the depth obtainable from examining individual primary studies. In response to frequent calls for in-depth analysis of how campaigns work (e.g. Cassidy et al. 179), it examined intervention characteristics that are associated with effectiveness. The review has particular relevance to the UK context, and we sought feedback from stakeholders to assess its usefulness.

Methodological limitations

We were unable to conduct statistical synthesis owing to the considerable heterogeneity across the studies. This makes it difficult to draw firm explanatory conclusions about the causes of the variability in results, noted by Ferri et al. 62 among others. In the review of reviews, some of the primary studies were published in the 1980s and 1990s and, thus, were discussing technologies that are no longer relevant. The searches for review A, the review of reviews, were conducted in January 2016. Searches for the subsequent reviews were conducted later (up to January 2017 for review C), reflecting the sequential nature of the project. Although offering a breadth that would have been unachievable if only primary studies had been examined, the focus on reviews meant that, at times, we lacked contextual and intervention details, and there may have been some overlap in studies between reviews. It was difficult to assess bias within the existing reviews because this would have entailed redoing their analyses. Inevitably, our results are limited by the quality of the primary studies, and reflect a publication bias in which weaker campaigns are rarely evaluated, and interventions with poor results are less likely to be written up and published. 166

As noted above, there are a number of issues involved in assessing the cost-effectiveness of mass media campaigns. Data are generally limited, with few studies; this is the case even regarding tobacco use, the health topic that tends to be most frequently examined in mass media studies and reviews. Part of the reason for the lack of studies is likely to be the challenges associated with extrapolating short-term effects (e.g. increase in quit attempts) to long-term costs and outcomes, which requires expertise in mathematical modelling. The evidence is also likely to be biased, with ineffective evaluations being unlikely to undergo economic evaluation, thereby running the risk that the intervention under examination is likely to look more cost-effective than it is on average. Finally, approaches to the synthesis of economic evidence are still being developed. Cost-effectiveness analyses, in particular, are very context specific, and it is challenging to conduct systematic reviews of such studies while maintaining global relevance.

Limitations in scope/definitions of our study

Although the study was wide-ranging, necessary parameters in terms of scope and how mass media campaigns were defined meant that there were inevitable gaps in the evidence we could review. We focused on six topics relating to preventable risk factors for disease: alcohol use, diet, illicit substance use, physical activity, sexual and reproductive health and smoking. The focus on disease prevention meant that campaigns addressing related behaviours but with a different focus were excluded: we did not include alcohol campaigns whose main focus was drink driving. Mass media campaigns seeking to raise awareness, counter stigma and encourage help-seeking behaviour in relation to mental health issues were excluded, as were road safety campaigns (e.g. targeting speeding or seat belt use) and campaigns encouraging skin cancer protection behaviours. Mass media campaigns whose primary aim was to encourage participation in screening programmes were excluded, although campaigns that sought to encourage screening in addition to more lasting behaviour change (such as campaigns that encouraged condom use and HIV/AIDS testing) were included because of the behaviour component. Because of our focus on population behaviour change, we excluded evidence relating to the use of the media in a media advocacy context, in which news media coverage and other forms of media messages are used to foster public and policy-maker support for policies or legislation to promote health, such as changes in taxation, drink driving laws or restrictions on marketing activity. 180–182 Previous research has suggested that media coverage achieved as part of a multifaceted advocacy campaign can be one of the factors leading to successful policy change, although the challenges of demonstrating the particular contribution of media to the outcomes are considerable. 183

We defined mass media campaigns as the intentional use of any media channel(s) of communication by local, regional and national organisations to influence lifestyle behaviour through largely passive or incidental exposure to media campaigns, rather than largely dependent on active help-seeking (adapted from Wakefield et al. 6 and Bala et al. 27). We adopted this definition of mass media as it is potentially the most useful for campaign planners seeking an estimate of the effectiveness of campaigns implemented in naturalistic settings that do not require active audience engagement with a given channel to achieve message exposure. 164 However, this meant that certain types of communications-based intervention were excluded. Interventions that require individuals actively to seek out the information (such as websites) or to opt in or sign up (such as SMS/text messaging for smokers trying to quit) would have been excluded because they require active engagement by target populations. Not including these interventions has limited what the review can say about new/digital/interactive media interventions. However, that missing evidence may be limited in itself. A fairly recent scoping review of how digital media (including visual, electronic and online media) are used in the area of public health184 found that current public health usage is predominantly the ‘outmoded approach of “telling and selling” and cast[s] the recipient in an individual and passive role’, based on 221 systematic reviews published between 2000 and 2013.

Large multifaceted community interventions that included a media element were included only if it was possible to relate effects to the media component of the campaign. For example, a review of interventions that involved a mass media campaign combined with health-related product distribution was included because the outcomes that related specifically to the mass media were reported and synthesised by the authors. 69 However, this meant that evidence from reviews such as the Cochrane Database Systematic Reviews of universal multicomponent prevention programmes for alcohol misuse and community interventions for preventing smoking76,185 and the ‘Change4Life Smart Swaps’ intervention study (Wrieden and Levy186) was not included in our reviews. We are therefore limited in what we can conclude about the extent to which mass media campaigns can interact with other interventions or services to improve health outcomes, which was one of our original objectives. However, our review of reviews found promising evidence regarding the ability of mass media campaigns to stimulate engagement with other services. We found that, overall, media campaigns can prompt calls to telephone quitlines for smoking cessation (although there is only moderate certainty in the strength of this finding). For helping to foster engagement with sexual health clinics or services, there was low-certainty evidence that campaigns can increase use of such services.

In examining characteristics of mass media campaigns that may be associated with effectiveness, we focused on campaign components that featured in the ‘Activities’ box of our logic model. We did not examine mass media campaign ‘Inputs’, which in our logic model included resources, staff, expertise/skills, technology and materials, although the full economic evaluation reviews in our rapid review of cost-effectiveness (review C) would have taken into account the upfront costs (resources) of mass media campaigns, when information was reported.

Search strategy

AB “systematic review” OR TI meta-analysis OR AB meta-analysis OR TX meta-analy* OR TI “systematic literature review” OR AB “systematic literature review” OR TI “meta-synthesis” OR AB “meta-synthesis” OR AB “integrative review” OR AB “integrative research review” OR AB “rapid review” OR TI “evidence based” OR AB “evidence based” OR TX “exclusion criteri*” OR research synthesis OR “narrative synthesis OR narrative review” OR “critical interpretive synthesis” OR “rapid review” OR “scoping review” OR “evidence synthesis” OR “research synthesis” OR “evidence review”

TI media OR AB media OR TI “mass communication” OR AB “mass communication” OR TI “social marketing” OR AB “social marketing” OR TI broadcast* OR AB broadcast* OR (MH “Communications Media”) OR (MH “Social Media”) OR “media” OR health campaigns