What are mass media interventions made of? Exploring the active content of interventions designed to increase HIV testing in gay men within a systematic review

Purpose Mass media HIV testing interventions are effective in increasing testing, but there has been no examination of their theory or behaviour change technique (BCT) content. Within a heterogeneous body of studies with weak evaluative designs and differing outcomes, we attempted to gain useful knowledge to shape future interventions. Methods Within a systematic review, following repeated requests to the authors of included studies for intervention materials, the Theory Coding Scheme, the Theoretical Domains Framework (TDF), and Behaviour Change Technique Taxonomy (BCTT) were used to extract data relating to active intervention content. Results Of 19 studies, five reported an explicit theoretical basis to their intervention. TDF analysis highlighted the key domains employed within the majority of interventions: ‘knowledge’, ‘social roles and identities’, and ‘beliefs about consequences’. BCT analysis showed three BCT groupings commonly reported within interventions: ‘Comparison of outcomes’, ‘Natural consequences’, and ‘Shaping knowledge’. Three individual BCTs formed the backbone of most interventions and can be considered ‘standard’ content: ‘Instructions on how to perform behaviour’; ‘Credible source’; and ‘Information about health consequences’. Conclusions This is the first study to examine and detail active intervention content in this field. It suggests future interventions should improve knowledge about testing, and use well‐branded and trusted sources that endorse testing. Future interventions should also provide clear information about the health benefits of testing. Our analysis also suggests that to improve levels of effectiveness characterizing the current field, it may be useful to elicit commitment, and action plans, relating to how to implement testing intentions. Statement of contribution What is already known on this subject? Interventions are urgently needed to increase HIV testing among men who have sex with men (MSM) and enable increased access to effective treatment for HIV infection. There is some evidence of the effectiveness of mass media interventions in increasing HIV testing among MSM. Nothing is known about the active components of existing mass media interventions targeting HIV testing. What does this study add? It describes the available literature concerning evaluated mass media interventions to increase HIV testing. It shows few interventions report any explicit theoretical basis although many interventions share common components, including coherently connected causal mechanisms and behaviour change techniques to moderate them. As a minimum, future interventions should improve knowledge about testing; use well‐branded and trusted sources that endorse testing; and provide clear information about the health benefits of testing. Our analysis also tentatively suggests it may be useful to elicit commitment and planning of how to implement testing intentions.


Statement of contribution
What is already known on this subject?
Interventions are urgently needed to increase HIV testing among men who have sex with men (MSM) and enable increased access to effective treatment for HIV infection. There is some evidence of the effectiveness of mass media interventions in increasing HIV testing among MSM.
Nothing is known about the active components of existing mass media interventions targeting HIV testing.
What does this study add? It describes the available literature concerning evaluated mass media interventions to increase HIV testing. It shows few interventions report any explicit theoretical basis although many interventions share common components, including coherently connected causal mechanisms and behaviour change techniques to moderate them. As a minimum, future interventions should improve knowledge about testing; use well-branded and trusted sources that endorse testing; and provide clear information about the health benefits of testing. Our analysis also tentatively suggests it may be useful to elicit commitment and planning of how to implement testing intentions.
Despite efforts to increase access to HIV testing services, it is estimated that worldwide around half of HIV-positive people remain undiagnosed (Conserve et al., 2017) and Public Health England estimated that 13% of HIV-positive men who have sex with men (MSM) were undiagnosed in 2016 . Undiagnosed infection is the primary source of onwards transmission (Skarbinski et al., 2015). The costs of HIV infection to society are difficult to assess, but estimated mean lifetime costs of treating one person are £360,800 with costs reducing to £179,000 if generic drugs replaced patented drugs (Nakagawa et al., 2012). Recent estimates suggest there are about 4,200 MSM who are currently living with undiagnosed HIV infection in the United Kingdom (Nash et al., 2018). HIV testing can be thought of as a gateway to a wider range of prevention approaches (Flowers, Estcourt, Sonnenberg, Burns, 2017). Within the United Kingdom, annual HIV testing is recommended for all MSM under existing guidelines, with more frequent testing (e.g., three monthly) recommended for men reporting two or more unprotected anal intercourse partners within the last year. Despite these recommendations, McDaid et al. (2016) report that when combining analyses from Scotland and England using both barbased and online samples, appropriate levels of HIV testing are not being met; for example, annual testing was reported by only half of these UK MSM (McDaid et al., 2016). New behaviour change interventions are urgently needed to increase levels of appropriate HIV testing in order to capitalize on the growing range of HIV prevention approaches such as pre-exposure prophylaxis -PrEP Flowers et al., 2017b;Witzel, Weatherburn, Rodger, Bourne, & Burns, 2017).
It is against this backdrop that the current study reports on one element of a wider intervention development study. Here, we report on work commissioned by a large NHS health board to conduct a review of the available literature concerning behaviour change interventions to increase HIV testing delivered through the mass media. This review work was complemented by further analyses and co-design projects engaging community members. Cochrane review-level evidence suggests the short-term effectiveness and the cost-effectiveness of mass media interventions to increase testing (Wei et al., 2011). Yet beyond this focus on effectiveness, there is a need to understand in more detail the active ingredients of such interventions in order to inform the development of future behaviour change interventions to increase appropriate HIV testing. A rich set of analytic tools can assist with the task of understanding the active ingredients of behaviour change interventions. Approaches such as the theory coding scheme, TCS (Michie & Prestwich, 2010), the Theoretical Domains Framework, TDF (TDFv2) (Atkins et al., 2017), and the Behaviour Change Techniques Taxonomy version 1 (BCTTv1) (Michie et al., 2013;Presseau et al., 2015) all share epistemological assumptions concerning the goal and value of developing cumulative knowledge to inform future interventions. These approaches, when used retrospectively within a systematic review, all seek to focus on what can be usefully learned from interventions that have been tested in the past. As such, their strengths lie in their systematic approach to specifying the shared elements of intervention content.
Many publicly funded real-world interventions are not designed with evaluation in mind, many are not adequately theorized in relation to how their content seeks to address intervention outcomes. In parallel where there is an evaluative framework, it is often plagued by weak research designs and the use of heterogeneous outcomes. These issues can reduce the potential of analytic tools such as the TDF and the BCTT and make the rigorous development of cumulative knowledge problematic. Within this paper, we explore what can usefully be learned from the application of these approaches within one field characterized by weak evaluative designs and heterogeneous outcomes.
The analysis presented here examines the content of evaluated HIV testing interventions in relation to (1) the role of explicit (i.e., TCS) and implicit theory (i.e., TDF) within interventions (in other words, the causal, or explanatory, mechanisms addressed by the interventions); and (2) the number and details of BCTs present within these interventions (in other words, intervention elements which moderate causal, or explanatory, mechanisms); (3) in addition, working within the constraints of the available studies and their diverse designs, we tentatively explore links between theoretical domains and BCTs across the interventions and their relationship to effectiveness.

Methods
The review Details of the systematic review are reported in McDaid et al. (2019); however, we have included a summary of the procedure undertaken for the review to provide a background to the current analysis.

Search strategy
Five electronic databases (CINAHL, Embase, Medline, PsycInfo, and Web of Science) were searched for studies published since 2009, using detailed search strategies and standard MESH terms for HIV, MSM, and mass media interventions (for details of an example search strategy see McDaid et al., 2019). In addition to database searches, reference lists of included articles were searched manually. A maximum of three requests were made to study authors for intervention materials.

Quality appraisal
No studies were excluded on the basis of quality. Details of the approach to quality appraisal can be found elsewhere .

Study selection
Studies published prior to 2010 and with intervention materials in languages other than English, Spanish, or Italian were excluded. Studies in which MSM constituted at least onethird of the sample included interventions that sought to change behaviour through noninteractive, visual or auditory means and included HIV testing as an outcome were included. The procedure for identifying relevant papers is shown in the PRISMA diagram in Figure 1.

Data extraction
Structured data extraction tools were developed to capture all required information from the included studies. Where intervention materials were made available, these were used to extract data of intervention content over and above the written intervention descriptions. Where no intervention materials were available intervention descriptions were used. Data extraction was completed by one researcher with a 10% sample validated by another. Discrepancies and disagreements were resolved through consensus or through discussion with wider team members.
Assessing theory within the interventions The current study uses two complementary yet distinct approaches to address theory present within the interventions: the TCS and the TDF.
In order to examine the role of explicit and formal theory within interventions, and in particular how formal theory was operationalized, the TCS was adopted (Michie & Prestwich, 2010). The current study used the first 11 items of the scheme to explore relationships between theory, target behaviours, and their role in information development and evaluation. The TCS appraises intervention content in relation to the extent it is theory driven; in other words, the way that formal theory suggests an array of theorydriven mechanisms and hypotheses. In contrast, the TDF (Cane, O'Connor, & Michie, 2012), with its roots in implementation science, takes a more inductive approach to understanding the role of theory. As a meta-theoretical perspective, the TDF can be used, post hoc, to code how the intervention was imagined working. In this way, it can illuminate the tacit use of theory; for example, if information provision was at the heart of what the intervention was aiming to do, there is a clear fit with the TDF domain knowledge. The TDF has a total span of 14 domains, each constituted from many more related theoretical constructs. The TDF domains provide a shared or common language in which to understand and describe the key causal mechanisms determining behaviour change across the full range of behavioural domains (be they complex behavioural domains such as the multifarious aspects of implementation or relatively simpler behavioural domains such as HIV testing). In this way, we used the TDF to conduct a retrospective assessment of the key causal mechanisms which we believed were being addressed within the interventions.

Behaviour change taxonomy v1
Intervention descriptions were also coded using the BCTTv1 as proposed by Michie et al. (2013). Intervention materials or intervention description (e.g., posters) were assessed for each of the 93 BCTs within the BCTTv1. Some elements of the intervention descriptions could be coded to multiple BCTs. In this way, we used the BCTT to code intervention elements moderating putative causal mechanisms underpinning the interventions.
Exploring patterns of intervention content and relative effectiveness Given the heterogeneity of the interventions and their primary outcomes, it was not possible to conduct meta-analytic approaches. Here, we categorize the studies to tentatively explore the patterning of theory and BCTs according to effectiveness. Effectiveness was defined by statistically significant results.
1. Intervention had a negative effect (i.e., decrease in uptake of HIV testing). 2. No evidence of a positive/negative effect. 3. Evidence of a positive effect in relation to the antecedents of HIV testing (e.g., knowledge of HIV testing increased). 4. Evidence of a positive effect on HIV testing itself.
Analysis TDF domains, individual and grouped BCTs were calculated across the included interventions to enable the description of 'standard' content, or in other words, typical, or regularly occurring content. These same elements were then mapped against the categories of effectiveness to enable the identification of unique components present only within the most effective interventions.

Results
Nineteen articles were identified from the review, focusing on 22 separate interventions. Following repeated requests to authors, we were able to locate intervention materials for a total of 14 interventions. Table 1 presents a summary of the intervention characteristics. Interventions were all mass media interventions, incorporating combinations of photographs, video footage, and cartoon animation. They encompassed posters, radio, TV, YouTube, and other materials.
The role of explicit and implicit theory within the interventions Using the TCS, our analysis found that only five of the included studies explicitly mentioned the use of theory, either in the development or in the delivery of the intervention (see Table 2).
Duplicate cases Five study conditions: (1) dramatic video; (2) documentary video; (3)  In relation to the TDF and the implicit use of theory, the interventions included in this study worked to increase HIV testing through focusing upon causal mechanisms such as the provision of 'knowledge' (n = 18), utilizing social roles and identities (n = 15), sexual or ethnic identities, and by focusing recipients' 'beliefs about the consequences' of their HIV testing-related behaviours (n = 14). Intervention content also sought to change behaviour through changing the 'environment context and resources' often by using particular spaces to prime behaviour change (n = 13). Finally, intervention content also worked through 'social influence' (n = 12) using mechanisms such as social identification or social norms around testing to cue behaviour change.
Behaviour change techniques Our analysis examined BCTs identified within interventions in terms of both BCT groupings (n = 16) and individual BCTs (n = 93).
In relation to BCT groups, within the included studies, 13 distinct groups of BCTs were employed, with the three most common groupings being 'Comparison of outcomes', 'Natural consequences' and 'Shaping knowledge' (see Table 2). In terms of individual BCTs, our findings identified 30 distinct BCTs were used across the interventions. We have grouped these BCTs into three main categories relating to frequency of use within the interventions: those that are very common in most interventions (n = 10-30), those that are fairly common (n = 5-10), and those that are rarely used (n < 4).
Our findings indicate that three 'very common' BCTs that form the backbone of most of the included interventions and could therefore be considered 'standard' content: 'Instructions on how to perform behaviour' (n = 19 studies), 'Credible source' (n = 19), and 'Information about health consequences' (n = 15).
The group of 'Fairly common' BCTs were often linked to each other and to those very commonly used (see above). They could be considered as reinforcing this 'standard content'. For example, 'Information about social and environmental consequences' (n = 7) clearly relates to the more prevalent technique of 'Information about health consequences' and encourages the recipient to consider the wider consequences of HIV testing for self, others, and the wider community beyond the consequences associated with health. Other fairly common BCTs which could be linked to contributing to the decision-making process regarding adopting, or not, the desired behaviour (HIV testing) were also identified within the contributing interventions. 'Social comparison' (n = 7), 'Pros and cons' (n = 5), and 'Vicarious consequences' (n = 5) encourage the recipient to consider their behaviours either in relation to perceived or observed consequences to self and/or others and task recipients with a decisional balance or reckoning of the reasons to test for HIV. Interventions employing video narratives were most likely to adopt the technique of 'Vicarious consequences' by depicting the consequences of testing and/or risk behaviour within a social setting. Similarly, the implementation of the BCT 'Social comparison' (n = 6) was often visual and sometimes nuanced, relying on the recipients to identify with the images or characters stories within the interventions. Other fairly common techniques adopted, related to the provision of resources, either through 'social support (unspecified)' (n = 5) and related technique 'social support (emotional)' (n = 4) or through 'adding objects to the environment' (n = 7). For example, several studies identified the use of approaches such as credit card-shaped leaflets that could be kept within the wallet. This technique linked to the use of 'Prompts and Cues' (n = 6), where materials were often situated in locations such as venues where testing could take place in order to prompt or cue testing.
In terms of the 'rarely used' BCTs, our findings suggest that interventions tended to employ BCTs that could be described as relatively passive and designed for delivery through one-off implementation.
We also identified key absences within the intervention content in that we found that BCTs relating to the groups 'feedback and monitoring' and 'goals and planning' appear to be under-used or absent within the included interventions. This potentially reflects the constraints of modes of delivery used within these mass media interventions.
Exploring patterns of intervention content and relative effectiveness Elsewhere, we detail the results of our analysis of the effectiveness of these interventions . The majority of studies (n = 12/19) reported results that were indicative of positive change in relation to HIV testing, whilst only five of the studies included indicated that the intervention had no effect.
In relation to explicit use of theory, and the degree to which we can say these interventions were theory driven, given the small numbers of studies explicitly discussing theory, we were unable to examine relative patterns of effectiveness in any meaningful way by the results of the TCS analysis.
In relation to implicit use of theory, or the key causal mechanisms that underpinned the included interventions, the TDF domains 'beliefs about capabilities', 'intentions', 'goals', and 'behavioural regulation' were only observed in interventions that showed evidence of a positive effect on HIV testing. Many of the more widely used TDF domains were present across at least two of the four categories of intervention effectiveness.
Across all categories of relative effectiveness, most of the groupings of BCTs (12/16) were observed. Interesting distributions were also observed in terms of patterning of individual BCTs across the interventions when they were grouped in relation to our four categories of intervention effectiveness. Single occurrences of 'Problem Solving' and 'Salience of consequences' were only observed in interventions showing no evidence of a positive, or negative, effect. In contrast, individual BCTs that featured within interventions showing evidence of a positive effect on HIV testing included one-off instances of 'Action planning', 'Commitment', 'Information about others approval', 'Material incentive (behaviour)', 'Information Material reward (behaviour)', 'Incentive (outcome)', 'Reward (outcome)', 'Reducing negative emotions'. Equally, 'Information about emotional consequences' and 'Restructuring the social environment' were observed within the effective interventions more often (three times each).
In relation to the BCTs, on average interventions reporting no effect were coded with a larger number of BCTs than those reporting a positive effect. However, some of these ineffective interventions were delivered by video, and therefore, we would suggest that the number of BCTs employed may be a reflection of intervention complexity and interactions with form of delivery rather than illuminating effectiveness per se.

Discussion
The current study has, for the first time, described the typical active content of previously evaluated international interventions designed to change HIV testing behaviour among MSM through the use of mass media. Moreover, within the constraints of heterogeneous study designs and diverse study outcomes, we have also tentatively explored patterns of effectiveness related to this active intervention content. These results represent a step forwards in directing future HIV testing intervention design for this population as they lever a level of hitherto under-specified detail regarding potentially useful intervention content. Our analysis specifies both the causal mechanisms, and the BCTs, used in many of these, largely effective, interventions. Our analysis also tentatively suggests the potential of further BCTs that may be uniquely associated with effectiveness. In this way, the analysis progresses the field but also demands further work for intervention development.
Across the contributing studies, our use of the TCS showed that interventions in this field are rarely theory driven; it was difficult to see coherent connections between these explicit formal theories and the active intervention content we identified. In other words, there is a disconnect between formal theory when it is mentioned, and the actual active elements of the interventions identified. It may be that formal theory is conceptualized in relation to the rather generic behaviour change goals of the interventions, rather than assisting with specifying particular content, which is tailored to the known barriers and facilitators of HIV testing behaviour. Alternatively, it may be that theory is added post hoc in relation to the discussion of the findings.
Our use of the TDF highlighted a strong sense of the key causal mechanisms which these interventions address. Whilst the links between the theoretical constructs reflected within formal theory and intervention content could be described as opaque, the TDF domains which these interventions address are logical, plausible, and work well with the medium of mass media. The interventions identified within the review focus on knowledge provision; speak directly to social roles and identities primarily in relation to gay men and gay cultures; have a clear focus upon targeting and attempting to change intervention recipients' beliefs about the consequences of HIV testing; address the environmental and contextual determinants of testing within the gay milieu; and capitalize on social influence particularly in relation to the social dynamics between gay men. In this way, these interventions collectively have a logical and coherent approach to addressing plausible causal mechanisms which may change HIV testing behaviour among MSM.
In relation to the BCT groups we identified, the three most common groupings were 'Comparison of outcomes', 'Natural consequences', and 'Shaping knowledge'. 'Comparison of outcomes' and 'Natural consequences' are similar in that both BCT groups encourage the recipient to consider the consequences of behaviour change; the first elicits a weighing up of options (e.g., pros and cons of testing or not), whilst the second focuses the participant to think through the consequences of testing, or not. 'Shaping knowledge' focuses on BCTs related to providing information and instruction about how to perform the behaviour (e.g., where to access HIV testing). Again, there is a clear sense of these groups of BCTs being logical and plausible intervention elements that may be useful in changing testing behaviour. Our analysis of individual BCTs showed very commonly used BCTs included; 'Instructions on how to perform behaviour', 'Credible source' and 'Information about health consequences'. Again, these aspects of active intervention content are logical and work well with the mass media format of intervention delivery.
Looking across the intervention content we identified, there is a clear sense of congruence between the tacit theory identified using the TDF (i.e., the causal mechanisms the interventions addressed) and the BCTs we could identify from the best available intervention descriptions and materials (i.e., the intervention content intended to change behaviour). In this way, there was good evidence of the underlying logic, or what could be described as a tacit theory of change, that underpinned many of these interventions. This logic tracked an arc connecting the likely determinants of behaviour (i.e., TDF domains) with congruent intervention elements that were logically tailored to them (i.e., BCT groupings and individual BCTs). For example, the theoretical domain of knowledge clearly relates to the BCT grouping 'Shaping knowledge' and concomitant BCTs 'information about health consequences' and 'instructions on how to perform a behaviour'. Equally, the TDF domain 'beliefs about consequences' relates well to both BCT groupings, 'Comparison of outcomes' and 'Natural Consequences'. In this way, the findings can be interpreted to suggest that BCTs that provide factual information from a trusted source and that also demonstrate support for desired behaviour change (i.e., HIV testing) and foster the intervention recipients understanding of the consequences of behaviour may be particularly common building blocks within mass media interventions to increase HIV testing. Our findings suggest that the underlying logic of these approaches is concerned with embedded BCTs that enable testing decisions based on credible information about the relative benefits of enactable behaviour change. In relation to the pragmatic aims of the project and our funders desire to develop an evidence-based mass media HIV testing intervention, our analysis of these published evaluations highlights the backbone of these largely effective interventions as being concerned, in the main, with a series of coherent and logically connected intervention elements. We believe these represent foundational elements for future interventions likely to yield similar degrees of effectiveness as those that contributed to the current review.
Our analysis of the patterning of BCTs across our categories of relative effectiveness is also interesting although limited given the heterogeneity of contributing studies and their outcome measures. Although it is not based on statistical analysis, it does suggest very specific ideas to take forwards into future intervention development work. It suggests TDF domains such as 'beliefs about capabilities', 'intentions', 'goals', and 'behavioural regulation' may be useful in improving intervention effectiveness beyond levels of effectiveness anticipated by the standard content outlined above. It also suggests that BCTs such as 'Action planning' or 'Commitment' may be worthy of further detailed exploration. These BCTs are interesting as they include a temporal element inviting intervention recipients to consider their behaviour in the future. It is worth noting that the use of these BCTs may be constrained by the mode of delivery within the mass media. It may be useful to consider how using the digital media may enable more interactive and sequential engagement with mass media interventions. Such sequential engagement may help with the formation of plans to implement intentions to test. Other BCTs worthy of further exploration include those which address the social and emotional aspects of HIV testing, these include 'Information about others approval', 'Reducing negative emotions', 'Information about emotional consequences', and 'Restructuring the social environment'. These BCTs are interesting as they acknowledge the non-health-related aspects of HIV testing and the immediate social context of HIV echoing many of the key local barriers to HIV testing .
Implications for practice, policy, and further research This study represents the first attempt to describe and interpret the active content of mass media interventions intended to increase HIV testing in gay men and other MSM. It provides hitherto unarticulated detail regarding the standard content of such interventions and makes granular suggestions of potentially useful future intervention content. It is useful as it provides a benchmark, albeit a snapshot, of the active content of recent mass media interventions to change HIV testing behaviour among MSM.

Limitations
Our analysis is limited by the legacy effects of the contributing studies. The included papers are mainly from the developed world and do not address the full range of social and cultural contexts inhabited by MSM who may benefit from increased HIV testing. We caution against the straightforward generalization of the findings reported here to MSM populations who do not receive the benefits of the social equality they deserve (e.g., in countries in which homosexuality is criminalized such as Jamaica, Saudi Arabia, or Uganda).
Our analysis is also limited by the range of materials describing the intervention content that were made available to the team. We systematically requested such materials three times from the authors of the included studies; 14 sets of intervention descriptions were eventually received. Intervention materials tended to provide richer accounts of the intervention content than that provided within published descriptions of interventions included within the primary evaluation papers. As such although we were consistent in using the best available intervention descriptions, it is important to note that poor reporting of intervention content within primary papers cautions against any simplistic interpretation of the findings reported here.
Furthermore, it is important to acknowledge that our synthesis and assessment of the patterns of intervention content and their relative effectiveness is not statistical. We have tried to glean useful information from the available evidence, but this does not relate to the specification of effect sizes or statistical insights into relative contribution of specific intervention elements to overall intervention effectiveness.
A further key limitation relates to our lack of appraisal of the cultural, technological, and temporal contexts within which these interventions were delivered. Whilst this was beyond the scope of the current paper, it begs questions to how effectiveness, and the role of specific active intervention content, may well relate to complex interactions between the intervention and the context it was delivered in. These limitations relate to wider weaknesses of the TDF and BCW approaches and stem from their tendency to deracinate interventions, and intervention content, from the key contexts in which they were implemented. Similarly, whilst the high degree of granularity associated with these approaches enables the specification of highly specific accounts of intervention content, it does not lend itself to understanding how diverse intervention components can work either synergistically, or inharmoniously. This particular point is relevant to considering how to interpret the specific BCT content of interventions, where interdependencies between BCTs may well be important in understanding effectiveness.
The team have already addressed some of these limitations with additional analyses (Flowers et al., in preparation), in which the findings reported within the current paper were synthesized with other diverse sources of evidence. These additional sources of evidence included: (1) a TDF and Behaviour change wheel (Michie, Atkins, & West, 2014) analysis of the perspectives from local gay men and other MSM about the immediate context and culture, in order to specify culturally appropriate intervention elements tailored to local men, (2) a synthesis, and subsequent analysis using the TDF and behaviour change wheel (Michie et al., 2014) of the international literature concerning barriers and facilitators to HIV testing to specify theory-informed, and evidence-based, potentially useful intervention content to complement the suggestions identified within the current study.

Conclusions
Our approach, detailing active intervention content in relation to TDF domains and BCTS, has identified key intervention elements that are likely to be useful to develop future mass media interventions to promote appropriate HIV testing. 'Standard' content was identified in most interventions: improving knowledge about testing, well branded and trusted sources endorsing the message and clear information about the health benefits of testing. However, it is also important to consider the use of BCTs that may boost or reinforce this 'standard content', for example, extending recipients understanding of the consequences of testing to include sexual partners and communities or eliciting commitment and planning of how to implement testing intentions. Further research that examines these issues within the local context and population is recommended to complement the results described here.