Mental health first aid training for high school teachers: a cluster randomized trial

Design

The study was a cluster randomized trial with schools as clusters and individual teachers the participants. A cluster design was used because it was not feasible to randomly assign individual teachers who were working in the same school because: (1) there may have been contamination of information provided across groups within the same school, and (2) schools may have responded to the training with changes in policy or procedures which would affect all teachers. Schools were randomly assigned to either receive training immediately or be placed on a wait list to receive training once the trial had finished. The trial has been registered with the Australian and New Zealand Clinical Trials Registry (ACTRN12608000561381).

Participants

Intervention

Teachers received a modified version of the Youth Mental Health First Aid course. To meet the scheduling needs of schools, the course was organized into two one-day parts of seven hours each. Part 1 was designed for all education staff and covered departmental policy on mental health issues, common mental disorders in adolescents (depressive and anxiety disorders, suicidal thoughts and behaviours, and non-suicidal self-injury) and how to apply the mental health action plan to help a student with such a problem. Part 2 was for teachers who had a particular responsibility for student welfare. It provided information about first aid approaches for crises that require a more comprehensive response and information about responses for less common mental health problems. Topics included how to give initial help to students who are experiencing a psychotic or eating disorder or substance misuse. Training was administered at the participants' school, with all available staff participating.

As documentation of the intervention, there was a lesson plan for each session, the existing Youth Mental Health First Aid manual [12] and a set of mental health factsheets. Lesson plans were developed by two Mental Health First Aid trainers of instructors who had previously worked as teachers. Additional material was added by staff of the Department of Education and Children's Services. Each course was conducted by two instructors, one from the Department of Education and Children's Services and the other from the Child and Adolescent Mental Health Service. These instructors received a one-week training program in how to conduct this modified Youth Mental Health First Aid course. They were trained by two experienced trainers, including Betty Kitchener who devised the Mental Health First Aid course.

Objectives

For teachers, the hypotheses tested were that mental health first aid training improves the following: mental health knowledge, stigmatizing attitudes, confidence in helping students, helping behaviours towards their students, knowledge of school policies and procedures for dealing with student mental health problems, support given to colleagues with mental health problems, seeking information about mental health problems and their own mental health. The primary outcome measure for the trial was teacher knowledge.

For students, the hypotheses tested were that the mental health first aid training of their teachers would lead to an increase in the information they receive about mental health problems from their teachers, and that their mental health would improve.

All hypotheses pertained to the individual rather than the cluster level.

Outcomes

The following teacher outcomes were measured at the individual level:

Sample size estimation

Required sample size was estimated using software for power analysis in cluster randomized trials [18]. Likely effect sizes were taken from a randomized trial of Mental Health First Aid in a workplace setting [4]. In this workplace trial, recognition of the disorder in a vignette improved 10% in the intervention group compared to 1% in the wait-list control group. Similarly, advising someone to seek professional help increased by 10% vs 1%. To detect this effect in an unclustered trial with 80% power at the 0.05 significance level, required n = 200. The average school was estimated to have 30 teachers, giving a cluster size of 30. The intra-class correlation (ICC) was unknown. Examining ICC values from .01 to .10, the number of required clusters varied from 10 to 28. A previous cluster randomized trial of MHFA in a rural area [5] found ICCs ranging from 0.002 to 0.15, with most < 0.05. We therefore assumed an ICC of 0.05, which required a minimum of 18 schools to be randomized. We managed to recruit 16 schools for the trial, 14 of which participated as randomized.

Randomization: sequence generation

The 16 schools were paired to be alike in socioeconomic characteristics. The pairing was carried out on the basis of: a scale of education disadvantage, size, location (metropolitan vs rural/remote), and gender (single vs mixed gender schools). Using the Random Integers option of Random.org, one school in each pair was randomly assigned to the immediate group and the other school to the wait-list group, by generating a 1 or a 2 for each pair (1 = immediate, 2 = wait-list).

Randomization: allocation concealment

Allocation was based on clusters rather than individuals, so that all teachers at a school received the same intervention. Schools were told about the allocation before their teachers completed the pre-test questionnaire. This was necessary so that they could schedule the staff training days.

Randomization: implementation

AFJ randomly assigned the schools. Participating schools were enrolled by a staff member of the Department of Education and Children's Services (HS) who informed them of their allocation after agreement to participate had been received.

Blinding

Blinding of participants was not possible. Post-test and follow-up questionnaires were self-completed by teachers who knew whether they had completed the training or not. Students were not informed about whether teachers at the school had received training, but no systematic attempt was made to blind them.

Statistical methods

The analysis of these multilevel or nested data required that the correlation of responses by individual participants between the measurement occasions and the correlation between participant responses within schools be taken into account. For that reason, mixed-effects models for continuous and dichotomous outcome variables, with group by measurement occasion interactions, were used to analyse the data. These maximum-likelihood based methods produce unbiased estimates when a proportion of the participants drop-out before the completion of the study, provided that they are missing at random [19, 20].

In the current study, all the participants included in the analyses completed the first questionnaire. Twenty-two percent of teachers did not complete the post-test questionnaire and 28% the follow-up questionnaire. In relation to the students, 24% did not complete the follow-up questionnaire.

All analyses were performed using Stata Release 10 [21].

Ethics

Ethical approval was given by the Youth and Women's Health Service Research Ethics Committee at the Women's and Children's Hospital.

Participant flow

Figure 1 shows the flow of participants at each stage of the trial. Sixteen schools agreed to be randomized. Because the schools had to timetable their teacher training days early in the school year, the randomization had to be carried out before the baseline questionnaires were administered. After randomization and before baseline questionnaires, two schools decided that they were unable to follow the allocation because of changes in timetabling constraints. They would have to either withdraw from the study or else would agree to do the training in the period that was not allocated to them. In the interests of maximizing school participation, it was agreed to swap the allocation for these two schools (one from intervention to control and the other from control to intervention), resulting in 14, rather than 16 schools receiving the intervention as randomized.

Numbers analysed

All participants who completed a pre-test questionnaire and were at one of the 14 schools that adhered to randomization were included in the analysis. However, a supplementary analysis was also carried out which included the 2 additional schools that did not adhere.

Participants' Characteristics

With the exception of a significantly larger proportion of intervention group teachers having less than 3 years teaching experience in schools (10.9% vs. 3.9%, χ ²⁽¹⁾ = 4.42 , P = 0.036), and a smaller proportion in leadership roles (17.4% vs. 27.2%, χ ²⁽¹⁾ = 4.16 , P = 0.041), the characteristics of teachers were similar between the intervention and control groups. In relation to the student sample, the only significant difference in characteristics was that intervention group students had a significantly larger proportion of year 7 students relative to the control group (3.2% vs. 1.2%, χ ²⁽¹⁾ = 6.29 , P = 0.012).

Teacher outcomes

At post-test, teachers who received training had greater gains in knowledge (mean difference = 2.08, P < 0.001) and these gains were maintained at follow-up (mean difference = 1.79, P < 0.001). The teachers who did two days of training showed greater gains in knowledge than those who did only one day, but the difference was not significant. Recognition of depression was high at pre-test and was not affected by the training. Beliefs about the effectiveness of different approaches became more consistent with those of mental health professionals at post-test (mean difference = 0.79, P = 0.006) and this change was maintained at follow-up (mean difference = 0.73, P = 0.013). A number of personal stigma items showed improvement in response to training. Trained teachers were less likely than untrained ones to see depression as due to personal weakness (OR = 3.07, P = 0.024 at post-test and OR = 2.47, P = 0.077 at follow-up) and they were also less likely to be reluctant to disclose depression to others (OR = 3.79, P = 0.012 at post-test and OR = 3.42, P = 0.029 at follow-up). Two of the perceived stigma items showed changes, with the trained teachers more likely than the untrained teachers to believe that other people see depression as due to personal weakness (OR = 1.10, P = 0.848 at post-test and OR = 3.01, P = 0.031 at follow-up) and the trained teachers more likely to see other people as reluctant to disclose (OR = 2.57, P = 0.041 at post-test and OR = 1.32, P = 0.555 at follow-up). Intentions towards helping students showed some greater gains in the trained group, with trained teachers more likely to say that they would discuss their concerns with another teacher (OR = 3.73, P = 0.013 at post-test, OR = 2.46, P = 0.094 at follow-up), discuss their concerns with a counsellor (OR = 3.87, P = 0.023 at post-test, OR = 2.98, P = 0.075 at follow-up) and have a conversation with the student (OR = 2.06, P = 0.162 at post-test, OR = 3.16, P = 0.032 at follow-up). Confidence in helping a student with a mental health problem also increased (OR = 8.09, P = 0.005 at post-test, OR = 7.02, P = 0.008 at follow-up), as did confidence in helping a work colleague (OR = 7.22, P = 0.005 at both post-test and OR = 11.65, P = 0.001 at follow-up). Teachers who were trained were more likely to agree with the following strategies to support a student with a mental health problem: review curriculum options/classroom practices (OR = 2.22, P = 0.071 at post-test, OR = 3.76, P = 0.004 at follow-up), review/change school policy (OR = 3.20, P = 0.029 at post-test, OR = 2.44, P = 0.108 at follow-up), and improve relationships within the school (OR = 3.09, P = 0.029 at post-test, OR = 3.26, P = 0.027 at follow-up). Finally, trained teachers were more likely to report that the school had a written policy to deal with students with mental health problems (OR = 4.57, P = 0.019 at post-test, OR = 7.28, P = 0.003 at follow-up) and that the policy had been implemented in the previous month (OR = 7.23, P = 0.070 at post-test, OR = 13.30, P = 0.028 at follow-up).

Contrary to the hypotheses, training did not affect helping behaviours of teachers towards either students or colleagues, teacher mental health or seeking of information about mental health problems.

Student outcomes

Contrary to the hypotheses, there was no difference in reported help received from teachers or in the students' mental health. A secondary analysis focussing just on students with worse mental health (above the cut-off on the Strengths and Difficulties Questionnaire) at baseline also did not support these hypotheses.

Adverse events

Given that this was an educational intervention with a non-clinical sample, there was no formal enquiry about adverse events. Informally, no adverse events were reported.