A cluster randomised controlled trial to determine the clinical effectiveness and cost-effectiveness of classroom-based cognitive-behavioural therapy (CBT) in reducing symptoms of depression in high-risk adolescents

Sample size

The study was powered to detect a difference of two points in mean Short Mood and Feelings Questionnaire (SMFQ) scores at 12-month follow-up between classroom-based CBT and each of the control arms. The pilot study provided estimates of intraclass correlation coefficients (ICCs; 0.025), mean year group size (n = 203), consent rate (84%) and SMFQ standard deviation (SD; 4.9) in the target population. Based on 80% consent, 80% retention and 20% of children being classified as high risk, a mean cluster size for analysis of 26 high-risk participants per year group was anticipated, requiring a minimum of 22 year groups to detect a difference of two points with 80% power and 2.7% Dunnett-correction two sided alpha.

Recruitment of schools

Information about the project was sent to 66 non-denominational comprehensive secondary schools in Bath and North East Somerset, Bristol, Wiltshire, Nottingham City and Nottinghamshire County. Schools that expressed an interest were contacted by the research team and a face-to-face meeting was arranged to discuss the project. Nine schools were recruited: one for the pilot study and eight for the main trial. In the participating schools, three year groups could not be included as they did not have discrete PSHE lessons. Therefore, 28 year groups were included in the randomisation process for the main trial. Details of participant flow are provided in Figure 1 (pilot study) and Figure 2 (main trial).

FIGURE 1.

Consolidated Standards of Reporting Trials diagram: pilot trial participant flow.

FIGURE 2.

Consolidated Standards of Reporting Trials diagram: main trial participant flow.

Inclusion/exclusion criteria

Interventions were provided during the school day as part of the school PSHE curriculum. All children were eligible unless they did not attend PSHE lessons (e.g. if they were away from school on long-term sickness absence, were withdrawn from PSHE for religious reasons or were educated elsewhere). Young people and their carers were contacted if they were identified as possibly clinically depressed during assessments (i.e. in the highest 2–3% of SMFQ scores). They were provided with information on how to seek further help should they require it, but these young people continued to participate in the programmes running in the school as usual.

Classification of the high-risk group

The target group for the effectiveness analysis were those who were at high risk of depression, that is those who had elevated and persistent symptoms of depression assessed using the SMFQ. 27 Those with scores of ≥ 5 at both screening and baseline assessments were classified as high risk.

Procedure for individual assessments with young people

Assessments were carried out at four time points: screening (SMFQ only), baseline, and 6 and 12 months from baseline. Some variation in assessment intervals occurred to fit around school holidays, examinations and other events. Mean times between assessments were baseline to 6 months = 176 days (SD 34.2); 6 months to 12 months = 181 days (SD 35.8); and baseline to 12 months = 359 days (SD 20.4). There was no difference between trial arms for the interval between time points for primary analysis (baseline to 12 months).

Assessments with young people were self-completed in schools during lesson time in sessions that were led by the research team. Following the pilot study, we anticipated that there would be differential dropout for the year 11 students who had left school by the time of their 12-month follow-up. To maximise response rates in this group in the main trial, any year 11 students who had stayed on for year 12 were assessed in school after 12 months. Those who had left school were contacted by post and/or e-mail (contact details and permission were obtained at 6-month assessment) and offered an option of filling in the questionnaire online and an incentive of entry to a prize draw (the prize was a gift voucher: 10 of £10 and 20 of £5 were available).

Classroom-based cognitive–behavioural therapy: the Resourceful Adolescent Programme UK

The RAP is a manualised depression prevention programme designed for use with groups of young people aged 12–15 years. It is based on a CBT model and interpersonal therapy principles. CBT recognises the importance of negative thoughts and low self-worth/image in the onset and maintenance of depression. These are actively targeted with core treatment components including psycho-education, identifying and challenging negative/dysfunctional thoughts, identifying personal strengths (thereby enhancing self-esteem/image), managing social problems, and learning to problem solve. Students complete their own workbook and group leaders have a detailed manual specifying key learning points and objectives.

For this study, the RAP was modified for use in the UK (RAP-UK). The content, key messages and goals were consistent with the original RAP, but the structure and method of delivery were revised to fit in with the UK education system. The key elements of RAP-UK – personal strengths, helpful thinking, keeping calm, problem solving, support networks and keeping the peace – are organised into nine sessions, each approximately 50–60 minutes long. RAP-UK was designed with a flexible method of delivery in mind to ensure that it could fit into a busy school timetable while retaining the key elements of the programme. Two additional booster sessions were offered to schools approximately 6 months after the initial programme had been completed. These provided an opportunity to review RAP-UK skills and to practise applying them to current difficulties. RAP-UK was delivered by two trained facilitators external to the school.

Attention control Personal, Social and Health Education

The attention control PSHE intervention involved similar time and contact with external providers to classroom-based CBT, but did not include the active components of the CBT intervention. The school delivered its usual PSHE curriculum, but the class teacher was joined by two facilitators from outside the school who assisted with delivering the lessons and engaging with young people. This controlled for the non-specific effects of interventions that are considered important in studies of depression. 29 As with classroom-based CBT, delivery of attention control PSHE was designed around nine 50- to 60-minute sessions, with flexibility to fit in with existing school PSHE programmes.

Usual Personal, Social and Health Education

‘Usual PSHE’ was the usual school PSHE curriculum provided by the school staff and did not involve any external input from the research team.

Programme facilitators

Facilitators had completed at least an undergraduate degree and had experience of working with young people and/or in a health-care setting. They were employed on a sessional basis. Different facilitators delivered the classroom-based CBT and attention control PSHE interventions to minimise therapist contamination. Facilitators received initial training, which covered the identification and management of mental health concerns, group management techniques and training in delivering their specific intervention. This was usually carried out with groups of new staff over 2 or 3 days, with condensed one-to-one training provided for staff joining at different points in the year as required. During the course of delivery, regular separate supervision groups were provided for classroom-based CBT and attention control PSHE facilitators. Teachers were present in all PSHE sessions and remained responsible for managing class behaviour.

Primary outcome

The SMFQ27 is a 13-item scale derived from the 33-item Mood and Feelings Questionnaire. 30 Each item consists of a simple statement (e.g. ‘I didn't enjoy anything at all’), which is rated as being ‘true’ (scores 2), ‘sometimes true’ (scores 1) or ‘not true’ (scores 0). The SMFQ correlates well with other measures of depression and has good test–retest reliability, and higher scores are associated with fulfilling diagnostic criteria for clinical depression. 31–33

Secondary outcomes

Secondary outcomes were the personal failure subscale of the Children's Automatic Thoughts Scale (CATS),34 Rosenberg Self-Esteem Inventory,35 Revised Child Anxiety and Depression Scale (RCADS),36 School Connectedness Scale,37 Attachment Questionnaire for Children38 and items relating to bullying (Olweus Bully/Victim Questionnaire39), self-harm, cannabis use, use of other ‘street drugs’ (e.g. amphetamines, LSD, ecstasy, cocaine, ketamine, crack, heroin), and alcohol use over the last 6 months.

Resource and service use

As part of their standard assessments, all participants were asked to complete the European Quality of Life-5 Dimensions40 (EQ-5D: a standardised measure of quality of life/health outcome) and a modified self-report version of the Client Service Receipt Inventory (CSRI). 41 The CSRI is widely used and has been adapted across the child and adult age range and for different services. 41 The wording and presentation of the questions were simplified and pilot tested for clarity and ease of completion by the younger age group during the pilot phase of this trial. The modified CSRI (see Appendix 2) retrospectively assessed time off school; receipt of mental health or other health services; educational support; antipsychotropic medication (i.e. antidepressants or others); and social work/care services over a 6-month period. The specific service use information recorded in the assessment for each type of service use is summarised in Table 2.

Sociodemographic data

Basic demographic data were collected (sex, age, who you usually live with, ethnicity). The Family Affluence Scale42 was used as an indicator of socioeconomic status.

Process evaluation

The process evaluation followed the RE-AIM (Reach, Efficacy/effectiveness, Adoption, Implementation and Maintenance) framework. 43 Data were collected on attendance and attrition for classroom-based CBT and attention control PSHE by programme facilitators. An independent observer attended 5% of classroom-based CBT sessions to assess treatment fidelity. Feedback was gathered from teachers, young people and facilitators using questionnaires and qualitative interviews.

Statistical analysis

Statistical analyses were conducted using Stata 11 (StataCorp LP, College Station, TX, USA). Descriptive statistics were used to assess the balance between the trial arms at baseline. Effects on the primary outcome (SMFQ at 12-month follow-up) were assessed by intention to treat without imputation. In order to take appropriate account of the hierarchical nature of the data, multivariable mixed-effects regression was used to compare mean SMFQ at 12 months for classroom-based CBT with attention control PSHE and usual PSHE, with adjustment for baseline SMFQ and randomisation variables. In a secondary analysis, further adjustment was made for variables that were imbalanced between the trial arms at baseline. These analyses were repeated for secondary outcomes.

For SMFQ, repeated-measures mixed-effects regression models were used to investigate convergence/divergence between trial arms over time. Preplanned subgroup analyses44 were carried out using interaction terms in the regression models between the randomised arm and the following baseline variables: symptom severity (SMFQ score 5–10, ≥11); self-harm (no/yes); alcohol and/or drug misuse (no/yes); year group (8–11); and family affluence.

Sensitivity analyses were conducted to assess the potential effect of missing data using multiple imputation chained equation methods,45 with the imputation model including all variables associated with absence of primary outcome at 12 months. Variance at class and school level was investigated by including these in the multilevel models. Neither of these sensitivity analyses made any material difference to the primary results and therefore the data presented are those from two-level (individual and year group) models based on observed data only.

Finally, the effect of actual attendance at allocated lessons was examined by estimating the compiler-average causal effect (CACE) using instrumental variable regression,46 weighted using inverse probability weights constructed using baseline SMFQ scores, randomised group and ‘adherence’ (defined as attendance at ≥ 60% of sessions).

Qualitative analysis

Interviews and focus groups were digitally audio recorded and transcribed verbatim. The transcripts were thematically analysed, broadly following the guidelines of Braun and Clarke. 47 Transcripts were coded using NVivo 9 (QSR International, Southport, UK) software. A predefined, broad and descriptive coding framework was developed to enable the generation of initial codes. Each coding unit was coded exclusively into just one category as this allows very clearly defined coding categories to be developed. 48 Three members of the research team independently coded three randomly selected transcripts. Inter-rater reliabilities calculated using NVivo 9 revealed coding agreement ranging from 78% to 100%, indicating satisfactory consistency of interpretation. Any coding inconsistencies were resolved by discussion and consensus. The remaining 14 transcripts were then divided between the researchers and coded using the same framework. Data were then reviewed by the researchers and the themes were redefined as required. The emergent inductively coded themes were examined for consensus and conflict across the different participant groups.

Economic analysis

The service and resource use and cost-effectiveness analysis was carried out according to current best practice methods for conducting economic evaluation alongside trials,49 and specifically alongside cluster randomised controlled trials. 50

The unit costs applied to different types of health service use and for visits to different types of professionals or services because of anxiety or depression are provided in Table 3. The two main sources for the unit costs were the Department of Health's National Schedule of Reference Costs51 [for primary care trusts (PCTs) and NHS trusts combined] and the Personal Social Services Research's Unit Costs of Health and Social Care52 (hourly costs of patient or client contact for various types of health or social care professional).

The service/resource use data were checked and cleaned manually where required. Reported reason(s) for inpatient stays were assessed as being either elective or non-elective and relevant unit costs were applied (see Table 3). For a small minority (< 1%) of participants who reported taking medication for anxiety or depression, the information provided on medication names and on how long they had been taken was too unreliable to use as a basis for estimating these costs (e.g. where young people simply recorded ‘dunno’ or ‘can't remember’, or stated that they used ‘paracetamol’, other over-the-counter medications or herbal remedies, which would have no cost implications for the NHS) and was therefore excluded from further analysis.

The classroom-based CBT and attention control PSHE programmes were costed using detailed project records of resource use. This included the paid time of facilitators delivering the programme, the cost of their training, travel costs, printing costs of course booklets, and an apportionment of the cost of recruiting schools. All costs were calculated as either the amount of resource used multiplied by a unit cost or the total amount incurred over the trial period divided by the number of pupils in participating classes, number of sessions delivered or number of schools, depending on the level at which the cost was incurred.

Cleaning and correction of resource use and EQ-5D data and calculation of service use costs were conducted using PASW Statistics 18 (SPSS Inc., Chicago, IL, USA). Models for analysing incremental cost-effectiveness were fitted using Stata 12.0 software. Two cost-effectiveness analyses were conducted, the first using the SMFQ score and the second using quality-adjusted life-years (QALYs) based on responses to the EQ-5D questionnaire. Derivation of the per-person QALYs from baseline to 12 months involved calculating the social preference weight (or utility) for all those who completed the EQ-5D at each of the three time points, estimating the ‘area under the curve’ between baseline and 6 months and between 6 months and 12 months and summing them. QALYs were therefore calculated only for students who had complete EQ-5D data at all three time points. The social preference weights for the EQ-5D were from a representative sample survey of the UK general population in 1993. 53

Incremental costs, incremental effects and, where relevant, incremental cost-effectiveness ratios (ICERs) were estimated, comparing each of the classroom-based CBT and attention control PSHE trial arms with usual PSHE. The incremental cost per unit decrease in the SMFQ score (as lower scores on the SMFQ indicate better outcome) and the incremental cost per unit QALY increase were estimated. Both unadjusted and adjusted analyses (adjusting for site, mode of delivery, number of students and number of classes) were carried out. The incremental cost per unit decrease in SMFQ was additionally adjusted for SMFQ score at baseline.

Random effects bivariate linear regression models54 were fitted to model cost and effectiveness (SMFQ or QALY) simultaneously, allowing for correlation within clusters and correlation between cost and effectiveness score within participants. Estimates of the mean difference in costs and corresponding standard error, the mean difference in effect and corresponding standard error and (indirectly via the variance–covariance matrix of the regression coefficients) the correlation between the mean cost difference were obtained from these models. Where the ICER was in the north-east quadrant of the cost-effectiveness plane (i.e. intervention has both higher costs and greater effectiveness than control), Fieller's method was applied to obtain a parametric estimate of the 95% confidence interval (CI) for the ICER and construct the cost-effectiveness acceptability curve (CEAC). 49 The degrees of freedom (df) for these calculations (implemented in Stata using the ‘fielleri’ and ‘accepti’ commands) were based on the number of clusters.

For all adjusted analyses, the joint distribution of the difference in costs and effects was displayed using a scatterplot with 2000 cost–effect data pairs generated parametrically (using simulation) from a bivariate normal distribution based on the estimates from the random effects bivariate regression model used to model the cost and effect outcomes. The parametric scatterplots were similar to those generated using a non-parametric bootstrap approach and so, for consistency, we present only the former. Note that because the cost of the intervention must be apportioned across all participants in a given trial arm, both the SMFQ- and QALY-based cost-effectiveness results are based on the whole sample with valid cost and outcome data (i.e. not just those assessed as high risk as in the primary effectiveness analysis). Given the short time horizon, neither costs nor outcomes were discounted to present values.

Classroom-based cognitive–behavioural therapy delivery

The classroom-based CBT arm included 79 classes. Intervention delivery was adapted to fit in with the existing structure of PSHE delivery within schools, with all programmes covering the core content and key tasks. The RAP-UK programme used was delivered in weekly lessons to 41 classes, while 22 classes had fortnightly lessons and 16 classes had other formats (project days or condensed programme over 3 weeks). All nine sessions were delivered to 71 classes, while four classes had a condensed eight-session programme and four classes had a seven-session programme. The RAP-UK programme was terminated after the fourth session for a whole year group (eight classes) in one school. This was due to disruption caused by adverse weather, which led to several days of school closure. Owing to competing priorities (i.e. making up time for examined subjects), the school was unable to complete the programme.

The additional RAP-UK booster sessions offered were completed by 40 of the classroom-based CBT classes. Of those classes that did not complete the boosters, eight were those that had terminated the programme at session 4, 16 classes were originally in year 11 so either had left school or were not doing PSHE the following academic year, eight classes had moved from year 10 to year 11 and no longer did PSHE, and one year group opted not to complete the booster sessions. Therefore, a total of 832 young people in the classroom-based CBT group (47.8%) attended at least one RAP-UK booster session to refresh and consolidate their skills approximately 6 months after completing the core programme.

Treatment fidelity

Of the 36 classroom-based CBT sessions observed to assess intervention fidelity, 31 covered all the core tasks, with at least 75% of core tasks being covered in the remaining five sessions. Differences in facilitator ratings of lesson content for classroom-based CBT versus attention control PSHE were examined using one-way analysis of variance and indicated that the classroom-based CBT focused more on self-esteem, emotional awareness and positive thinking, and less on topics traditionally covered in PSHE, i.e. sex education, ethical issues, diversity, religion and citizenship (all p < 0.05). There was no difference between classroom-based CBT and attention control PSHE in facilitators' views of student engagement with lessons.

Evaluation questionnaire data

Questionnaires evaluating their PSHE lessons were completed by 482 boys and 506 girls from seven of the eight schools (Table 22). Of these, 386 (39.1%) received the classroom-based CBT programme (RAP-UK), 316 (32%) received attention control PSHE and 286 (28.9%) received usual PSHE.

Taking the sample overall, the majority of young people did not rate their PSHE lessons positively, with only 38.3% of the usual PSHE group and 36.5% of those who received classroom-based CBT positively rating PSHE (6 or more out of 10 for liking their lessons). Similarly, 42.1% of those in the usual PSHE group and 35.4% of those who received classroom-based CBT rated the usefulness of PSHE in everyday life as 6 or more out of 10. Approximately half felt that their lessons were relevant for their age (usual PSHE, 53.7%; classroom-based CBT, 49.6%). Usual PSHE was rated more positively than both classroom-based CBT and attention control PSHE for liking (F = 7.11, df 2,970; p < 0.01), usefulness (F = 6.46, df 2,966; p < 0.01) and relevance for their age (F = 8.84, df 2,963; p < 0.01).

In the high-risk group, no differences were found in the ratings of the three trial arms (n = 203). Further, when the analysis was split by sex, there were no differences between trial arms in girls' ratings of PSHE, but boys preferred usual PSHE to either classroom-based CBT or attention control PSHE (liking F = 5.49, df 2,469; p < 0.01; usefulness F = 7.26, df 2,465; p < 0.01; relevance for their age F = 7.72, df 2,465; p < 0.01).

Costing the intervention

Table 24 shows the key data and costs that were used to calculate the mean intervention cost per student.

The classroom-based CBT intervention cost an estimated £41.96 per child, while attention control PSHE cost an estimated £34.45 per child (Table 25). These costs exclude the costs of developing or adapting the new materials (these are ‘sunk costs’ and it is assumed that they would not be incurred again) or the estimated proportions of people's time, which are as a result of the research/trial context of programme delivery. The costs do, however, include a share of the initial training costs of the facilitators (time of trainers and facilitators, room hire and subsistence). Therefore, they do reflect the relatively high ratio of facilitators to total number of classes delivered, which might not be as high if the classroom-based CBT or attention control PSHE was rolled out on a larger scale and for longer. These ratios were 39 facilitators to 79 classes for classroom-based CBT, and 35 to 72 for attention control PSHE. Furthermore, with attention control PSHE there are no booklet costs and outside of a randomised controlled trial we assume that there would also be no costs associated with school recruitment to such a programme.

Cost of health and social care usage

Table 26 (hospital health care costs) and Table 27 (community-based care costs) show the health and social care service and resources data and costs that were used to calculate the mean service user cost per student in the cost-effectiveness analysis (i.e. for the subsample with complete EQ-5D responses and costs at all three study time points). Overall, these data show that the total care costs for adolescents were similar across the three trial arms at all three time points. Between 26% and 33% of adolescents in each trial arm had hospital care and between 37% and 45% had seen a general practitioner (GP) during the previous 6 months for any health reason. However, rates of service use or seeing care professionals for worry, anxiety or unhappiness were much lower: 1.6–4.6% of adolescents saw a GP for such problems; 0.5–3.0% saw a school nurse; and < 1% saw either the Child Mental Health Service or a child psychologist during a 6-month period. Of the 4.0–6.7% who saw ‘someone else’ for their psychological problems, most reported seeing or talking to their parents or other close relatives; teachers; friends; other doctors; youth workers; or a mentor/counsellor.

In total, over the 18 months for which service use data were collected, the per-person cost of services used by adolescents in each trial arm were £484 (SD £1294) for classroom-based CBT, £385 (SD £1169) for usual PSHE and £483 (SD £1294) for attention control PSHE. We did not test for the statistical significance of any service use cost differences between trial arms both because the study was not powered to detect such differences and because, in a cost-effectiveness analysis, it is more important to estimate the CIs around the joint distribution of cost and effectiveness differences.

Impacts on health-related quality of life (European Quality of Life-5 Dimensions)

Figure 3a shows the EQ-5D index scores by trial arm and time point that were used to calculate incremental effectiveness in relation to health-related quality of life, and used in the cost-effectiveness analysis. For completeness, Figure 3b shows the equivalent results for the EQ-5D visual analogue scale scores.

FIGURE 3.

(a) European Quality of Life-5 Dimensions index scores; and (b) visual analogue scale scores by time point and trial arm.

Both these figures and the data on which they are based (Table 28) show that any mean changes over time or mean differences between trial arms in either of these scores were minimal [varying less than 0.03 (EQ-5D index) or less than 2 points (EQ-5D visual analogue scale) over time or between trial arms]. As with the cost data, we did not test for the statistical significance of any service use cost differences between trial arms both because the study was not powered to detect such differences and because, in cost-effectiveness analysis, it is more important to estimate the CIs around the joint distribution of cost and effectiveness differences.

The detailed profiles of health-related quality of life by individual EQ-5D question and response level, by both trial arm and study time point, are shown in Tables 29 and 30. These tables show that the vast majority of self-reported health problems (affecting between 15% and 20% of adolescents) related to either pain/discomfort or anxiety/depression, while relatively few reported health problems impacting on mobility, self-care or usual activities.

Incremental cost-effectiveness

Tables 31 and 32 show the incremental cost-effectiveness for each of classroom-based CBT and attention control PSHE per person compared with usual PSHE for the 12-month period of the trial. This includes SMFQ score at 12 months and QALYs from baseline to 12 months as outcomes, and adjusted and unadjusted analyses are shown. Attention control PSHE was both less effective and more costly than usual PSHE; that is to say, both the point estimate ICER and most of the joint distribution of incremental costs and effects are in the north-west quadrant of the cost-effectiveness plane. All of the analyses except one found classroom-based CBT to be both less effective and more costly than the usual PSHE control. For the adjusted analysis of the incremental cost per QALY of classroom-based CBT compared with controls there is a very small estimated per person QALY gain (0.00054) for the additional cost of £100 per person.

Scatterplots of incremental costs and QALYs (Figures 4 and 5) and the related cost-effectiveness acceptability curve (Figure 6) were produced. The cost-effectiveness acceptability curve shows that there is an approximately 5% probability that classroom-based CBT is less costly than usual PSHE, and a 46% probability that it is less effective than usual PSHE. An incremental cost per QALY gained of £185,000 (95% CI is undefined) was estimated for classroom-based CBT in the adjusted analysis. This is both higher than the normally accepted willingness to pay for QALYs when they relate to the adoption of health technologies within the NHS58 and highly uncertain. Further to the statistically non-significant effectiveness results for nearly all comparisons and analyses conducted, it can also be concluded that neither classroom-based CBT nor attention control PHSE would be cost-effective compared with usual PSHE. Given the small and highly uncertain differences in cost and effectiveness (SMFQ or QALYs) between classroom-based CBT and attention control PHSE, it would not be informative to estimate the incremental cost-effectiveness of these comparisons.

FIGURE 4.

Cost-effectiveness plane. Classroom-based CBT vs. usual PSHE, ratio of incremental cost to QALYs gained (adjusted analysis).

FIGURE 5.

Scatterplots of incremental costs and effects for the adjusted analyses. (a) Classroom-based CBT vs. usual PSHE: SMFQ; (b) attention control PSHE vs. usual PSHE: SMFQ; (c) classroom-based CBT vs. usual PSHE: QALYs; and (d) attention control PSHE vs. usual PSHE: QALYs.

FIGURE 6.

Probability that classroom-based CBT is cost-effective compared with usual PSHE based on cost per QALY (adjusted analysis).

Reach and effectiveness

The reach of classroom-based CBT within participating schools was good, with the vast majority of young people on roll attending PSHE and therefore being able to access the intervention. Over 90% of eligible young people assented to complete assessments, and over 80% of young people completing baseline were retained at 12-month follow-up. Attendance of PSHE was good and attrition from the PSHE programmes was low. Nonetheless, initial recruitment of schools to the trial was challenging, with only 9 of the 66 schools approached agreeing to take part. This was partly due to other pressures on schools [e.g. Ofsted (Office for Standards in Education, Children's Services and Skills) inspections, staffing changes, participation in other projects]. It may have also been due to the additional demands of a research trial in terms of assessments and research procedures (e.g. randomisation, consent/assent, assessments). However, this may have been indicative of a wider lack of engagement with the project, particularly given other pressures on their time and resources and the general emphasis on academic targets within the UK education system.

The classroom-based CBT programme used (RAP-UK) specifically targeted factors reported to protect against the development of depression, that is thinking styles, emotional recognition and regulation and coping skills. 59 However, we observed that when variables that were imbalanced at baseline had been adjusted for, there was some evidence that classroom-based CBT could have had a small but potentially harmful effect on symptoms of depression compared with usual PSHE in the high-risk group. Furthermore, the high-risk group receiving classroom-based CBT reported more feelings of personal failure at 12 months than those who received usual PSHE. This may indicate a detrimental effect, reflecting limited improvement in mood over time in the high-risk group. While the possibility of a small harmful effect needs to be acknowledged, higher rates of symptoms may reflect greater self-recognition and acknowledgement of exisiting symptoms of depression and negative thinking styles. Furthermore, there was some evidence of potentially beneficial effects of classroom-based CBT on some of the secondary outcomes (i.e. RCADS depression at 12 months compared with usual PSHE, self-harm thoughts at 6 months, and SMFQ as a binary outcome at 12 months compared with attention control PSHE). Similarly, when all participants were included in the analysis regardless of risk status, both potential benefits and potential harms of classroom-based CBT were identified, although all effects were modest and unlikely to be clinically significant.

Unfortunately, our data do not extend beyond 12 months so the longer-term effects of the interventions are not known. Nonetheless, in the short term, our results suggested that, compared with no intervention, classroom-based CBT targeting symptoms of depression in high-risk adolescents was associated with increased levels of self-reported depressive symptoms. The possibility of worse outcomes following school-based intervention programmes has also been found for other mental health disorders such as attention-deficit hyperactivity disorder, and an improved understanding of the mechanisms underlying possible adverse outcomes merits further investigation. 60

Adoption and implementation

Classroom-based CBT was adopted and successfully implemented in the participating schools, but this process proved to be immensely challenging. Schools are very busy and complex environments where there are many competing demands. Furthermore, PSHE is generally undervalued, under-resourced and not prioritised in secondary schools. Finding time, resources and the enthusiasm required to implement a focused mental health programme in this context is difficult. RAP-UK was designed to fit in with existing school structures and enormous flexibility was required to attempt to fit the programme into schools. To accommodate this, a team of over 70 trained facilitators was required to deliver the interventions, predominantly owing to timetabling issues. This posed a major challenge in terms of training and co-ordination of the intervention delivery.

Our experience was that successful delivery of the interventions required a great deal of organisation on the part of schools and the research team, involving regular communication, fitting in around timetables and other demands (e.g. examinations, sports days, work experience) and managing unexpected events (e.g. school closures due to adverse weather). In cases where lessons were cancelled, it could be very difficult to rearrange these. For example, the classroom-based CBT intervention was terminated after session 4 in a whole year group in one school because the school had lost several days to closures during adverse weather and they needed the time allocated to the programme to catch up on examined subjects. Therefore, despite having agreed the best way of implementing the programme with PSHE co-ordinators in schools, we experienced regular ongoing difficulties with delivery because of competing pressures.

Feedback from teachers and facilitators indicated that a shorter and more fast-paced programme might have been more engaging and easier to implement. However, reviews of depression prevention programmes in adolescents indicate that between 8 and 12 sessions are required for these to be effective. 8,9 Part of the issue for teachers appeared to be insufficient content in the RAP-UK programme given the amount of time allocated to the classroom-based CBT intervention; therefore, including more materials and activities to ensure that the students remained engaged and time was used constructively might have been helpful, rather than reducing the programme in length. Alternatively, Wolfe et al. 61 have argued that, to enhance maintenance, shorter programmes should be considered. Programmes could potentially be delivered more flexibly over a longer period of time. Continuous and extended programmes are advocated by Wells et al. 62 and may help to sustain a given programme's benefit. 63 Nonetheless, the current format of delivering nine classroom-based CBT sessions within an academic year was difficult to implement and it is unlikely that this could be sustained outside the context of a research trial.

Engagement with and sustainability of school-based intervention programmes is likely to be strongly influenced by levels of teacher and pupil satisfaction. 64 A number of previous studies have reported that RAP was positively received and perceived to be useful by teachers and young people,16,61,65 and this was our experience in the pilot study. However, in the main trial where the interventions were implemented under more diverse conditions, the feedback was more mixed. In the main trial, a number of positive aspects of the classroom-based CBT were noted in the feedback received from young people and teachers, particularly relating to the content and structure of the RAP-UK sessions, the use of hands-on and interactive tasks, and the usefulness of some of the components of the programme (e.g. recognising body signals and problem solving). Nonetheless, the feedback relating to the acceptability and sustainability of classroom-based CBT was not generally very positive, despite us having revised the intervention and training for facilitators following the feedback from the pilot study.

One of the major challenges faced was that PSHE in general was not positively rated by students regardless of trial arm, and teachers noted that it was an undervalued subject in schools. Given that PSHE is a mandatory part of the school curriculum and deals with issues that are central to the health and well-being of young people, further work needs to be undertaken to make PSHE more engaging, useful and relevant for students. This seems especially important, as fitting programmes into a curriculum that is already so negatively viewed is unlikely to be useful. The feedback questionnaires indicated that classroom-based CBT was more negatively rated than usual PSHE by boys and by the low-risk group, but that there was little difference in the way the programmes were rated in the high-risk group and by girls. Based on the qualitative data, the classroom-based CBT programme appeared to be more acceptable to younger students (year 8) but was less well received in the older year groups, and it was suggested that there was a need for greater differentiation and flexibility, and more emphasis on hands-on tasks.

Overall, teachers were not positive about classroom-based CBT in terms of helping students with their mood. This may reflect a lack of general understanding about childhood depression or a failure to clearly communicate the purpose and aims of the classroom-based CBT programme. Although the research team invested considerable time engaging with the senior management teams and PSHE coordinators at participating schools, interacting with individual teachers proved more difficult. Schools are very busy places, and it was not always possible to find an opportunity at the start of the project for either the research team or facilitators to meet and discuss the project with individual PSHE teachers. Teachers were consulted during the pilot phase of this study to gather feedback and inform the development of the programme before it was rolled out for the main trial. However, it became evident that the teachers involved in the main trial would have liked to be more directly involved in developing the programmes and tailoring them to their students. Teachers in a study of RAP in New Zealand also felt that they could have delivered the programme more effectively if they had been able to deviate from the manual and adapt the way some of the concepts were taught. 25 Involving teachers in programme development and planning may have improved engagement with teachers and young people, but this process would affect treatment fidelity unless all the key active elements were retained, so care would need to be taken with this approach. Furthermore, our experience was that teachers were very busy and had many demands on their time. Finding time to work with facilitators on planning, preparation and reviewing progress often proved impossible in reality, even when teachers were very motivated and engaged with the programme.

The majority of classroom-based CBT facilitators felt positive about their role, enjoyed their work and felt they had sufficient training in general. However, almost half of the facilitators did not feel that the classroom-based CBT would improve mental health, with similar numbers feeling that young people were not engaged with the programme. Facilitators' knowledge and confidence have been identified as important factors in relation to the effectiveness of a programme. 66 Providing more training, including an acclimatisation period in schools, allowing more time for facilitators and teachers to work together and develop a relationship and providing more time to familiarise with the programme could have improved acceptability for classroom-based CBT facilitators. However, this may have been less of an issue outside the confines of a research trial if facilitators were able to work regularly with a group of schools over a longer period of time.

Maintenance

Feedback from teachers and PSHE co-ordinators indicated that delivering classroom-based CBT in secondary schools in its current form would not be sustainable. During interviews, teachers unanimously felt that the classroom-based CBT would be too expensive and lengthy to deliver. Training school staff in mental health programmes provides an alternative and potentially more cost-effective model, although school staff tend to be less effective than trained facilitators or health professionals. 9,19 Around half of the teachers involved in classroom-based CBT did not feel confident that they could deliver the programme on their own, suggesting the need for a significant teacher training and support programme if all teachers were expected to deliver it. Teacher feedback also suggested that if they were to deliver the RAP-UK programme used in the current study, it would be reduced and amended, indicating that treatment fidelity would be compromised. Furthermore, enthusiasm and motivation from teachers would be required to engage with young people in delivering classroom-based CBT, and this was not always present. Therefore, neither delivery by external facilitators (more expensive) nor delivery by teachers (likely to be less effective) appeared to represent sustainable models for delivering classroom-based CBT in schools in its current form.

Economic evaluation

Whether based on estimated differences in QALYs or on differences in SMFQ scores, this research provides no evidence that either classroom-based CBT or attention control PSHE is a cost-effective way of reducing symptoms of depression in adolescents in schools. For all of the analyses bar one, the point estimate of the ICER was in the upper-left quadrant and so it would not be a meaningful number to report. The probability that classroom-based CBT or attention control PSHE were both less effective and more costly than usual PSHE ranged from 43% to 98% in the adjusted analyses. Although the adjusted estimate of the mean incremental cost per QALY of classroom-based CBT compared with usual PSHE is about £185,000 per QALY, the 95% CI surrounding this ratio is not definable. The very small estimated QALY gain and the considerable uncertainty surrounding this estimate are consistent with the overall effectiveness results of the trial in relation to the SMFQ outcome. Even for this analysis, however, there is still a 43% estimated probability that classroom-based CBT is both more costly and less effective than usual PSHE, and only a 25% probability that the cost-effectiveness ratio is less than £20,000 per QALY, which is the current threshold below which the National Institute for Health and Care Excellence is likely to recommend treatments on the basis of value for money.

There are currently very few economic evaluations of similar group-CBT programmes for low mood or depression, and none that is for depression prevention trials targeting school children. However, a study comparing individual versus family CBT for children with anxiety disorders in the Netherlands similarly found no significant differences in costs or effectiveness (QALYs). 67 However, individual CBT was found to be both more effective and less costly than family CBT in most sensitivity analyses. 67 In contrast, an earlier modelling-based cost-effectiveness study in Australia, comparing individual CBT with ‘current practice’ to treat major depression in children and adolescents (aged 6–17 years), estimated that CBT saved disability-adjusted life-years (DALYs) at a cost of A$9000 per DALY saved. 68 Therefore, it is not possible to judge whether or not our findings are consistent with those of similar studies because of the lack of trial-based economic evaluations of group CBT to prevent depression, low mood and anxiety in children.

1.1. Acronym

PROMISE: PROmoting Mental health In Schools through Education

2.1. Research Objectives

A number of systematic and meta-reviews have highlighted the potential efficacy of CBT in the prevention and treatment of depression in children. 1,79–82 Although these reviews have been positive a number of issues have been identified including the absence of comparisons with appropriate placebo groups, the need for longer term follow-up and the absence of randomised controlled trails within a UK context. This study aims to address these issues and in particular will:

1. Examine the effectiveness of school based CBT Personal Social and Health Education (PSHE) vs. an attention PSHE (delivered by people external to the school) vs. usual PSHE (school delivered) on the prevention of depressive symptoms in high risk adolescents aged 13-16 as assessed by the Short Mood and Feelings Questionnaire at 6 and 12 months. Our primary follow-up will be 6 months after completing the intervention.

2. Examine the effectiveness of CBT (6 & 12 months) on the secondary outcomes of negative thoughts, self-esteem and anxiety.

3. To undertake a secondary sub-group analysis to investigate the effect modification by school connectedness, bullying, self harm, alcohol and drug misuse on treatment outcome (6 months).

4. Assess the cost-effectiveness of the intervention in terms of health-related quality of life (and cost-utility) at 6 months.

5. To undertake a process evaluation to assess factors associated with adherence and acceptability of the intervention including participant perception of usefulness, examples of on-going skill usage, and satisfaction (6 & 12 months).

3.2. Existing Research

Epidemiological studies suggest that over a six month period up to 2.5% of children and 8.3% of adolescents suffer from a major depressive disorder. 2 Cumulative rates indicate that up to 20% will suffer at least one clinically depressive episode by the age of 18. 1 In the UK prevalence estimates are 1.4% amongst 11–16 year olds in the community and around 20% amongst 13–16 year olds attending primary care. 83,84 Adolescent depression causes significant impairment, impacts on developmental trajectories, interferes with educational attainment and increases the risk of attempted and completed suicide as well as major depressive disorder in adulthood. 2–5 In addition, sub-threshold depressive symptoms in adolescence carry a similar risk to major depression for developing depression and suicidal behaviours later in life. 85 Whilst approximately 50% of children are estimated to spontaneously recover for the remaining half symptoms persist and significantly impair functioning. 82 It is therefore particularly concerning to note that depression in adolescents often remains unrecognised and untreated. 6,7

Depression is an important public health issue and, in view of the above, there has been growing interest in the development of interventions designed to prevent depression in adolescents. The better evaluated depression prevention programmes have tended to be based upon cognitive behavioural therapy (CBT). These include the Penn Resiliency Programme,14 Coping with Stress,16 Problem Solving for Life,17 FRIENDS,18 Resourceful Adolescent Programme10 and other CBT programmes. 19 Some have been delivered as selective interventions to adolescents with elevated symptoms of depression23,24 whilst others have been delivered as universal interventions to whole populations. 10,25 Most preventative CBT interventions have produced immediate short term gains in terms of reductions in depression scores although these are not always maintained at follow-up. For example, Spence, Sheffield and Donovan17 published the largest study to date which involved 1500 children aged 12–14 .The 8-session Problem Solving for Life programme resulted in significant post-treatment reductions in symptoms of depression and an increase in problem solving skills compared to a non-intervention comparison group. However, when assessed at 12 month follow-up this difference was no longer significant. 20

The Cochrane review notes that targeted preventative interventions result in marginally greater reductions in depressive symptoms. 1 This may partly be explained by initial levels of depressive symptoms in the control groups. 71 In targeted interventions the sample is chosen on the basis of risk status and the control group is therefore likely to have higher levels of depressive symptoms at baseline and follow-up. This is in contrast to universal interventions where initial levels of symptoms are lower and may not therefore be high enough to demonstrate a preventative effect at follow-up. However targeted approaches encounter significant recruitment problems that severely limit their potential impact and use. For example, in one targeted study 3935 parents with depression or depressive symptoms were identified. 23 Of these, 2995 were judged appropriate for the study although baseline interviews were conducted with only 551 youth parent dyads, 18% of the eligible sample. Similarly, in a recent study only 271 out of a total population of 6000 responded or were eligible to receive the preventative intervention and of the 147 young people allocated to the active intervention 41 (29%) did not attend. 14 Other reviewers have noted recruitment rates of less than 50% and attrition rates of approximately 30% in targeted preventative studies. 12 A further limitation of targeted approaches is their failure to provide any input to low risk children that would prevent low level symptoms escalating. In a study involving 260 adolescents, 1.2% of the healthy adolescents receiving a preventative intervention moved into the high risk group at follow-up compared with 10.1% in the control group. 10 Selective depression preventative programmes therefore have a limited effect and reach and engage only a small proportion of adolescents who are identified as at risk for depression.

Universal prevention programmes provide an alternative approach. Whilst not all recipients of universal approaches present with elevated depressive symptoms they nonetheless provide a pragmatic alternative, reduce possible negative effects of stigma and labelling, result in lower rates of dropout and greater participation rates. 10,16,17 Indeed, universal approaches achieve recruitment rates of 67%–88%. 10,11,17 Proponents of universal population based approaches argue that whilst they might have a more limited effect on individuals they will nonetheless reduce far more disorders in the population as a whole than a highly effective targeted approach. 26 Indeed, Rose86 argues that a large group of people exposed to a low risk (e.g. minor depressive symptoms) will ultimately generate more clinical cases than a small group of people exposed to a high risk. There is therefore a strong rationale for pursuing universal approaches for the prevention of depression in adolescents.

Of the evaluated universal depression prevention programmes the Resourceful Adolescent Programme (RAP) appears particularly promising. In the initial efficacy study, 260 adolescents were assigned to RAP, RAP plus family involvement or a no intervention group. 10 In terms of reach 85% of the eligible cohort took part and attrition was 5.8%. Adolescents who received either version of RAP reported significantly lower levels of depressive symptomatology at post intervention and 10 month follow-up compared with the no intervention group. In terms of health status, 71% of the RAP group who were classified as “at risk” on the basis of initial depression scores had moved into the healthy range when assessed at post intervention compared with 31% in the control condition. This was maintained at 10 month follow-up where 75% of the RAP high risk group and 41% of the control high risk group scored within the normal range on measures of depression. There was also evidence of a preventative effect since none of the “healthy” adolescents in the RAP group moved into the “at risk” range at follow-up compared to 10.5% of those in the control condition.

In a New Zealand adaptation of the programme, RAP-Kiwi, 392 students were assigned to either RAP-Kiwi or an attention placebo condition. 11 Once again recruitment rates were high (73%) and attrition low (9% attrition at 6 months). Depression scores were significantly lower in the RAP group post intervention. In clinical terms, sixteen students in the RAP group, as determined by scores on the Beck Depression Inventory moved from the moderate/severe to the minimal/mild category at post-test compared to six in the placebo group. The results were less conclusive at 18 month follow-up where although the difference on the Reynolds Adolescent Depression Scale remained significant scores on the Beck Depression Inventory were not.

A large multi-site randomised controlled effectiveness trial in Australia of RAP involving 2664 students from 12 schools has recently been undertaken. 12 RAP participants recorded significantly lower levels of depressive symptoms than those in the control condition at both post-intervention and 12-month follow-up. Based on initial depression scores, significantly more (49.1%) of the at-risk students in the RAP condition moved into the healthy category at post-intervention compared with 35.3% in the control group. This difference was maintained at 12 month follow-up. In a subsequent qualitative evaluation with 109 young people, 61.2% of girls and 46.6% of boys were able to identify specific examples where they had used skills learned during RAP. 12 The authors concluded that within the context of a real world effectiveness study RAP appears to positively affect the health status of “at risk” students.

The absence of long term follow-up and the tendency for post-intervention effects of depression prevention programmes to diminish after 6 months have been noted. 1,20 It would therefore seem important for preventative interventions to include additional booster sessions in order to maintain short term benefits. In addition, there is a need to determine whether improvements are due to the specific treatment components of the intervention or the non-specific factors associated with therapy and whether these improvements are significantly greater than would normally occur over time. Active interventions therefore need to be compared with placebo conditions and usual care. In view of the promising results from RAP this project intends to undertake a pragmatic trial to evaluate the effectiveness of the universal group RAP CBT programme in the UK context. RAP will be delivered via PSHE in schools by trained mental health professionals and will be compared over a 12 month period with an attention PSHE and usual PSHE (treatment as usual).

3.3. Research Methods

3.5. Planned inclusion/exclusion criteria:

3.6. Ethical arrangements

3.6.3. Consent/Assent:

Evaluation of opt-in and opt-out recruitment strategies suggest that opt in strategies result in lower recruitment rates and healthier participants. 89 This led the authors to suggest that opt-out approaches for obtaining parental consent should be the default recruitment strategy for interventions that pose a low risk to participants. The participants in this study are not referred NHS patients, the intervention is low risk, and as such we propose an opt-out approach. At the start of the project a letter will be posted to the carers of all eligible young people informing them about the study. Although all young people will be expected to participate in the interventions only those who opt-in and provide signed consent/assent will complete the assessments. The letter will therefore inform carers that they can opt out of the assessments if they do not wish their child to complete the questionnaires. In addition, the project information sheet will be read to the young people and child consent/assent obtained before completing the questionnaires i.e. dual carer/child consent/assent will be required. Young people will be asked for alternative contact details (i.e. e-mail address/mobile number/home address) in case the research team should need to contact them in relation to the project. This is particularly relevant for Year 11 students who may have left the school before the 12-month follow up.

3.7. Risk and Benefits

The risks of participating in this study are considered to be small. At worse, the proposed interventions may not result in any additional lasting benefits although it is considered highly unlikely that they will have any detrimental effects. A study information sheet, provided at the outset of the study, will be prepared for participants informing them of risks and benefits. If the intervention is shown to have a positive effect then the potential to promote positive mental health in adolescents is a significant benefit to society and the individuals who took part.

3.8. Proposed sample size

The HTA commissioning brief (06/37) specifies a shift in the distribution of depression scores as an important outcome. The primary outcome for this trial will therefore be score on the SMFQ as a continuous measure. The pilot study (n = 711) provided estimates of ICC (0.025), mean year group size (n = 203) and consent rate (at least 80%). A range of target differences for effect sizes of 0.36–0.42 SDs are detectable with 80% power and 5% two-sided alpha with 20–27 year groups. The eight participating schools have a total of 28 participating year groups, meaning that the study will have > 80% power at the 5% alpha level to detect a planned minimum difference of 0.36 SDs.

3.9. Proposed Outcome measures

3.9.1. Psychological Functioning

• Primary Outcome Measure – Short Form Mood and Feelings Questionnaire (SMFQ;27 This 13 item scale is derived from the long form (33 item) Moods and Feeling Questionnaire 30. Each item consists of a simple statement (e.g. I didn't enjoy anything at all) which is rated as being either “true” (scores 2), “sometimes true” (scores 1) or “not true” (scores 0). The short form is designed as a brief self-report screening instrument that can be used to assess severity of depression in community samples 31. The SMFQ is a unifactorial scale with a robust single factor structure. 90,91 Criterion validity (i.e. ability to predict clinical diagnosis) has been established within both clinical27 and community samples33,92 and with children ranging in age from 7–16. The scale correlates well with other measures of depression, has good test/re-test reliability with higher scores tending to be associated with children who fulfill diagnostic criteria for clinical depression. 31–33

• Secondary Outcome Measure – Children's Automatic Thoughts Scale (CATS;34 This self completed scale assesses a range of negative self statements in children and young people aged 7-16. For each item the child is asked to rate whether they have had a similar thought over the past week. Each item is rated as “not at all” (scores 0), “sometimes” (scores 1), “fairly often” (scores 2), “often” (scores 3) or “all the time” (scores 4). Confirmatory factor analysis identified 4 distinct but correlated factors relating to thoughts about physical threat, social threat, personal failure and hostility. 93 Internal consistency for the total score was high (Cronbach Alpha = 0.95) with acceptable test–retest reliability (0.79). The scale has been found to effectively discriminate between a community and clinical sample with the personal failure sub-scale being the strongest predictor of depressive symptoms. 94 The 10 item personal failure sub-scale will be used.

• Secondary Outcome Measure – Rosenberg Self-Esteem Inventory35 The Rosenberg Self-Esteem Scale is a 10-item self-report measure of global self-esteem. It consists of 10 statements related to overall feelings of self-worth or self-acceptance. The items are answered on a four-point scale ranging from, strongly agree to strongly disagree. The Rosenberg Self-Esteem Scale has demonstrated good reliability and validity across a large number of different sample groups and has been validated for use with male and female adolescents as well as substance abusers and other clinical groups, and is regularly used in treatment outcome studies.

• Secondary Outcome – Revised Child Anxiety and Depression Scale (RCADS;36 This self-report scale assesses symptoms of DSM-defined anxiety disorders and major depression in children aged 9–16. The scale is an adapted version of the Spence Children's Anxiety Scale95 and after psychometric evaluation was reduced to 25 items. 96 Each item is rated on a 5 point scale to indicate frequency and can be combined to form 5 sub-scales assessing symptoms of generalised anxiety disorder, separation anxiety disorder, social phobia, panic disorder and major depressive disorder. The scale has good internal consistency, test-re-test stability, and good convergent and divergent validity. 97

FIGURE 1.

PROMISE Procedure Flowchart.

3.10. Data analysis and management

3.11. Research governance

There will be three main management committees: 1) Trial Steering Committee (TSC) This will provide overall supervision of the trial. It will meet at least once a year and its role will be to monitor and supervise the progress of the trial towards achieving its goals; to advise the investigators in general scientific and management issues; and to ensure that there are no major deviations from the trail protocol. The Lead applicant will inform the Chair of the TSC who may call additional meetings when there are matters arising from the conduct or management of the trial that might require their advice. 2) Data Monitoring and Ethics Committee (DMEC). This will monitor data and advise the TSC on whether there are any ethical or safety reasons why the trial should not continue. This will have access to unblinded data and will consider the need for any interim analysis. The DMEC will meet at least annually. 3) Trial Management Group: A separate Trial Management Group will be established to oversee the operational running and progress of the project. This will be chaired by the Lead Applicant and will include the senior researcher and researchers, and other co-applicants as appropriate. The meeting will involve monthly teleconferencing as well as 3 monthly face to face meetings.

In addition, local monthly supervision groups will be established in each of the 4 localities for the psychology assistants delivering the interventions. Separate groups will be established for the RAP and the attention interventions. Each group will meet monthly for 2 hours. Supervisors will be experienced Senior Clinicians from the local CAMHS teams. They will participate in the initial intervention training sessions so that they are familiar and knowledgeable about the intervention. The Lead applicant will offer support and advice to the group supervisors.

3. Project timetable and milestones

Month 1–4 (Sept–Dec 08): Recruitment, induction, establishment of project infra-structure, links with project schools. Training of intervention leaders and research assistants for pilot.

Month 5–10: (Jan–June 09): Feasibility and piloting of recruitment, assessments, randomisation, RAP and attention placebo intervention and process evaluation in 1 school (n = 711).

Month 11–12 (July–August 09): Training of intervention leaders and research assistants for main trial.

Month 13 (Sept 09–January 10): Initial screening and baseline assessments of study cohort (T1).

Month 13–22 (Sep 09–June 10): Intervention starts and is provided across study schools during the next 3 terms (9 months).

Month 22–31 (March–July 10): 6 month (T2) assessments.

Month 28–37 (Sept 10–January 11): Final 12 month (T3) assessment.

Months 31–38 (February 11–June 11): Complete data entry/database cleaning.

Months 38–40 (July–December 11): Data analysis, final project report, preparation of papers for publication.

5. Service Users

RAP is an established manualised programme and thus the content and structure is already determined. However we intend to involve young people to ensure that the content and wording is appropriate for the UK and in seeking their advice and guidance about our proposed process evaluations. We therefore plan to work with young people in two main ways:

Consultation: during the pilot and feasibility stage we will establish 4 focus groups within the pilot school (one for each year group). These will provide opportunities to discuss with young people issues of implementation, ways of maintaining the profile of the project, ideas to maximise assessment completion, their views about the content and wording of the interventions and assessments and how sensitive issues such as deliberate self-harm, drug and alcohol misuse can be presented in a clear and acceptable way.

Collaboration: two young people from the pilot schools will be recruited to become members of the project steering group. The user researcher for this project, Ms Stevens, will be instrumental in recruiting and promoting the importance of this to young people and will undertake a promotion exercise in which research is demystified and their important role and contribution highlighted. They will be supported by the user researcher to attend and participate in the Trial Steering Committee and will therefore be active partners in the management of the project including monitoring progress and potential difficulties, interpretation of findings summarising conclusions & identifying key lessons. An allowance has been made to pay for their time and to support their travel and attendance at meetings.