Key findings and recommendations

The STASH study intervention was feasible to implement in six Scottish schools with varying levels of deprivation and educational attainment. It was highly acceptable to the students, staff and trainers involved. This chapter presents key findings against the study objectives.

Key findings against study objectives

Objective 1

Finalise the design of a school-based STI prevention intervention, in which influential PSs use online social networks and face-to-face interactions to influence norms, knowledge, competence and behaviour, and promote use of sexual health services.

We met the first objective via reviews of literature and sex education resources; intensive consultation with young people and professional stakeholders (patient and public involvement); co-production and optimisation of a website of curated resources and a training manual; and an intervention pilot. Intervention development work identified key motivations for schools in terms of meeting national targets for health and well-being, and providing leadership opportunities for senior students. Our consultations and pilot affirmed that the peer nomination process used in ASSIST would be palatable to students and staff, but that we might need to overinvite students (25% of the year group rather than 17.5%) to the recruitment meeting, to train 15% of the year group as PSs (the proportion deemed critical in ASSIST).67

In response to discussions with stakeholders regarding the sensitivity of the topic, we developed additional nomination questions focused on trust and listening skills to pilot alongside the original ASSIST questions (focused on respect and leadership). Most development activities contributed to the design of a curated set of shareable digital resources (housed on the STASH study website) and a bespoke training manual. Over various iterations, the STASH study website became increasingly visual and decreasingly reliant on text and signposting to other websites (as students in the pilot rarely clicked on these links).

The PS role crystallised incrementally in response to ideas from young people and stakeholders, as well as learning from the pilot. The role was eventually clarified as information giving, attitude influencing, promoting the website and signposting to sexual health services. Professional stakeholders were clear that the PSs would require training on handling sensitive disclosures and that they should be advised to avoid a counselling role. A PS charter was created to signal to PSs the responsibility inherent in the role.

The STASH study sought to adapt to contemporary modes of social interaction by introducing a social media option for dissemination of messages. A review of current platforms identified Facebook as the only one to meet the project requirements of closed and private groups and ability to interface with a website. Consultations with young people established that Facebook was not a first choice of platform for young people, but nevertheless it was widely used and broadly acceptable. A key design feature, initiated in response to concerns of teachers, was to stipulate that the STASH study trainers would monitor and support online activities via membership of each PS Facebook group. Development activities, and particularly the pilot, highlighted the need to increase awareness of the STASH study across school staff and year group students, but suggested a flexible approach, allowing each PS team to decide a course of action appropriate to their school.

Objective 2

Assess the recruitment and retention of PSs, as well as feasibility and acceptability of the intervention among PSs, participants and key stakeholders.

We anticipated that role uptake would be lower than in ASSIST57 and the ASSIST extensions128 because we were delivering the intervention on a more sensitive topic to a year group undertaking public examinations. Across the six schools, 52% of nominated students (63% of those attending the recruitment meeting) were trained. Variation across schools (a high of 77% in school 5 and a low of 33% in school 6) was consistent with process evaluation findings of variable levels of school buy-in and support of the intervention, as well as variation in the quality of the ‘pitch’ given by trainers at the recruit meeting. Role retention was very high, with 97% of PSs fulfilling the role (defined as posting three or more STASH study messages on Facebook or having two or more conversations and attending two or more follow-up sessions). The intervention was highly acceptable to PSs (85% said they liked the role), more so to young women (94%) than to young men (71%).

The majority of PSs liked the STASH study website and found it easy to navigate and share messages. The STASH study appeared to achieve a good balance between allowing flexible engagement and ensuring trainer monitoring and support.

Over 70% of PSs said that they found the STASH study more relevant, helpful and interesting than their usual SRE lessons. Of the other students engaging with STASH study messages via the PSs (exposed students), 74% thought that the way the STASH study was run was acceptable and 78% thought that the information provided was acceptable. Again, acceptability was higher for young women than for young men. We did not interview parents directly, but only one student reported that their parent was unhappy about them being a PS.

The eight staff members interviewed were positive about the ethos of the STASH study, the leadership opportunities it afforded, the rapport between trainers and PSs, and the high-quality training. Staff acknowledged the additional workload involved in delivering the STASH study intervention and the challenges for students in terms of missing classes for training and follow-up sessions. One teacher raised concern about school capacity to deliver the intervention. His concerns related as much to the burden of the evaluation as the intervention, and partly reflected that the school was undergoing a major transition at the time.

Objective 3

Assess the fidelity and reach of intervention delivery by trainers and PSs, including barriers to, and facilitators of, successful implementation.

The STASH study was implemented with excellent fidelity. All nomination exercises, recruitment sessions, external trainings and 35 of 36 follow-up sessions were delivered as planned. Attendance at follow-up sessions was high (average 80%), particularly given that this was an examination year. The PSs were highly active on social media, sharing 1343 messages from the STASH study website to their Facebook groups over the course of the intervention period (5 weeks in three schools and 10 weeks in the other three schools). On average, PSs reported six face-to-face conversations about the STASH study with their friends.

Social network analysis showed that students trained as PSs were well positioned across their year group friendship networks. They had greater potential to reach across their year group than students who were not nominated (5.3 incoming friendship ties vs. 3.5 for non-nominated students). PSs were also well distributed across friendship clusters, ranging from presence in one-third of friendship groups in one school to four-fifths in another. Across all schools, PSs had direct friendship ties to one-third of students.

The intervention had good reach across the S4 year group, with 58% (n = 611) of students at follow-up reporting some level of exposure (trained as a PS, visited the website or member of a Facebook group, or had a conversation with a PS). Young women were more likely to report exposure than young men (63% vs. 53%, respectively). Educational attainment and sexual experience at baseline both appeared linked to likelihood of exposure, but not socioeconomic deprivation, ethnicity, religiosity or sexual orientation. Key facilitators identified were buy-in and support from schools; visibility of the STASH study intervention to the rest of the year group and school; good rapport at training; high-quality training; regular direct communication between trainers and PSs via Facebook, including reminders for follow-up meetings; the use of private and closed groups to give PSs control over message sharing; and well-planned and regularly spaced follow-up sessions that included an interactive activity, as well as progress check.

Key barriers identified were reliance on schools for scheduling of activities; academic pressures, such that follow-ups scheduled during class time were difficult for some teachers and students; technical issues with the website (these were small in number); infrequent use of Facebook among some students; low levels of awareness of the intervention across the rest of the school; low levels of intrinsic motivation of non-PSs to engage with PS activities; and poorer engagement by young men compared with young women, probably due to higher embarrassment and less maturity in discussing sexual health issues.

Objective 4

Refine and test the programme theory and theoretical basis of the intervention.

We used the feasibility trial to test the STASH study programme theory, leading to a refined theory in which mechanisms of change, conditions of success and context were more clearly articulated (Figure 17). We also explored ways to visualise the theory to reflect the dynamic interaction of intervention components in a format accessible to stakeholders and delivery partners (Figure 18). Figure 17 shows the version of the programme theory following the feasibility trial, but prior to the full analysis of the results. Figure 17(a) shows the full logic model from problem identification to outcomes and Figure 17(b) ‘explodes’ the intervention component to detail the conditions of successful implementation and mechanisms of action. In comparison to the post-pilot, pre-feasibility trial programme theory, the version showing in Figure 17 has a number of key refinements. It more clearly sets out the theoretical underpinnings of the trial at micro, mid and macro level. It articulates the mechanisms of change for key actors in the intervention and it more clearly establishes the conditions of successful implementation, linking these to their respective intervention components. In addition, we refined our expectations of the key intermediate and end outcomes. We added competence at last sex (equally willing, no regret, no alcohol) and ease of talking about sex as intermediate outcomes, we moved relationship well-being to the end outcomes, and we added confidence and well-being among PSs as an end outcome. These refinements to the end outcomes emerged from discussion – with the TMG, TSC and project stakeholders – about the importance of recognising the wider potential impact of the trial beyond STI prevention. Although we captured qualitatively the broader psychosocial benefits for PSs, our measures of well-being (at least in this small sample) did not pick this up. Similarly, our new measure of relationship well-being was unable to detect a difference between intervention and control groups (see Objective 6).

FIGURE 17 The STASH study programme theory: feasibility trial stage. (a) Full programme theory; and (b) details of conditions of successful implementation and mechanisms of action. TDF, theoretical domains framework. Reproduced from Forsyth et al.62 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original figure. (Figure not reproduced here; see the original article.)

FIGURE 18 An alternative visualisation of the programme theory designed primarily to explain the STASH study to stakeholders (using an animation to highlight different stages). The visualisation highlights the conditions and mechanisms relevant to each intervention stage. (Figure not reproduced here; see the original article.)

Our analysis of the feasibility trial data largely confirmed that the programme theory presented in Figure 17 adequately captured the preconditions of success and contextual factors influencing effectiveness. In general, the key conditions were met, with variability across schools. Some conditions were harder to meet than others.

The feasibility trial was able to interrogate the mechanisms of change to some extent (see Figure 17). Table 18 shows the mechanisms in a separate box, alongside key evidence from the study relating to that mechanism and suggested revisions to the hypothesised mechanisms for a future version of the programme theory. The more challenging mechanisms were those dependent on PS delivery of activities (the aspect of the intervention over which we had least control). In particular, there appeared a limit to which the PS felt socially validated in the role by their peers (they appeared neither validated nor disproved of). A key finding was that there appeared a limit to which non-PSs were interested in messages and saw them as relevant. Although the training led to enthusiasm and awareness of sexual health issues among PSs, it appeared difficult for them to generate the same level of interest among their friends via brief messages and conversations. As these conditions were only partially met, the extent to which the intervention supported autonomy and intrinsic motivation (a purported mechanism of change) is unclear.

Diffusion of innovations theorists suggest five stages in the adoption of new behaviours (knowledge, persuasion, decision, trial and adoption).129 Owing to variation in the pace at which people move through the stages of adoption, considerable time can elapse between first and last adopters, even for clearly advantageous behaviours.24 The theory suggests that the rate at which a new idea or behaviour is adopted depends, among other things, on the past experiences and needs of potential adopters.59,129 A challenge for sexual risk reduction messages in mid-adolescence is that only a proportion of students are sexually active [approximately 20% reported vaginal or oral sex in this study (see Table 11)]. The ‘potential adopters’ of new messages on sexual behaviour may be relatively small (particularly as some of those who are sexually active are already practising safer-sex behaviours). The STASH study messages were wide-ranging in recognition of this variation among the target group; even so, students who were not sexually active may have simply decided that the STASH study was less personally relevant to them.

Given that sexual experience and ‘readiness’ for messages about sexual health will always be variable within a S4 student year group, it is worth reinterrogating our programme theory to ask whether or not supporting increased autonomy and intrinsic motivations is (1) realistic and (2) an essential mechanism of change. The architects of self-determination theory, Ryan and Deci,83 describe how social and environmental factors can facilitate or undermine intrinsic motivation (cognitive evaluation theory,83 a subtheory within self-determination theory). In the STASH study, PSs were able to create their own groups, and to choose which messages to share, how and with whom, and non-PSs had full autonomy over if and how they engaged with messages. The practice and ethos of the STASH study was thus autonomy supportive, particularly in comparison with more controlled didactic classroom-based methods. What seems less clear is whether or not this autonomy led to intrinsic motivation. Interestingly, the top two motivations cited for PSs continuing in the project were ‘learning information’ and ‘fun’ (see Chapter 5, Motivation to participate), and not extrinsic rewards, such as benefit to CV, indicating some level of intrinsic motivation among the PS group. Among non-PSs, intrinsic motivation to engage with the topic appeared low, but it is worth questioning whether or not this is a realistic expectation across the year group, and whether or not it is essential as a mechanism of change. It is possible that norms can be diffused across a peer network with relatively low levels of intrinsic motivation to engage with the intervention. Therefore, in the next iteration of the programme theory, increased intrinsic motivation should perhaps be limited to PSs and not the year group as a whole. In Recommendations for future evaluation, we also suggest ways to enhance non-PS engagement in the intervention.

Objective 5

Enhance understanding of the potential of social media, when used by influential peers, to diffuse norm change and facilitate social support for healthy sexual behaviour.

The students interviewed in the STASH study engaged with social media for primarily social reasons, using a range of platforms. Social ‘gossip’ unfolded seamlessly across online and offline spaces, such that not keeping up online could mean missing key developments in social events. Non-PS student reactions to receiving STASH study posts via Facebook ranged from openness and interest to indifference. Some saw the opportunity to engage online as less awkward, whereas others preferred face-to-face conversation or a mix. To the extent that non-PSs registered and responded positively to messages, it appeared important that messages were sent by someone they knew and seen by others in the group, such that there were opportunities for discussion. In other words, the offline social context was key in both legitimising and augmenting online messages.

The process evaluation confirmed findings of the intervention development stage in suggesting a strong preference for humorous, colourful, eye-catching images, with as little text as possible. The opportunity to share messages within a private group rather than on a general profile page was key to feeling comfortable, and PSs were discerning about who they invited. Issues of stigma and unease at being linked to sexual health content, identified in previous social media interventions in sexual health,130,131 did not emerge as particular issues in the STASH study. This is most likely as a result of the use of private groups. A possible cost of this privacy is that STASH study interactions may have been viewed as parallel to, rather than embedded in, PSs’ usual online social interactions.

The presence of a trainer within the private groups appeared reassuring for many, although only a small number felt the need or wished to engage with them in this forum (primarily young women). In terms of practical considerations, the use of social media necessitates hiring trainers who are comfortable using it, and there are challenges in establishing a design that is sufficiently flexible to keep pace with constantly evolving platforms and technology.

Objective 6

Determine key trial design parameters for a possible future large-scale trial, including recruitment and retention rates, strategies, outcome measures, ICC and sample size.

All six schools were retained in the study. Response rates for the student surveys were 80% for the controls, and 82% at baseline and 79% at follow-up. Among the intervention year, 81.2% of students completing the baseline questionnaire also completed the follow-up questionnaire. Response rates were slightly lower than other recent school-based studies.128,132 As parental opt-out (n = 6) and student non-consent (n = 3) were low, the response rate primarily reflected the number of students who attended school on survey days, and made it to the class in which the questionnaire was administered. There was a suggestion from teachers that the class lists (on which the denominator was based) were not always up to date and included students who did not regularly attend school. Our finding that students who did not complete the follow-up questionnaire were more likely to be male, from more deprived backgrounds and with lower educational entertainment highlights both the importance of conducting the STASH study in schools (where the fullest range of young people can be reached) and the methodological challenges of doing so.

We investigated a number of candidate primary outcomes, identifying a composite measure of ‘always practising safer sex’ as a potential primary outcome for the trial. We defined this as ‘No condomless (vaginal or anal) sex in the last 6 months’. A future intervention would seek to increase the proportion of students practising safer sex either by not having sex (the safest option) or by having sex but always using a condom. The odds of reporting always safer sex in this study were in a direction suggestive of benefit, although the study was not powered to demonstrate a significant difference. Like all self-report measures of sexual behaviour, this outcome is subject to social desirability bias and, to a lesser extent, recall bias. A further limitation is that it does not include unprotected oral sex and also does not account for situations in which two young people in a long-term monogamous relationship may have stopped using condoms following negative STI tests. These behaviours will be measured in the outcome questionnaire.

Delayed initiation or no sex and condom use are the most common outcomes in youth-targeted sex education interventions,3,31 but these are usually reported separately. Some studies use a measure of consistent condom use, whereas others focus on condom use at last sex.3,31 For instance, the school-based ‘Safer Choices’ intervention had three separate primary outcomes: (1) delayed initiation of sexual intercourse, (2) frequency of sex with a condom among those reporting intercourse and (3) the number of sexual partners with whom students had condomless intercourse in the last 3 months (among those reporting intercourse).133 Our composite outcome captures the first and second of these dimensions within a single measure. In a subsequent evaluation, we would not propose to measure the third dimension, because the rate of partner change within our time frame for follow-up is too low (an average of one partner) (see Chapter 7, Descriptive statistics for primary outcomes). Reviews of condom use measures in evaluating human immunodeficiency virus prevention show substantial heterogeneity in measurement, with no trend towards standardisation over time.134 Condom use at last sex is the most common measure, followed by consistency of condom use; however, current recommendations do not suggest clear superiority of one over the other.134 With respect to youth-targeted interventions, a challenge is that behavioural outcome measures can present an inconsistent picture. For instance, the STAND study found an increase in condom use at last intercourse, but negative changes in other outcomes (e.g. condom use in anal sex, unplanned pregnancy and STIs).41

In terms of secondary outcomes, our question banks on knowledge and attitudes had high response rates. Reliability was good for the knowledge question bank (Cronbach alpha 0.7) and medium or low response rates for the question bank on attitudes (Cronbach alpha 0.52). Knowledge was higher, and attitudes more positive, in the intervention responses than in the control, suggesting potential benefit of the intervention (again, not powered for significance). The question bank on confidence had good response rates and reliability (Cronbach alpha 0.77); however, we found an inconsistent picture (control year higher on some items and follow-up on others), possibly reflecting that individual assessment may vary according to experience of trying a behaviour (such as negotiating condom use). If taken forward to a future trial, these measures should be limited to those with sexual experience. Our newly designed measure of relationship well-being had high reliability (Cronbach alpha 0.79) for the positive subscale in particular (measuring openness, happiness, comfort and respect) and correlated with related constructs (e.g. distress about sex life), but did not differentiate between control and intervention groups. This needs to be considered in the decision about whether or not to take this forward. Our adapted measure of sexual competence99 (willingness, regret, use of alcohol or drugs) has medium reliability (Cronbach alpha 0.66) and merits inclusion, but condom use should be added to the measure (as a key dimension of competence). The measure should also be asked of most recent sex, rather than first sex, to increase the number of students for whom the question is relevant.

Our analysis of potential effect modifiers suggested several that might affect the primary outcome and should be retained: FSM, ethnicity, educational attainment, religiosity, peer group sexual behaviour and alcohol use. Other modifiers did not demonstrate an association with the primary outcome and could possibly be omitted from the questionnaire going forward. These include self-regulation, parental monitoring, sexual attraction and importance of social media to social life.

Although a study objective, we ultimately opted not to propose a sample size calculation for a definitive trial at this stage. This is partly because the estimate of effect from this small sample is imprecise and yet any sample size calculation would be highly sensitive to it. The recent DELTA2 recommendations state that pilot trials are typically too small to form a reliable basis for a sample size calculation, although they can indicate a realistic target difference.95 At the stage of designing a full-scale evaluation, we will follow the recent DELTA2 guidelines to calculate the sample size, basing it on a wider range of sexual health interventions and on an assessment of what would be considered meaningful to key stakeholders.95 Evidence from previous school-based sexual health interventions suggests that any differences in behaviour are likely to be small;3 for instance, the effective and widely adopted ‘Safer Choices’ intervention found estimated OR effect sizes (intervention vs. control) of 0.83 for sexual initiation (non-significant), 0.63 for frequency of sex without a condom in last 3 months (p = 0.05) and 0.73 for number of sexual partners without a condom in last 3 months (p = 0.02).133 In our case, an OR of 1.2 may be realistic for an outcome of ‘always safer sex’. In terms of what is considered meaningful, a drop in the primary outcome of a risk behaviour may be less important to educational stakeholders than improved well-being, and health sector stakeholders may be happy to accept a small change in primary outcome when other secondary outcomes (such as knowledge about where to seek help and self-efficacy in condom use) improve. Such considerations are key, particularly when only small changes in the primary outcome are realistic. A further consideration is the heterogeneity in measurement of primary outcomes for youth-focused STI prevention. The DELTA2 guidance recommends performing a sensitivity analysis to consider the effect of uncertainty in the estimate of the target difference. This study provides useful information to contribute to our review of existing literature: a control proportion of 0.854 for practising ‘always safer sex’ and an ICC estimate (based on follow-up and control students) of 0.0198.

Objective 7

Determine the key components of a future cost-effectiveness analysis and test data collection methods.

The total cost of the STASH study intervention was approximately £42 per student or £4258 per school for a year group of approximately 100 students. This varied from £23 per pupil to £93 per student, dependent on year group size. This suggests that the intervention may be more costly to deliver in smaller schools, a fact that should be borne in mind for future evaluation and roll-out. The question bank on sexual health-related resource use was completed well, with relatively low levels of missing data. We established that the CHU-9D was unsuited to calculating QALYs in the context of the STASH study because it was not associated with the primary outcome of always practising safer sex. This analysis confirmed our lack of theoretical expectation that the intervention would have an effect on the domains measured by the CHU-9D, making the CHU-9D an inappropriate and insensitive measure for any effects on well-being that might be achieved by the STASH study. Lower CHU-9D scores in the control arm made the resulting incremental cost-effectiveness ratios, cost-effectiveness acceptability curves and cost-effectiveness planes difficult to interpret. A possible approach for the economic evaluation in a future trial of the STASH intervention could be to model long-term outcomes of STIs and unintended pregnancies related to increases in safer sex, accounting for the disutility of a STI and unplanned pregnancy.

Objective 8

Establish whether or not pre-set progression criteria are met and if a larger-scale trial is warranted.

The study set seven progression criteria regarding the feasibility and acceptability of the intervention and evaluation. We adopted a traffic light system (green, amber, red), with green targets as follows:

• In each of four schools, 60% of nominated students are recruited and complete the training.
• In each of four schools, 60% of PSs complete the training, send three or more messages or have three or more conversations, and attend two or more follow-up meetings.
• In each of four schools, 60% of PSs report that they ‘liked’ the role.
• In each of four schools, 60% of students who are exposed to the STASH study agree that the intervention was acceptable.
• No major acceptability issues raised.
• Less than 15% of PSs report that their parents or carers were unhappy about them being a PS.
• In each of four schools, achieve student response rates of > 70% at baseline and follow-up.

We met green targets with ease for all criteria bar the first. This first indicator (red) was discussed at length with the TSC, as stipulated in our protocol. At the trial outset, we agreed that in the case of amber or red, a mitigating strategy should be identified and continuation should be considered only if all other targets were green. Given that the green targets were easily surpassed for all other criteria, the TSC were supportive of continuation.

Although we did not achieve 60% role uptake among those nominated, we were still able to train 13% of the year group as PSs. This is just short of the 15% deemed critical in the popular opinion leader model of peer education on which ASSIST is based.66 We achieved this partly by overinviting the top nominees (taking the top 25%, rather than the 17.5% invited in ASSIST). Nominees who took up the role received more nominations than those who declined to participate, suggesting that uptake was more likely among the more influential students. This suggests that we may have thrown the net a little wide and that inviting a smaller proportion to recruitment (for example 20%) might still achieve a critical mass of 15%, but with a higher proportional uptake among those invited. The number of friendship ties was broadly similar across those who did take up the role and those who did not take up the role, except that the trained PSs had higher two-step reach centrality (mean of 27 students reached in two steps vs. a mean of 24 students reached for those who did not take up the role). The PSs were well distributed across friendship clusters, with results similar to those in ASSIST.40 PSs also had higher potential to reach across their network than non-nominated students, and were reasonably representative of their year group, being similar in terms of ethnicity, age, religiosity and sexual orientation, but slightly more academic and more likely to be eligible for FSM (a proxy for low-income household). PSs were slightly more likely to report some sexual experience at baseline, such that role uptake may have been related to perceived personal relevance of the topic and/or readiness to hear messages about sexual health.

The target of 60% was based on the experience of ASSIST and extensions of ASSIST67,128 [in which role uptake was high (> 90%)], but may have been set too high for the STASH study given the topic and age group of the students. Among younger students in ASSIST, the prospect of a 2-day training out of school provided a strong motivation, whereas in the STASH study this element received a mixed response, because of academic pressures. There is little evidence from peer-led sexual health interventions to guide decisions on an appropriate target for recruitment. Of five peer-led studies included in a recent European systematic review,32 none used peer nomination to select peer educators (two employed teacher selection, two asked for volunteers and one did not report the selection method). The USA-based STAND study41 is, to our knowledge, the only other youth-focused sexual health intervention to employ peer nomination. The STAND study41 was also a school-based study targeting 15- to 16-year-olds. Their peer nomination process identified 42 opinion leaders, of whom 21 were trained (50%), a very similar proportion to the STASH study. This suggests that 50% uptake may be a more reasonable target. Nonetheless, it is important to think about how we might improve uptake in the future. The target of 60% was far surpassed in one school (77%) and nearly met in two schools (58% and 56%), suggesting that more intensive work to prepare and support schools may be important. In establishing the appropriate proportion of nominees invited to the recruitment meeting, it will be important to identify the percentage (somewhere between 17% and 25%) that achieves the highest proportional uptake while still recruiting 15% of the year. Other possible mitigation strategies regarding the recruitment process are discussed in Recommendations for future evaluation.

Recommendations for future evaluation

Adaptation to the intervention

• Further strategies to enhance awareness of the intervention in the year group and among school staff should be developed. Strategies employed in this feasibility trial included announcements in registration class, assemblies, bulletins and staff meetings. The PSs were given flexibility to ‘announce’ the project according to their preference. In practice, the effectiveness of these activities depended on the support they received from school staff. Going forward, the intervention might supplement these activities via a standard trainer-delivered SRE session, which would give key STI prevention messages and introduce both the project and the PSs. Such a session would help generate a more supportive context for PS activities, reducing the ‘surprise’ element of sexual health messages appearing in student social media threads. It would also reduce the inbuilt inequality of exposure being dependent on proximity to a PS and increase the coherence60 of the intervention for participants (in understanding why only a proportion of the year received training). The cost implications in terms of additional trainer time would need to be considered.
• Uptake of the PS role could be improved in a number of ways. On average, 13% of nominees did not attend the recruitment meeting and so did not have a chance to consider participation. The importance of the recruitment session needs to be emphasised to schools, along with practical suggestions as to how they can support the process. For larger schools, in particular, it may be prudent to split the recruitment meeting into two sessions, on the basis that students may feel more comfortable in smaller groups. The second session would have an additional benefit of providing a mop-up for those who missed the first. We found that in addition to school support, the ‘pitch’ given at the recruitment meeting is very important. Equipping trainers to deliver a strong pitch, with a focus on the key motivators (identified as curiosity, benefit to CV and fun) may also help to increase uptake. Anonymous PS testimonies given in this study could be integrated into recruitment materials and if the STASH study is delivered in subsequent years within the same school, there would also be the option of inviting previously trained PSs to talk about their experience. This is likely to be both reassuring and motivating.
• The study established social media as effective at diffusing intervention messages when used in conjunction with face-to-face elements. At a practical level, the social media element needs to be ‘future-proofed’ to cope with evolving engagement in social media and technological advances. In addition, the appropriate level of control over social media use requires confirmation. Ideally, PSs would have flexibility to engage with STASH study messages using a social media platform of their choice, giving them greater ownership and potential to tailor to the social media preferences of their friendship group. The absence of any untoward incidents in our study suggests that some loosening of control and enhanced flexibility may be justifiable and beneficial; however, this will require careful consideration and monitoring. On a practical level, updates to the website should aim for simplicity for users and minimisation of technical issues.
• A challenge for the STASH study was the range and complexity of messages required to fully equip young people for positive sexual health and avoidance of STIs. This challenge was intensified by the wide range of sexual experience and existing knowledge across the year group, such that individuals differed in the messages they required and found relevant. Our literature review strongly supported a holistic intervention content (rather than a focus on STIs and risk)11–13 and flexibility for PSs to engage with messages according to their interest. Although beneficial in terms of ownership and engagement, this high level of choice was challenging for some PSs. In future, the intervention might supplement the curriculum with a set of streamlined, simple and memorable ‘basic set’ messages that PSs are encouraged to engage with as a minimum.
• Any future roll-out of the intervention will need to address the gender bias in engagement. A finer-grained analysis of motivations to participate might identify particular motivators for young men that could be used to strengthen the appeal for them.
• In terms of timetabling, the peer nomination process could be moved back to June (prior to summer break) so that the baseline questionnaire could be implemented as soon as schools return from the summer break. This would have an important additional benefit of giving more time for intervention delivery before the Christmas break. It is possible, although unlikely, that friendship networks change over the summer break, so this change may need to be piloted (e.g. by repeating the nomination survey in a single school).
• Our health economics analysis suggests that delivery in a small school is potentially more costly.
• In the longer term, the ideal implementation model for the STASH intervention could be an annual cycle of delivery for successive S4 year groups. Trained PSs moving into S5 and S6 (the final 2 years of secondary school) could support the PSs and deliver additional activities, such as supporting school drop-in clinics, facilitating SRE sessions for younger years and undertaking outreach. A wider implementation model such as this might facilitate embeddedness within schools and contribute to longer-lasting impact. After the study ended, the STASH trainers continued to work with PSs in several schools post intervention (e.g. to support the school drop-in clinic) and informal learning from this work will contribute to future discussion on intervention implementation.
• A significant alternative adaptation to the model would be to train S6 students (aged 16–18 years) as PSs, rather than S4. The S6 students could use social media and face-to-face conversation to diffuse messages to their peers in S6 (and S4 and S5 if desired, as senior students are often in mixed year group classes), as well as facilitating classroom sessions with S4 students. The project could be delivered in a 3-year cycle, in which S4 students initially taught by S6 PSs eventually become trained PSs themselves. This approach could potentially address some of the limitations identified in this study. First, the older age group would be more palatable to schools because there is less academic pressure in S6 (the main public examinations are undertaken in S5) and students tend to have free periods in which follow-up sessions could be scheduled. Second, there would be a greater proportion of sexually active students for whom the STASH study messages feel personally relevant, thereby potentially addressing some of the identified challenges around intrinsic motivation. Third, an older age group would not require such close control and monitoring of social media use, enabling students to engage more flexibly with different social media platforms. Finally, in later school years, young men tend to ‘catch up’ in terms of reaching a level of maturity conducive to full engagement with the role. Although it could be argued that S6 is too late to prevent unsafe or unhealthy first sexual experiences, the fact that the STASH study is intended to augment rather than replace sex education justifies waiting until a greater proportion of the target group perceives the messages as personally relevant. In an English educational context, this would translate to the lower sixth year group. In schools with a separate sixth form, the intervention may need to take place later in the school year, as new friendship groups will need time to establish.

Adaptation to data collection methods

• Administering the questionnaire was burdensome for schools and a future trial should consider ways to reduce this burden. A return to a paper-and-pen questionnaire is not suggested for reasons of data quality.117,118 One option for a web-based survey, would be for the fieldwork team to carry mobile Wi-Fi ‘hotspots’, which would allow students to remain in class and complete the survey on their mobile phones. A pilot of this approach in another ongoing school-based study within our unit suggests that this is feasible (Equally Safe Research team, University of Glasgow, 2020, personal communication). A small number of tablets could be made available to students without phones or charge. This approach would reduce logistical difficulties and increase privacy. Survey administration may be easier for schools if there is a clearly identified back-up or support teacher for each contact teacher. This person should be present at all early planning meetings to ensure that they understand the study and can stand in effectively.
• The questionnaire was slightly long, both for students undertaking it and for administration within a school period. Reducing the length of the questionnaire should be straightforward to achieve, as analysis of the outcome measures identified a number of items that can be omitted. We have identified specific ways to improve questionnaire wording and filtering. We could also do more to work with teachers to identify language issues ahead of time and put in place measures to support students on a school-by-school basis.
• The questionnaire was experienced as intrusive by some students. This is, to some extent, inevitable in a survey on attitudes and behaviour related to sex. Although we were careful to explain the questionnaire topic in the participant information sheet, and at the start of the questionnaire, we could expand the introduction to the survey on the fieldworker script. A stronger explanation as to why we need to collect the data could be provided, focusing particularly on the questions that raised concern about confidentiality, such as the postcode and friendship module. Reassurances could be repeated throughout the questionnaire. At follow-up, an additional explanation should state the reasons for repeating the same questions, as higher levels of missing data at follow-up may reflect students feeling that they have already answered the questions.
• With respect to the PS questionnaire, judicious use of open-ended questions would allow us to collect further insight into varied perspectives across the PS cadre, augmenting the qualitative data collection.

Methodological strengths and limitations

This study has a number of strengths. Intervention development followed a rigorous and theory-driven approach, involving significant co-production with young people, professional stakeholders and the trainers who subsequently delivered the intervention. Consistent with the existing literature,135–137 we found this process indispensable in establishing what was appealing to young people and feasible to implement. A strong feature of the evaluation was the wide range of data sources and use of synthesis and triangulation to develop a rich picture of intervention experience. In addition to establishing the potential of the intervention approach, the study has made key contributions to understanding the social network position of individuals rated as influential by their peers, and the potential of using social media within the context of peer-led and school-based sexual health interventions.

Some of the study limitations relate to those of feasibility studies in general. The small sample means that the study cannot give an indication of future effectiveness. It also placed limits on the extent to which we were able to investigate the reliability and validity of candidate measures. Owing to the lack of precision, we avoided reporting p-values.

A limitation of the qualitative methods employed in the process evaluation was that embarrassment and stigma associated with the topic appeared to constrain the depth and candidness of student accounts. Conversations were sometimes stilted or dominated by banter that served to mask embarrassment and avoid loss of face. This was particularly the case in a group (albeit friendship) setting and school context. It was therefore difficult to interpret whether or not lack of recall of detailed information about the STASH study related to minimal engagement or reluctance to talk about the topic in front of other group members. This was much less an issue for PSs, who were more comfortable with the topic following their training. In terms of sampling, we were limited in the extent to which we could follow-up PSs who had dropped out or interview students beyond those in PS friendship groups. We also did not interview parents, relying on PSs and teachers to report their reactions. An absence of complaints by parents does not necessarily imply that all parents were happy with the intervention or with their child being a PS. A practical limitation was that group interviews had to take place within a school period. Once participants had gathered, and informed consent had been taken, there were sometimes only 30 minutes remaining for discussion.

Finally, in common with much feasibility work, the STASH study took place within a small geographical area and transferability to regions with different educational contexts will require careful attention to differences in context, particularly those relating to preconditions of the intervention.

Conclusion and first steps towards future evaluation

This study has demonstrated that the STASH study is a feasible approach to undertaking STI prevention within the context of Scottish secondary schools. It can be implemented with good fidelity and is acceptable to those involved. We have established the potential of carefully managed social media use when supported by face-to-face conversation. All bar one of our progression criteria were met, with the weight of evidence supportive of continuation to a full-scale evaluation. Prior to this, we believe that small-scale work may be prudent to ‘future proof’ the website to interact with a range of social media platforms. We should also pilot more flexibility and less monitoring of social media, enhanced ways to increase the salience of the intervention in schools, enhanced methods of peer recruitment and implementation in an older year group. As the feasibility and acceptability of the intervention has been established, it seems reasonable to test any proposed changes to intervention delivery (including the switch to an older age group) as an internal pilot. We envisage the next step as a multicentre evaluation across the UK. This will involve adjustments to other UK education systems, for example in relation to the timing of the STASH study within the academic year, timetabling of activities, school pastoral support structures and existing provision of sex education in schools. In England, sex education has recently become a statutory requirement and the STASH study is likely to appeal to school leaders as a student-led initiative that delivers on this requirement. Other key policies, such as the use of mobile phones in school, are subject to change in future and may impinge both on intervention and evaluation designs. The STASH study will also need to keep apace with technological and legal changes in relation to social media apps. The project will need to build in flexibility to adjust to these differences and potential changes, most of which will be outside the research team’s control.

The starting point for the scaled-up evaluation is further refinement to our programme theory. We will discuss the refinements proposed by this study with key stakeholders, and undertake a literature review and stakeholder consultation to confirm the optimal study design, primary outcome measure and sample size.