Trial design
Our choice to use a mixed-methods trial design is guided by the principle of pragmatism,50 our overarching goal being to comprehensively evaluate different facets of TWM. While our study primarily focuses on the programme’s efficacy, as reflected in the primary research questions, we also seek to elaborate on these results and expand the study’s focus by investigating questions of a more exploratory type. Therefore, the two main purposes of our mixed-methods design are complementarity and expansion.51 In line with this rationale, we have chosen a sequential exploratory design (QUAN → qual), which encompasses three steps: (1) a quasi-experimental online survey, (2) focus groups, and (3) qualitative interviews with stakeholders. While step 1 pertains to the ‘QUAN’ phase, steps 2 and 3 form the ‘qual’ phase. The findings from the quantitative and qualitative phases will be connected in the intermediate stage of the study52, such that the findings of step 1 will influence both steps 2 and 3. A visual model depicting the trial design can be found in figure 1.
Visual model depicting the mixed-methods trial design. ANOVA: analysis of variance. ANCOVA: analysis of covariance.
In the online survey, we will collect quantitative data in three waves: preintervention, postintervention and at 6-month follow-up. As the intervention is being carried out in a different cultural context for the first time, our focus at this stage will be to evaluate the intervention as such rather than contrasting it with other active treatments. Consequently, we have chosen a quasi-experimental design with a passive control group.
In step 1, we will focus on answering the efficacy-focused research questions, that is, questions 1–5. Additionally, this step will help to identify ‘candidate mediators’ and mechanisms (of change)41 targeted in question 6.
The focus groups with intervention participants will take place approximately 1 month after the intervention. Participants will be stratified according to target group (students, employees and managers) and, if possible, gender (women, men; other gender identities will not be used for stratification purposes).
In step 2, we will focus on both elaborating and expanding the quantitative results obtained in step 1. The goal of the elaboration process will be to develop a deeper understanding of the quantitative data, for example, by obtaining a more fine-grained understanding of mechanisms of change. The goal of the expansion process will be to develop a broader understanding of the quantitative data, for example, by giving participants the opportunity to freely share their experiences with the programme (thus expanding the range of programme ‘outcomes’).
The qualitative interviews with stakeholders will take place after the first implementation round, that is, after the first 6-month follow-up assessment. Programme stakeholders will include people working in mental health and psychosocial services at the German universities in which the programme will be implemented. Further, representatives of the universities’ personnel department and the staff council will be contacted for an interview.
In step 3, our sole focus will be stakeholders’ perspectives on factors facilitating and hindering the sustainable implementation of the programme, that is, research question 7. Although the target group of this step is different from that of steps 1 and 2, the previously obtained results provide an important basis for answering question 7. Therefore, development51 may be considered the third function of our mixed-methods design.
Patient and public involvement
The programme’s target groups, that is, students, employees, and managers, have been involved in the entire research process. They were included in the development of the study design, and the study was designed based on their needs and priorities. The research questions regarding stigmatising attitudes, mental health literacy and openness were derived from panel discussions and expert interviews. They were also part of the process of culturally adapting the programme (including focus group discussions, translations, etc). Further, the target groups were involved in the adaptation process of two of the outcome measurement instruments used in this study (Opening Minds Scale–Workplace Attitudes (OMS-WA) and Intergroup Anxiety Scale (IAS); see the ‘Data collection’ section). Programme participants will also be involved in the choice of outcome measures in the qualitative parts of the study. As described above, they will be asked to reflect on the ways in which the programme may be beneficial to them, either personally (step 2) or as seen from their professional perspective (step 3). Moreover, they will be part of the recruitment of participants by forwarding invitations and presenting the study to their peers, and in a similar way, they will also be part of the dissemination, as leaflets and factsheets will be created that summarise the main findings of the study in lay language to be disseminated to the public. To ensure their continued involvement in the research, the study was connected to an internal advisory board at the university (consisting of representatives of all groups) that reflects on and discusses the progress of the study.
Sample size
We have calculated the sample size based on the anticipated mean effect for research question 1. An a priori power analysis for a 2×3 analysis of variance (ANOVA) (intervention vs control; preintervention, postintervention, and follow-up assessment) with repeated measures with a power of 0.80, α=0.05 and an anticipated medium effect size (d=0.38; see meta-analysis on TWM23; indicates a required total sample of n=50, which means approximately n=25 per group (ie, control and intervention groups) if the groups are equally populated. As we will carry out the intervention in different target groups, the required number of participants increases: For the three target groups, this results in a total planned sample of 25×2×3=150 people, of whom around a third are to be reached again for follow-up.
For the focus groups, participants of the intervention group will be stratified according to target group (students, employees, management) and, if possible, gender (male, female), thus ideally resulting in six groups of six people each (n=36).
Participant eligibility criteria
To be included in the study, participants have to be aged 18 years or older, to be members of a German higher education institution, and to belong to either of the intervention target groups: students, employees or managers. Given the focus of our intervention, managers are defined as having personnel responsibility. Individuals who do not meet these criteria will be excluded from the study.
Participant timeline
In the following, we will describe the different steps of the study procedure. Figure 2 provides a schematic overview of the schedule.
Schedule of participant acquisition, enrolment, interventions and assessments. IAS, Intergroup Anxiety Scale; IASMHS, Inventory of Attitudes to Seeking Mental Health Services; MHL-W-G, Mental Health Literacy Tool for the Workplace; OMS-WA, Opening Minds Scale–Workplace Attitudes; PHQ-9, Patient Health Questionnaire-9; SSMIS-SF, Self-Stigma of Mental Illness Scale–Short Form; SSOSH, Self-Stigma of Seeking Help; SSRPH, Stigma Scale for Receiving Psychological Help; VASI, Value-based Stigma Inventory.
Participant acquisition and enrolment
Since the intervention is aimed at students and staff of German universities and piloted at the University of Greifswald (a medium-sized town in the northeast of Germany), participants for the intervention group are currently being acquired via the university’s channels. The intervention has been included in the catalogue of health offers available to all members of the University of Greifswald. In addition, the intervention dates are being included in the university calendar, and flyers are being distributed across the campus. The employee and manager versions of the programme are also being advertised via central university mailing lists, whereas the student version will be advertised via various student Instagram channels. University members interested in taking part can enrol via the catalogue. Control group participants will be acquired via other German universities’ channels, for example, mailing lists.
Assessments and interventions
Participants of both groups will be invited to a quantitative preintervention assessment (t1) a few days prior to the intervention. As described above, the student and employee versions of the programme will take place on 1 day, whereas the manager version of the intervention will be spread over 2 days (max. 2 weeks apart). At the end of the intervention, only the intervention participants will be informed about the focus groups and asked to indicate whether they consent to the research team contacting them about participating. Quantitative postintervention assessment (t2QUAN) will take place shortly after the intervention for both groups; qualitative postintervention assessment (t2QUAL) will take place approximately a month after the intervention. Finally, a quantitative follow-up assessment (t3) will take place around 6 months postintervention for both groups. The study, that is, first participant enrolment, started on 1 February 2024, after trial registration (31 January 2024) and submission of the first version of this protocol (31 January 2024). The study is planned to end on 30 June 2025.
Intervention
TWM23 is a Canadian group-based intervention with the primary aims of reducing stigmatisation of mental illnesses and people affected by these illnesses and promoting help-seeking. Its secondary aim is to strengthen positive mental health by providing adaptive coping strategies and relevant help options so that stress can be dealt with more effectively. As a result, mental illnesses may develop less frequently and take a less severe course. The programme is based on a combination of psychoeducational, contact-based, and coping-oriented elements and thus reflects the current state of research (see the ‘Introduction’ section). As described above, there are different versions of the intervention. Given the target groups of our study, we will implement the culturally adapted student, employee, and manager programme versions. Table 1 lists the programme modules for each target group.
The Working Mind (TWM) programme modules for the different target groups
The information will be delivered using a PowerPoint presentation along with a facilitator guide. Furthermore, participants will receive a handout and additional print information on specific topics (eg, a resilience guide). All the original materials have been translated into German and adapted culturally. Intervention facilitators have a professional background in mental health and have successfully completed a week-long facilitator training for the intervention.
The intervention will be provided in small face-to-face groups of up to 15 people. The intervention will take place on the premises of the University of Greifswald; a first round is planned from February 2024 onwards.
Each module is around 1 hour long, resulting in a total duration of around 4 hours. The programmes for students and employees are carried out in 1 day while the programme for managers is divided into two sessions of 4 hours each.
Measures
In the following, we will provide an overview of the constructs to be examined in our study. All the constructs listed below can also be found in figure 2.
Outcomes
The study’s primary outcomes, examined in research questions 1–3, encompass mental illness stigma (H1), openness to mental health problems (H2), and willingness to seek help/utilisation of support offers (H3).
Secondary outcomes relate to research questions 4 and 6. Research question 4 comprises positive mental health outcomes: resilience (H4) and subjective well-being (H5). Research question 6 comprises possible intermediary outcomes (ie, mechanisms of change): mental health literacy (H8), mental health continuum beliefs (H9), empathy towards people with mental illness (H10), and intergroup anxiety towards people with mental illness (H11).
We will aggregate individual participant data by calculating mean values. Due to our aim to measure changes in outcomes, we will examine the variables listed above:preintervention (quantitative), postintervention (quantitative and qualitative), and at 6-month follow-up (quantitative). The only exception will be the utilisation of support offers, which we will only collect at follow-up.
Covariates
Several measures will be collected and examined as potential moderators of programme outcomes. As stated in research question 5, we will examine gender (H6) and personal values (H7) as potential moderators. Further, we will collect other sociodemographic and contextual variables identified as relevant factors of influence in the literature15–17 19 23 26 53: age, education, professional situation, field of study/work, current level of mental distress, as well as direct and indirect experience with mental illness and treatment options.
Again, we will aggregate individual participant data by calculating mean values. All covariates will be collected preintervention. Mental distress will additionally be measured at quantitative postintervention and 6-month follow-up, personal values will additionally be measured at 6-month follow-up. Gender, personal values, and experience with mental illness (treatment) will additionally be collected at qualitative postintervention. An overview can be found in figure 2.
Data collection methods
Quantitative data will be collected via online questionnaires, using the online platform SoSci Survey. Qualitative data will be collected in focus groups and interviews using semistructured interview guidelines. As the findings of step 1 will inform the subsequent steps of data collection, the interview guidelines will be finalised after analysis of these data. Both the online questionnaire and the interview guidelines will be uploaded to the OSF (https://osf.io/qrjce/?view_only=562269481229499c9467d750b7021e4c). The interview guidelines can also be found in online supplemental files 1,2. Student research assistants supporting both quantitative and qualitative assessment will be trained for the different tasks involved. Participants in the intervention group will receive a voucher worth €10 for each completed questionnaire and €30 for taking part in a focus group to increase participation rates.
Supplemental material
Supplemental material
In the following, we will briefly describe the psychometric features of the quantitative measurement instruments used. If not stated otherwise, we will calculate the sum scores for each of the instruments.
Primary outcomes (questions 1–3)
(H1) Mental illness stigma will be examined with various questionnaires to adequately capture the construct’s multifaceted nature.
Public mental illness stigma in the workplace/study context will be assessed with the OMS-WA.23 25 26 54 55 The OMS-WA exists in different versions tailored to different contexts; given our target groups, we will use the student,26 employee23 25 55, and manager version.54 While the student version measures stigmatising attitudes related to the study context (and the phase of life associated with it), the employee and manager versions measure stigmatising attitudes in the workplace. The student version comprises 23 items, the employee version 22, and the manager version 11, such as ‘people/employees with a mental illness could snap out of it if they wanted to.’ Participants are asked to rate the items on a 5-point Likert scale from 1=‘strongly disagree’ to 5=‘strongly agree’. Since the OMS-WA is central to our study and the versions we will use have not yet been translated into German, we translated them using a forward-back translation procedure, consulting an expert panel and pretesting them with the target group. Internal consistency is good to excellent for the English student version (0.88≤α≤0.92)26 and excellent for the employee version (α=0.90).54 Validation of the manager version is reported to be in progress.54 A comprehensive validation study on the employee version is currently under review56; the student version will be validated subsequently.
Public mental illness stigma (context-unspecific) will be assessed with the German version of the Self-Stigma of Mental Illness Scale–Short Form57 58 Agreement subscale. Whereas the scale as a whole was developed to measure self-stigma in people living with a mental illness, the Agreement subscale captures public stigma (which represents a necessary component of self-stigma). Participants are asked to rate the five items, such as ‘I think most persons with mental illness are dangerous.’ on a 5-point Likert scale from 1=‘strongly disagree’ to 5=‘strongly agree’. Studies across multiple clinical samples and languages indicate good construct validity of the scale; internal consistencies for the Agreement subscale are acceptable (72≤α≤0.7957). The use of the subscale for a general workplace/student sample will be discussed in the context of the study’s limitations.
In addition, we will examine prognostic pessimism, that is, the view that people with mental illness are unlikely to recover, which represents another important component of mental illness stigma.59 Prognostic pessimism will be assessed with a single item measure based on the one used by Lebowitz and Ahn.60 It asks participants to rate, on a 7-point Likert scale from 1=‘not at all permanent’ to 7=‘very permanent’, how permanent they consider a mental illness to be.
Value-sensitive mental illness stigma will be assessed with the VASI.38 Participants are asked to rate the 15 items, such as ‘It damages my reputation if a mental illness becomes known in my family.’ on a 5-point Likert scale from 1=‘strongly disagree’ to 5=‘strongly agree’. The items form five subscales: Self-Realisation, Personal Enrichment, Reputation, Meritocratic Values, Security. Rieckhof et al
38 report good internal consistency (α=0.88) and good convergent and construct validity of their questionnaire.
Public stigma of seeking help will be assessed with the German version of the Stigma Scale for Receiving Psychological Help.61 Participants are asked to rate the five items, such as ‘Seeing a psychologist for emotional or interpersonal problems carries social stigma.’ on a 4-point Likert scale from 0=‘strongly disagree’ to 3=‘strongly agree’. Internal consistency for the scale is good (α=0.81)61 and the original study indicated good construct validity62; however, information on its validity remains sparse.
Self-stigma of seeking help will be assessed with the German version of the Self-Stigma of Seeking Help scale.61 Participants are asked to rate the 10 items, such as ‘I would feel worse about myself if I could not solve my own problems.’ on a 5-point Likert scale from 1=‘strongly disagree’ to 5 ‘strongly agree’. Internal consistency for the scale is acceptable to good (0.80≤α≤0.84).61 63 Information on the validity of the German version is not yet available, but the original version has proven valid in terms of construct, criterion, and predictive validity.64
(H2) Openness towards mental health problems will be assessed with the German version of the Inventory of Attitudes to Seeking Mental Health Services65 Psychological Openness subscale: Participants are asked to rate the eight items, such as ‘There are certain problems which should not be discussed outside of one’s immediate family.’ on a 5-point Likert scale from 0=‘disagree’ to 4=‘agree’. Internal consistency for the Psychological Openness subscale is acceptable (α=0.70).61 Convergent validity for the original scale was demonstrated by Mackenzie et al.66
(H3) Willingness to seek help will be assessed by asking participants to rate how likely it would be for them to use different support offers, using a 6-point Likert scale from 1=‘not at all likely’ to 7=‘very likely’. They will be provided with a list of 8–10 (depending on the target group) support offers discussed during the intervention and asked to rate the likelihood for each of the offers.
Utilisation of support offers will be assessed by asking participants if they have used one or more support offers (0=‘no’, 1=‘yes’). Participants will be able to choose from the options provided when assessing their willingness to seek support, and they will also be able to state any other support offers they have sought out in a free text box.
Secondary outcomes
(H4) Resilience will be assessed with the German version of the Brief Resilience Scale (BRS)67 Participants are asked to rate the six items, such as ‘I tend to bounce back quickly after hard times.’ on a 5-point Likert scale from 1=‘strongly disagree’ to 5=‘strongly agree’. Internal consistency for the scale is good (α=0.85).67 The BRS is moderately correlated with optimism and social support, indicating convergent validity.67
(H5) Subjective well-being will be assessed with the German version of the WHO-5 Well-being Index (WHO-5).68 Participants are asked to rate five items, such as ‘I have felt calm and relaxed’. Participants are instructed to refer to the last 2 weeks, answering the items using a 6-point Likert scale from 0=‘all of the time’ to 5=‘at no time’. Internal consistency for the scale is good to excellent (0.89≤α=0.92).68 The WHO-5 is moderately correlated with somatic well-being, indicating convergent validity.68
(H8) Mental health literacy will be assessed using the Mental Health Literacy Scale (MHLS)69 for students and the German version of the Mental Health Literacy Tool for the Workplace (MHL-W-G)70 for employees and managers. As regards the MHLS, there is currently no validated German translation available; therefore, we will use the German translation developed and piloted in an unpublished bachelor’s thesis.71 Depending on the item, participants are asked to rate the 35 items on either a 4-point Likert scale from 1=very unlikely/unhelpful to 4=‘very likely/helpful’ or a 5-point Likert scale from 1=‘strongly disagree/definitely unwilling’ to 5=‘strongly agree/definitely willing’. They are, for instance, asked to rate statements such as ‘To what extent do you think it is likely that the diagnosis of Drug Dependence includes physical and psychological tolerance of the drug (ie, require more of the drug to get the same effect)?’. Internal consistency for the English version of the scale is good (α=0.83), and correlations with help-seeking intentions—although in the low range—point to construct validity.69 The MHL-W-G assesses workplace-related mental health literacy based on four vignettes. Participants are asked to rate their competence on the basis of four items per vignette, that is, 16 items in total, using a 5-point Likert scale from 1=‘strongly disagree’ to 5=‘strongly agree’. Internal consistency of the scale is good to excellent (0.88≤α≤0.92).67 The scale is moderately correlated with general health literacy and related variables; therefore, construct validity can be assumed.70
(H9) Continuum beliefs will be assessed with the newly developed (German) Continuum Beliefs Scale.72 Participants are asked to rate the nine items, such as ‘Now and again most of us have symptoms of a mental illness.’ on a 5-point Likert scale from 1=‘strongly disagree’ to 5=‘strongly agree’. As the scale has been developed as part of an ongoing project,72 psychometric validation is still outstanding.
(H10) Empathy will be assessed with a three-item scale used in previous stigma research73 74 which we have translated into German. Participants are asked to rate items such as ‘If a person with a mental health problem I knew was feeling sad, I think that I would also feel sad.’ on a 7-point Likert scale from 1=‘strongly disagree’ to 7=‘strongly agree’. Internal consistency of the empathy scale is good (ω=0.87). The scale has been shown to correlate with behavioural facets of stigma,73 which could be interpreted as a first hint to criterion validity.
(H11) Intergroup anxiety will be assessed with the IAS.49 75 The IAS instructs participants to imagine a situation in which they are to interact with people with a mental illness, and where they themselves are the only person without a mental illness. Subsequently, they are asked to rate to what extent they experienced certain emotions compared with a situation in which they interacted with people without mental illness. As regards the specific items, we will rely on Potts et al,49 whose version we have translated into German (for a detailed description of the translation process, see OMS-WA): Participants will be presented with a list of 13 adjectives (ie, items) to be rated on a 5-point Likert scale from 0=‘not at all’ to 4=‘extremely’. Positive emotions will then be reverse scored, and all items will be averaged. According to Potts et al,49 the scale shows good internal consistency (α=0.86) and associations with related constructs such as intergroup contact, which they interpret as evidence for construct validity. Individuals reporting personal experience with mental illness treatment and elevated levels of mental distress will not be asked about intergroup anxiety.
Assessment of potential covariates
(H6) Gender will be assessed by asking participants which gender they identify with, the response options being 1=‘female’, 2=‘male’, 3=‘non-binary’, 4=‘gender-fluid’, 5=‘agender’. Alternatively, participants can also use a text box to fill in the gender that they identify with.
(H7) Personal values will be assessed with the Portrait Values Questionnaire 21 (PVQ21, ESS21)66, adapted in a gender-neutral form.38 There is some evidence76 77 that partial scalar invariance of this short form exists for seven of the ten values postulated by Schwartz35 in his Theory of Human Values. Higher order measures of values (self-transcendence, self-enhancement, conservation, and openness to change) have shown more acceptable model fits.77 Personal values measured with PVQ21 have been shown to be stable over a 3-year period, similar to personal traits.78 The PVQ21 consists of 21 items presented in the form of short verbal portraits that describe a person’s goals, aspirations, or desires that point explicitly to each value, for example: ‘It’s important to them to be rich’. The respondents are asked to rate the similarity on a scale from 1=‘very similar’ to 6 = ‘not similar at all’.
Age, education, professional situation, and field of study or work will each be assessed with single items. Age will be assessed by asking participants to indicate their age in years. Education will be assessed by asking participants to indicate their highest school-leaving certification, with the response options ranging from 1=‘none’ to 11=‘doctorate’, including a free text box for certifications/qualifications not listed (not all response options listed as they are specific to the German education system). Professional situation will be assessed by asking employees and managers to indicate their current professional situation, with various nominally scaled response options provided: 1=‘full-time employed’, 2=‘part-time employed’, 3=‘partially retired’, 4=‘marginally employed’, 5=‘occasionally employed’, 6=‘in vocational training/an apprenticeship’, 7=‘in voluntary service’, 8=‘on parental leave/leave of absence’, including a free text box. Students will be asked whether they pursue gainful employment alongside their studies, with the following nominally scaled response options provided: 1=‘no’, 2=‘yes, full-time employed’, 3=‘yes, part-time employed’, 4=‘yes, self-employed’, 5=‘yes, marginally employed’. Field of work will be assessed by asking employees to indicate their field of work, with the following response options provided: 1=‘administration’, 2=‘academic staff’, 3=‘other’. Students will be asked to indicate their field of study by choosing one of the following response options: 1=‘social sciences’, 2=‘arts’, 3=‘teacher training’, 4=‘mathematics, sciences’, 5=‘medicine, health sciences’, 6=‘humanities’, 7=‘economics, law’.
Mental distress will be assessed with the German version of the Patient Health Questionnaire-9 (PHQ-9),79 a screening instrument for depression. Participants are asked to rate nine items describing different symptoms of depression, such as ‘Feeling down, depressed or hopeless’ on a 4-point Likert scale with the response options 1=‘not at all’, 2=‘several days’, 3=‘more than half the days’, 4=‘nearly every day’. The internal consistency of the scale is good (α = 0.89). The PHQ-9 has been shown to negatively correlate with various indicators of positive mental health, which can be considered an indicator for convergent validity.80
Experience with mental illness (treatment) will be assessed with four questions capturing different types of experience. Participants are asked (1) whether they are currently in treatment for mental illness, (2) whether they have been in treatment for mental illness at some point in their life, (3) whether someone from their immediate social environment has been in treatment for mental illness, and (4) whether they have already worked with someone with a mental illness. The items are to be answered in a yes-no response format (0=‘no’, 1=‘yes’); for questions 3 and 4, the option 2 = ‘I don’t know’ is also provided. Participants who answer questions one and/or two with ‘yes’ will additionally be asked about the type of treatment they have sought, with the response options (multiple answers possible) being 1=‘medical treatment (psychiatric treatment, eg, psychotropic drugs’), 2=‘psychotherapeutic treatment (eg, talk therapy)’, 3=‘art, music and/or sports therapy’, 4=‘self-help group’, 5=‘counselling and/or coaching services (eg, educational counselling, life counselling)’, 6=‘online therapy and/or telephone therapy’.
Data management
Participants will enter the quantitative survey data via the platform SoSci Survey. SoSci Survey works with SSL encryption (HTTPS) of the data when filling out the questionnaire and when retrieving the collected data. These data will then be stored pseudonymously on the file servers of the University of Greifswald. Plausibility checks of the data (eg, range checks) will be carried out before data analysis. The statistical analysis of the quantitative data will be carried out using the software R. After collection of the 6-month follow-up data, the quantitative data will be completely anonymised.
The qualitative focus group and interview data will be audiotaped, stored on the file servers of the University of Greifswald and transcribed manually. The data will be anonymised in the transcription process, and the audio files will be deleted once the transcription is completed. The qualitative data will be analysed using MAXQDA.
The data collected will be compiled in the data processing facility and stored on the secured server of University of Greifswald’s data centre. The anonymised data will be stored beyond the end of the project, with the sole purpose of evaluating and reporting them in scientific publications. Once analysed, the anonymised dataset, aggregated at the group level, will be made accessible via the OSF.
Data analysis
The quantitative data collected in step 1 will serve to answer research questions 1–6. Hypotheses H1–5 will be analysed with a 2×3 ANOVA with repeated measures to take into account the two groups (intervention vs waitlist) and the three time points of quantitative assessment. Given a significant main effect, moderation hypotheses H6 will be tested using a 2×3×2 ANOVA with group, time point, and gender as predictors, testing the interaction effect of all three predictors. H7 will be tested using a 2×2 ANCOVA with group, time and personal values (continuous) as predictors. If the assumptions of normality and heterogeneity cannot be met, an ordinal logistic regression will be calculated. Mediation hypotheses H8–11 will be tested using parallel multiple mediation analyses.
To minimise the risk of missing values, participants are compensated for taking part in the survey. Further, the survey is set up so that participants are reminded whenever they have left any items unanswered. Before proceeding to the next question, they are asked to confirm they wish to proceed without answering said item(s). By reducing inadvertent omissions, this strategy may particularly reduce one mechanism under which missing values occur: values missing completely at random. In order to manage data missing at random, attrition will be analysed, examining selective dropout across the three time points. This procedure allows for the identification of potential auxiliary variables (eg, mental distress), which may then enable the use of the full information maximum likelihood method, provided that an appropriate analytical model can be specified with the available data.81 Otherwise, alternative methods such as multiple imputation, using mixed-effect models, will be considered.81
The qualitative data collected in the focus groups (step 2) and the stakeholder interviews (step 3) will be analysed using qualitative content analysis, according to Kuckartz.82 A combination of inductive and deductive approaches to analysis will be used.83 More specifically, we will follow Kuckartz’82 content-structuring approach, which encompasses seven steps: (1) initial textual work, including writing memos and case summaries; (2) deductive development of main categories based on the focus group or interview guideline; (3) coding the data based on the main categories; (4) inductive development of subcategories; (5) coding the data based on the subcategories; (6) analysing the data along the main and subcategories, including associations between categories and (7) putting into writing the findings and documenting the process. The steps described may be repeated over several cycles, thus resulting in an iterative process of data analysis.81 We will analyse the fully transcribed focus group or interview transcripts, using the group as the unit of analysis for the focus groups. As regards the coding units, no minimum or maximum length will be specified since a coding unit equals a unit of meaning in Kuckartz’ qualitative content analysis.82 The transcripts will be coded and analysed by two researchers (ES and EN) as well as trained student assistants, such that each transcript will be coded by two coders, at least one of them ES or EN. To ensure a shared understanding of both the research questions and the method, we will organise a dedicated meeting with all German team members (ES, ST, EN and the student assistants) prior to starting qualitative data analysis. ES and EN both hold a postgraduate degree in psychology; the research assistants are undergraduate students of psychology. ES is a licensed psychotherapist and has extensive expertise regarding the systematic analysis of interview data and qualitative content analysis; EN is a psychologist and has no prior experience with qualitative content analysis but has familiarised herself thoroughly with the method on the basis of the relevant literature and under supervision of ST, psychologist, who has led and conducted several qualitative research projects.
The findings from the quantitative and qualitative phases will be integrated at the data interpretation stage52 (see figure 1).
This post was originally published on https://bmjopen.bmj.com