Training School Teachers to Deliver a Mindfulness Program: Exploring Scalability, Acceptability, Effectiveness, and Cost-effectiveness

Background There is growing research support for the use of mindfulness training (MT) in schools, but almost no high-quality evidence about different training models for people wishing to teach mindfulness in this setting. Effective dissemination of MT relies on the development of scalable training routes. Objective To compare 4 training routes for school teachers wishing to deliver MT differing in intensity and potential scalability, considering teaching competency, training acceptability, and cost-effectiveness. Methods Schools were randomized to an existing route comprising an 8-session instructor-led personal mindfulness course, combined with 4-day MT program training, or 1 of 3 more scalable, lower intensity, alternatives: an instructor-led personal mindfulness course combined with 1-day MT program training, a self-taught personal mindfulness course (delivered through a course book) combined with 4-day MT program training, and a self-taught personal mindfulness course combined with 1-day MT program training. Results Attrition from training was substantial across all routes. The instructor-led course was more effective than the self-taught course in increasing teachers’ personal mindfulness skills. Even the most intensive (existing) training route brought only 29% of the teachers commencing training, and 56% of those completing the study protocol, to the required minimum competency threshold (an advanced beginner rating on an adapted version of the Mindfulness-based Interventions Teaching Assessment Criteria). The differences in levels of competency achieved by existing training compared with the more scalable alternatives were modest, with economic evaluation suggesting that the existing route was both more expensive and more effective than lower intensity alternatives, but with no statistically significant differences between routes. Conclusions This research questions the move toward abbreviating teacher training to increase scalability and suggests instead that many teachers require additional support to ensure competency from first delivery of MT in the classroom.

groups and to discuss issues they may not have had the chance to raise in the larger group, as well as to discuss and plan how they envisaged mindfulness being implemented and supported in their school ('The School Challenge'); Periods of Mindfulness Practice -Extended periods of mindfulness practice were included throughout the training (six sessions in total) in order to allow participants to deepen their personal mindfulness practice over the four-days; 'Our Journey so Far' -these smaller group work sessions involved creating posters charting the structure and flow of the .b lessons in order to understand the scaffolding and logic that lies beneath them; Video clips -observing what it is like to teach .b in the classroom -real lessons being taught to real classes.
One-day Training. The one-day training followed the four-day course with the following adaptations: Lesson Modelling -whilst two key lessons were taught in full, the remaining lessons were partially 'talked through' -with participants being given an overview and brief discussion of lesson content, rather than seeing each lesson being taught in full.
Leading Practice and Enquiry -one opportunity was provided to discuss what it was like to lead mindfulness practice and enquiry with young people, some of the typical issues and challenges that could arise, and how teaching mindfulness to young people might differ from teaching mindfulness to adults; Teach Back -participants had a single opportunity to practice teaching a mindfulness practice and to receive supportive feedback; Home Groups -home group sessions were not included; Periods of mindfulness practice -two shorter periods of mindfulness practice were included in the training day; 'Our Journey so Far' -this exercise was excluded from 1-day training; Video clips -participants watched a real lesson being taught to a real class and discussed their observations. Each domain and overall teaching competency is rated on the following scale: 1incompetent; 2 -beginner; 3 -advanced beginner; 4 -competent; 5 -proficient; 6advanced. For the purposes of the study reported, the overall six point rating was dichotimised at the incompetent or beginner / advanced beginner or above boundary. This is because people reaching advanced beginner level or above are generally regarded as having reached an adequate minimum competency threshold and able to begin teaching under supervision. In the context of a teacher delivering MT in schools this was considered to reflect a reasonable expected level of competency on first delivery of the program.

References
Independent raters who were experienced MBI teachers, experienced in the delivery of the .b MT program, and trained in use of the MBI-TAC received additional training in the administration of the MBI-TAC TEACH addendum in a two day workshop led by WK. Each domain of the MBI TAC TEACH was reviewed and joint ratings were conducted with the use of illustrative teaching video extracts, in order to establish initial consistency. Following the training workshop, all trainees independently rated videos of two participants delivering the .b course, viewing two sessions from each teacher, and rating the six competency domains as well as providing a composite rating based on an overall impression of teaching competency for each teacher. Videos for this phase were benchmarked through consensus rating by WK and LL (an experienced MBI-TAC trainer & rater, and an experienced teacher & .b instructor, respectively) with sessions selected to expose trainees to a broad range of classroom contexts and teacher competency levels. Raters participated in a video conference to review their independent ratings and explore discrepancies. Following this, the reliability of raters was assessed. Raters reviewed videos of a further three participant teachers and rated competency of each. Their ratings were compared to benchmark values (WK & LL) and an average disagreement score was calculated for each rater. All six raters had an average discrepancy of less than one point on the MBI-TAC Teach from the benchmarks, indicating a high degree of consistency between independent raters and the benchmark. During subsequent independent video coding, all raters participated in periodic video conferences to review processes and raise any issues arising during coding.
Raters reviewed two anonymised lessons to assess participants' teaching skills in both the first and second halves of the course, and to view sessions which contained activities and elements addressing all the domains on the MBI-TAC TEACH. Sessions for rating were randomly assigned from Sessions 3&6, 4&6, or 3&7. In some instances, teachers did not capture all lessons on video, because they forgot, mislaid their camera, experienced technical difficulties or had changes in scheduling that meant they could not access the camera at the required time. Where randomly assigned sessions were not available, Session 8 was substituted if available. For n = 88 participants (43% of those randomised, 92% of those rated for competency) two teaching videos were assessed, but for n = 8 cases (4% of those randomised, 8% of those rated for competency) competency had to be assessed via review of a single videoed session.

Reasons for Participant Withdrawal, by Study Phase and Training Route
In the tables that follow we provide reasons for attrition broken down by study phase and training route. It should be noted that this information is accurate to the best of our knowledge and is provided to give an overview of typical reasons for attrition in each study arm. However we relied on participants', and in some instances their colleagues', reports on reasons for attrition, and this should be taken into account in interpreting the data. Note: Six participants withdrew and one returned having missed the T1 data point

Reasons for Participant Withdrawal following Program Training
Instructor-Led, Four-day Training N Camera and memory card stolen within school 1 Memory card and camera lost within school 3 Teacher reported teaching course but did not return video card 1 Teacher could not be timetabled to teach course 4 Total 9 Instructor-Led, One-day Training N Did not teach course 3 Camera/recording failure 3 Teacher did not return memory card 1 Total Self-taught, Four-day Training N Teacher did not teach course 1 Teacher left the school 1 Participants / School were non-responsive 3 Total 5 Self-taught, One-day Training Teacher on sabbatical / teacher left school 3 Camera/recording failure 1 Not able or interested in teaching course 9 Participants /School were non-responsive 2 Teacher reported teaching course but did not return a memory card 1 Total 16 Appendix 4

Full Results of Cost-Effectiveness Analysis
Results of the primary cost-effectiveness analysis, considering all teachers that were randomised, suggest a larger difference in both costs and outcomes between the standard training route, instructor-led, four-day, and one-day training routes compared to the selftaught, four-day training route (Table S3). The comparison between instructor-led, four-day and self-taught, one-day shows an adjusted difference between costs of £729.80 and adjusted difference between effects (proportion of competent teachers) of 12.15%, generating an ICER of £60.04 per percentage point increase in proportion of teachers deemed competent.
Comparing instructor-led, four-day with instructor-led, one-day shows a difference in costs of £479.14 and a difference in effects of 9.81%, resulting in an ICER of £48.53 per percentage point increase in proportion of teachers reaching the threshold for adequate minimum competency. The comparison of instructor-led, four-day with self-taught, four-day shows relatively smaller differences in costs (adjusted mean difference: £361.21) and effects (adjusted mean difference: -3.62%), resulting in an ICER of -£99.73 per percentage point increase in proportion of competent teachers. However, in the scenario analysis the comparison of instructor-led, four-day with self-taught, four-day shows a difference in costs of £281.09 and a difference in effects of -17.75%, due to the higher rates of competency for self-taught, four-day route in this analysis (63% competent in scenario analysis versus 24% competent in the primary analysis; Table 7). Comparison of instructor-led, four-day with both one-day routes for the scenario analysis shows similar results as in the primary analysis.
The cost-effectiveness plane for the primary analysis ( Figure S1) shows that most scatter points for comparisons between instructor-led, four-day training and self-taught, oneday training (blue), and between instructor-led, four-day training and instructor-led, one-day training (grey line) fall mainly in the northeast quadrant, indicating that instructor-led, fourday training is more costly, but also more effective. Comparing instructor-led, four-day training with self-taught, four-day training however results in more scatter points falling across the northeast (instructor-led, four-day training more costly and more effective) and northwest (instructor-led, four-day training more costly and less effective) quadrants. This indicates a lower probability of instructor-led, four-day training being cost-effective compared to self-taught, four-day training, than of instructor-led, four-day training being cost-effective compared to either of the one-day routes.
The associated cost-effectiveness acceptability curves ( Figure S2) suggest that the probability of instructor-led, four-day training being cost-effective compared to one-day training, whether with instructor-led or self-taught MT (blue and grey lines), is heavily dependent on society's willingness to pay for an increase in competent teachers; the higher the willingness to pay, the higher the probability of instructor-led, four-day training being cost-effective compared to the one-day training options. In contrast, the results comparing instructor-led, four-day training and self-taught, four-day training (orange line) suggests that the probability of instructor-led, four-day training being cost-effective is below 50% irrespective of willingness to pay. In other words, there is a higher probability of self-taught, four-day training being cost-effective compared to instructor-led, four-day training.
The scenario analysis, considering participants that completed the study protocol and submitted codeable videos shows greater variability in estimates of costs and effects for all three comparisons (Figures S3 and S4). However, results are very similar to those in the primary economic analysis, with the cost-effectiveness of instructor-led, four day training compared to either instructor-led or self-taught, one-day training, being uncertain and dependent on willingness to pay, whilst the comparison with self-taught, four-day training is clearer and suggests a greater probability of self-taught, four-day training being cost-effective compared to instructor-led, four-day training.
Appendix 5 Supplementary Table Legends   Table S1. Comparison of baseline characteristics of those providing and not providing data at each time point. Table S2. Comparison of teacher primary outcomes and mechanism measures at postintervention follow-up (T2), adjusting for baseline (T0). Table S3. Summary of costs used in economic analyses. Table S4. Differences in mean costs and effects (proportion of teachers reaching minimum competency threshold) among those who were randomised (primary analysis) and those who completed study protocol (scenario analysis).       IL4D: instructor-led, four-day; ST4D: self-taught, four-day; IL1D: instructor-led, one-day; ST1D: self-taught, one-day * From model including intervention group and BL variables -gender, age, cluster size (<5) Figure S1. Bootstrapped mean differences in costs and effects (proportion of teachers reaching the minimum competency threshold) among those who were randomised (primary analysis).

Supplementary Figure Legends
IL4D: instructor-led four-day; ST4D: self-taught four-day; IL1D: instructor-led one-day; ST1D: self-taught one-day NW: northwest (more costly, less effective); NE: northeast (more costly, more effective); SE: southeast (less costly, more effective); SW: southwest (less costly, less effective) Figure S2. Cost-effectiveness acceptability curve showing the probability that standard training (IL4D) is cost-effective compared to less intensive training routes for different values of willingness to pay for percentage point increase in teachers reaching the minimum competency threshold among those who were randomised (primary analysis).
IL4D: instructor-led four-day; ST4D: self-taught four-day; IL1D: instructor-led one-day; ST1D: self-taught one-day Figure S3. Bootstrapped mean differences in costs and effects (proportion of teachers reaching the minimum competency threshold) among those who completed study protocol (scenario analysis).
IL4D: instructor-led four-day; ST4D: self-taught four-day; IL1D: instructor-led one-day; ST1D: self-taught one-day NW: northwest (more costly, less effective); NE: northeast (more costly, more effective); SE: southeast (less costly, more effective); SW: southwest (less costly, less effective) Figure S4. Cost-effectiveness acceptability curve showing the probability that standard training (IL4D) is cost-effective compared to less intensive training routes for different values of willingness to pay for percentage point increase in teachers reaching the minimum competency threshold among those who completed study protocol (scenario analysis).