This parallel-group design followed the CONSORT guidelines  and checklist  for conducting RCTs. A computerised random number generator was used to allocate caregivers randomly, using block randomisation, 1:1 to intervention and wait-list control group. Allocation was concealed: it was done off-site, in Australia, and was undertaken by a researcher who was not directly involved in recruitment. Randomisation was then emailed to the researcher who allocated participants to intervention and control groups. After trial commencement, no change of original protocol occurred.
Participants and recruitment
Primary caregivers of patients attending a psychiatric outpatient department for treatment of depression (who had consented to take part in a separate RCT of GSH [24, 25]) participated. Caregivers were recruited through mental health clinicians at Suan Prung Psychiatric Hospital outpatient department in Chiang Mai City in northern Thailand. Clinicians provided brief to them information about the study; those who expressed interest in participating were referred to the researcher, who provided detailed information about participation.
(a) primary caregiver of an adult receiving outpatient department treatment for moderate depression ((F32.1), ICD-10 classification) , (b) aged 18–60 years, able to read and write in Thai, and (c) had a telephone at home.
A caregiver currently receiving treatment for an acute episode of mental illness.
Sample size and power
IBM® SPSS® SamplePower (Vers. 2.0) was used to calculate sample size, which showed that a standardised difference between both groups of 0.8 (considered an effect size of clinical significance) could be achieved with 80 % power with a sample of 52, with a type I error probability of 0.05. To allow for some attrition, the sample size was increased to 56.
The Family Questionnaire (FQ) , a 20-item self-report instrument assessing EE directed by, in this instance, a primary caregiver towards a family member with depression, was used. It comprises two sub-scales (critical comments, emotional over-involvement) rated on a 4-point Likert scale, ranging from 1 (never/very rarely) to 4 (very often). Critical comments are adverse statements about the behaviour or personality of the person with depression, while emotional over-involvement comprises over-protectiveness, over-intrusiveness and over-identification with the person, by the caregiver . Higher scores (maximum of 80) signify high levels of EE, which are associated with a greater likelihood of relapse; lower scores (minimum of 20) indicate low levels of EE (supportive comments and lack of over involvement), which are associated with a reduced risk of relapse . The FQ has been validated against the Camberwell Family Interview, which is regarded as the benchmark or 'gold standard' measure of EE . In the current study, Cronbach's alpha was used to assess internal reliability of the FQ at baseline, showing good reliability (α = 0.900).
As no pre-existing Thai version of the FQ existed, the WHO guidelines for translating instruments  was used to translate it into Thai. Research outcomes were assessed at baseline (Week 0), post-test (Week 8) and follow-up (Week 12).
The study was undertaken in caregiver participants' homes in several provinces in the upper northern part of Thailand, including Chiang Mai City. Control group participants received standard outpatient department support when they accompanied the family member with depression to the department for consultations and prescription of antidepressant, or a combination of antidepressant and anti-anxiety, medications. At the same time, the caregiver received limited information and support from mental health clinicians about how to support the family member with depression.
Intervention group participants were provided with a copy of The Good Mood Guide: A self-help manual for depression [31, 32], which had been translated into Thai. The manual was based on self-help and CBT principles and contained eight modules: (i) Summarises depression. Helps the person to assess depression and distress levels and promotes physical exercise. (ii) Emphasises the importance of maintaining social contact and physical activity. (iii) Assists carers to comprehend how people with depression feel and think and how they can recognise and label automatic thoughts and link situations and emotions to events in their lives. (iv) Shows how to change the person with depression’s thought patterns from negative to positive. (v) Highlights how healthy living, social support and problem-solving can help the person to change their behaviour and overcome their depression. (vi) Provides skills to improve the person with depression’s sleep pattern and how to maintain positive thoughts, emotions and behaviours. (vii) Shows carers how the person with depression can practise progressive muscle relaxation to deal with stress and enhance time management. (viii) Emphasises to carers how previously learned skills in changing behaviours, thought challenging and learning how to cope with difficult situations can assist the person with depression.
Caregivers were shown how to use the manual and requested to complete one module each week. On average, it took two hours to finish each module, which comprised reading and writing and some homework exercises. While the manual was developed initially as GSH for individuals with depression, caregivers could also use it.
Both groups were contacted by telephone each week, for approximately 5 min, by a researcher. The purpose of the call was to answer caregivers’ questions, provide limited support, and, for the GSH group only, to answer caregivers’ questions about using the manual.
Treatment fidelity was assessed during the weekly telephone calls by asking GSH group participants a prearranged set of questions about the content of the module completed that week.
Victoria University, Melbourne and the Mental Health Department, Public Health Ministry of Thailand, Bangkok human research ethics committees gave approval to conduct the study. All participants gave written, informed consent.
IBM® SPSS® Statistics Version 20.0 was used to analyse the data, by a researcher who was blind to group allocation. An insignificant amount (<0.5 %) of FQ item scores were recorded as missing. Descriptive statistics were used to analyse demographic data, with summary characteristics provided across the whole cohort and for both treatment groups. The balance of the groups at baseline was monitored for imbalances of any demographic characteristics across groups. Established procedures were followed whereby decisions on covariate adjustment in the case of baseline imbalance were made by prior knowledge of an influence on an outcome. A mixed analysis of variance (ANOVA) technique, incorporating between-participants and within-participants factors, was adopted to assess the effect of grouping, time and interaction effects. Effect sizes were evaluated using the partial eta-squared statistic. Intention-to-treat analyses were undertaken.