This study used both qualitative and quantitative methods to optimise the intervention, test feasibility and pilot the randomised controlled trial. The target population was women from disadvantaged areas, where the prevalence of depression is often highest [9
]. This trial was designed exclusively for women, as female only groups have shown better recruitment and lower dropout rates [10
] and, it was thought that mutual support might be achieved more effectively in a single sex group with a restricted age range. The recruitment focused on two disadvantaged areas of Bristol: an inner city area with large Black and minority ethnic populations and a more peripheral urban area with predominantly white residents.
Ethical approval was given by Frenchay Research Ethics Committee (07/H0107/60).
Study inclusion criteria were: women aged 30 to 55 with clinical depression according to the Patient Health Questionnaire (PHQ-9). Exclusion criteria were: severe depression (this criterion was later relaxed in the second round of groups); drug or alcohol abuse; currently attending specialist psychiatric services (including psychotherapy); or being unable to speak English. Current or past use of antidepressants or benzodiazepines was not an exclusion criterion.
Local GPs, mental health professionals and community organisations such as health training teams and domestic violence organisations were all invited to refer potential participants. Posters and fliers were distributed around community venues to encourage women to refer themselves. The information circulated did not mention the term 'depression', but asked women to get in touch if they felt 'stressed' or 'not able to cope'. Medical notes searches were also conducted in 5 GP practices to identify potential participants who had a diagnosis of depression or anxiety, were currently taking antidepressants, or with a recent PHQ-9 score between 5 and 24. Practices wrote to women identified asking them to contact the research team if interested in participating. The 5 practices involved had 30,383 registered patients, with individual practice list sizes ranging from 4,848 to 7,300.
Intensive recruitment was conducted over a restricted period of two months in each area to ensure that potential participants were not left waiting too long between being recruited and the groups starting. Comparison of the feasibility and efficiency of each of the recruitment methods was done by maintaining a record of all referral sources for all potentially interested people contacting the research office. Information about recruitment was also collected during follow up calls to recruiters and a limited number of interviews.
There was a three stage process of recruitment. After an expression of initial interest by a potential participant through one of the routes detailed above, further information about the study was given over the telephone. For those still interested after the telephone conversation, detailed written information and consent forms were sent and the person invited to attend an appointment with the researcher. At this face-to-face assessment, potential participants completed a PHQ-9. Those with a score of between 10 and 20 (first round of groups) or 10 and 24 (second round) and no reason for exclusion were considered eligible for the study. PHQ-9 scores of 5, 10, 15, and 20 represent mild, moderate, moderately severe, and severe depression, respectively [12
]. We increased the upper threshold for the PHQ-9 in the second round of groups since experience from the first recruitment round suggested that the lower threshold excluded some women who appeared suitable for the group intervention. All eligible participants were then asked to provide written consent, complete a baseline questionnaire, and to participate in a semi-structured qualitative interview. Details of the measures included in the baseline and follow-up questionnaires were as follows:
• Patient Health Questionnaire [13
• Anxiety symptoms: Beck Anxiety Inventory [14
• Health Status: Physical and Mental component scores from the Short Form-12 Questionnaire [15
• Dysfunctional thoughts: shortened Automatic Thoughts Questionnaire [16
• Social support: shortened Medical Outcome Study Social Support Survey [17
• Recent health service use
• Socio-demographic characteristics e.g. ethnic group, household income, housing tenure.
After baseline assessment, participants were randomised to the intervention or control arm.
Concealed allocation of eligible participants to either the intervention or usual care arm was achieved using an automated randomisation system accessed via the web. Minimisation was used to ensure balance in PHQ-9 score (≤15 vs. ≥ 16) and GP practice. The allocation ratio was 2:1 in favour of the intervention arm in order to ensure that groups were viable. Participants were randomised and informed of their treatment allocation at the end of baseline interview. The allocation was not concealed from either the participants or the researcher, although analysis was conducted blind to allocation.
Building on the findings from previous studies [10
] the intervention was based on principles derived from CBT and problem-solving approaches. Using homework and skill based approaches the group intervention sought to empower participants through improving their self-management skills. It is important to note that the course was designed to be used by non-professional facilitators following a manual. It should be considered as an attempt to design a low-cost group intervention based on CBT principles which could be made widely available rather than being equivalent to individual CBT from a trained therapist.
Facilitators and groups: A pair of trained facilitators delivered the intervention. There was a deliberate strategy to recruit facilitators who were not professional therapists in order to make the group intervention widely available, to ensure that facilitators were socially connected to the target population in deprived areas, and to keep the cost of the intervention low and therefore to maximise its potential cost-effectiveness. Facilitators were recruited via advertisement in local newspapers, community magazines and word-of-mouth. There were over 150 applicants. Ten (female) applicants were recruited. Prior to the study none of the facilitators had received CBT training but most had counselling skills, some had experience of group work and one had lived experience of mental ill health. Facilitators received a 5-day training course from experienced CBT trainers (JC and RA). Facilitators also continued to receive regular fortnightly supervision from the CBT trainer (JC). Four intervention groups were run in 2 rounds in 2 locations, with 1 morning and 1 evening group in each location.
The intervention consisted of 12 sessions delivered over 10 consecutive weeks with 2 booster sessions after a gap of 2 to 4 weeks at the end. The course content was based on previous courses and interventions [18
] and is shown in summary below:
1. Introduction and group rules
2. Checking activity levels
3. Raising activity levels
4. Catching negative thoughts
5. Balancing negative thoughts
6. Managing anxiety
8. New Ways of Solving Problems
9. Integrating and catch up session
10. Making a personal plan
11. Booster session I
12. Booster session II
All participants received a free manual which described the content of each session, along with exercises and space for notes. Considerable attention was given to ensuring the course manuals were clear, attractively laid-out and in a language acceptable to targeted users. Facilitators followed a similar course manual with instructions and suggested timings. Fidelity of the intervention was ensured by regular supervision and observation of one session for each of the groups. Two members of the research team (JC and HC) assessed the delivery and quality of the intervention by using a modified version of the CBT rating scale (CTSR), adapted to make it suitable for a group [23
]. Where consensus was not reached a third member of the team (CS) decided.
Participants in the control arm were given an information booklet. This contained details of local support organisations such as local mental health organisations, counselling services, carers groups and Black and minority ethnic services. An information booklet was given in addition to usual care because it was thought that it might improve study retention in the control arm. Participants in both arms were allowed to continue taking (or start) any antidepressant or other medication prescribed by their GP.
Data collection and outcome measures
Patients were assessed at baseline, and 3 and 6 months after starting the groups. The primary outcome measure was the PHQ-9. Secondary outcomes are listed above. All measures were used at baseline and after 3 and 6 months. Follow-up data were collected either during an appointment, by postal questionnaire or over the telephone.
Qualitative methods were used to support the design of the intervention and to help understand issues of acceptability and feasibility. Interviews were conducted with all (75) participants at baseline and at 3 and 6 months exploring motivation to join the study, own definitions of improvement and experiences of the groups. Subsequent sampling and detailed analysis was done with 20 participants from interviews at baseline, 3 and 6 months (14 in intervention and 6 in control arm at each of the 3 time points). Purposeful random sampling was done to ensure those selected represented a range of illness severity (based on PHQ-9 scores) and high, low or no attendance at groups. In addition, interviews with 2 GPs and 5 community staff were conducted and feedback collected from the facilitators during 2 focus groups and on a weekly basis. Interviews, focus groups and observations were all digitally recorded with consent and transcribed. The interviews were analysed thematically by the constant comparison method and by taking a Framework approach [24
]. Some data, such as interviews with recruiters, was organised in Atlas ti, data from participant interviews was organised using Excel spreadsheets and the Framework approach. HC led the analysis, although the credibility of the Framework categories was checked by other members of the team.
We used descriptive statistics such as means, standard deviations and proportions to describe the characteristics of patients allocated to the intervention and control arms at baseline. Three individuals who had been randomised to the control arm were invited to the intervention arm in order to bolster the numbers required to form a viable therapeutic group. These individuals were chosen at random and all agreed to this change. In the analysis these individuals were treated as members of the intervention arm (treatment allocated) because they received the intervention but we also performed sensitivity analyses in which we analysed these individuals according to their randomisation to account for this protocol alteration. Because this was a pilot study, not expected to provide definitive evidence about effectiveness, no power or sample size calculation was performed. Comparisons across groups were done using linear and logistic regression. The primary outcome variable (PHQ-9) was used as a continuous variable and also as a binary variable (proportion of patients 'improved' (50% score reduction) and 'recovered' (score <10)) to compare groups as allocated to treatment, with adjustment for PHQ-9 baseline scores. Similar regression models were used for secondary outcomes. All these analyses were repeated for primary and secondary outcome variables at 3 and 6 month follow-up periods. Secondary analyses compared participants according to the treatment actually received (on the basis of records of therapy sessions attended). These analyses of complier-average causal effect used instrumental variables in linear regression models for PHQ-9 at each follow-up as a continuous score. We also dichotomised attendance according to whether or not participants attended at least half (six) of the sessions. All statistical analysis was performed using STATA version 9.0.