Peer-provided Problem Management Plus (PM+) for adult Syrian refugees: a pilot randomised controlled trial on effectiveness and cost-effectiveness

Aims Common mental disorders are highly prevalent among Syrian refugees. Problem Management Plus (PM+) is a brief, transdiagnostic, non-specialist helper delivered, psychological intervention targeting psychological distress. This single-blind pilot randomised controlled trial (RCT) on PM+ delivered by peer-refugees examined trial procedures in advance of a definitive RCT, evaluated PM+ 's acceptability and feasibility, and investigated its likely effectiveness and cost-effectiveness among Syrian refugees in the Netherlands. Methods Adult Syrian refugees (N = 60) with elevated psychological distress (Kessler Psychological Distress Scale (K10) score >15) and reduced pychosocial functioning (WHO Disability Assessment Schedule 2.0 (WHODAS) score >16) were randomised into PM+ in addition to care as usual (CAU) (PM+/CAU; n = 30) or CAU alone (n = 30). Primary outcomes were symptoms of depression and anxiety (Hopkins Symptom Checklist; HSCL-25) at 3-month follow-up. Secondary outcomes were pychosocial functioning (WHO Disability Assessment Schedule; WHODAS 2.0), symptoms of posttraumatic stress disorder (PTSD) (PTSD Checklist for DSM 5; PCL-5) and self-identified problems (Psychological Outcomes Profiles; PSYCHLOPS). Changes in service utilisation and time out of employment and/or adult education were estimated (adapted version of the Client Service Receipt Inventory; CSRI). Semi-structured interviews on the implementation of PM+ were conducted with stakeholders (i.e. six PM+ participants, five non-specialist helpers and five key informants). Results Recruitment, randomization and blinding procedures were successful. PM+ was generally perceived positively by stakeholders, especially regarding the intervention strategies, accommodation of the intervention and the helpers. Two serious adverse events not attributable to the trial were reported. At 3-month follow-up, the HSCL-25 total score was significantly lower for the PM+/CAU group (n = 30) than CAU group (n = 30) (p = 0.004; d = 0.58). Significant differences in favour of PM+/CAU were also found for WHODAS psychosocial functioning (p = 0.009, d = 0.73), PCL-5 symptoms of PTSD (p = 0.006, d = 0.66) and PSYCHLOPS self-identified problems (p = 0.005, d = 0.81). There were no significant differences in mean health service costs (p = 0.191) and the mean costs of lost productive time (p = 0.141). This suggests PM+ may potentially be cost-effective with an incremental cost from a health system perspective of €5047 (95% CI €0–€19 773) per additional recovery achieved. Conclusions Trial procedures and PM+ delivered by non-specialist peer-refugee helpers seemed acceptable, feasible and safe. Analyses indicate that PM+ may be effective in improving mental health outcomes and psychosocial functioning, and potentially cost-effective. These results support the development of a definitive RCT with a larger sample of refugees and a longer follow-up period.


Introduction
Recent years have witnessed a dramatic increase in the number of asylum seekers and refugees worldwide, largely accounted for by refugees fleeing from the war in Syria to its neighbouring countries and Europe (UNHCR, 2018). Common mental disorders are prevalent in refugees, with estimated prevalence rates of 30.8% for depression and 30.6% for posttraumatic stress disorder (PTSD) (Steel et al., 2009). Recent studies among Syrian refugees specifically also report elevated levels of distress (Tinghög et al., 2017;Poole et al., 2018).
Refugees are individuals who are 'unable or unwilling to return to their country of origin owing to a well-founded fear of being persecuted' (UN General Assembly, 1951). The steep increase in refugees carries significant public health implications (Priebe et al., 2016). Since the outbreak of the Syrian war, there has been an increase in studies evaluating mental health and psychosocial support programmes for Syrian refugees in Europe (e.g. Lehnung et al., 2017) and the Middle East (e.g. Weinstein et al., 2016). Meta-analytic evidence supports cognitive behavioural therapy (CBT) and narrative exposure therapy to treat PTSD in refugees (Nosè et al., 2017;Turrini et al., 2017). Although high-income countries such as the Netherlands have specialised mental health staff and programmes available for refugees, the treatment gap is large. Studies among refugees/migrants found that a large proportion did not receive adequate mental health care (Lamkaddem et al., 2014;Priebe et al., 2016). Access to specialist mental health care is hampered by various barriers such as long waitlists, communication difficulties and stigma (Satinsky et al., 2019).
To overcome barriers to mental health care for communities affected by adversity, the World Health Organization (WHO) developed Problem Management Plus (PM+). PM+ is a brief, transdiagnostic psychological intervention targeting symptoms of depression, anxiety and distress, and is based on CBT and problem-solving therapy strategies (Dawson et al., 2015). The intervention comprises five face-to-face sessions with a nonspecialist helper (Dawson et al., 2015). Randomised controlled trials (RCTs) on individual PM+ in violence-affected communities in Pakistan and Kenya showed that participants who received individual PM+ had fewer symptoms of depression, anxiety and PTSD, higher levels of functioning and fewer self-identified problems (Rahman et al., 2016;Bryant et al., 2017). Another RCT on group PM+ in distressed females in the Swat area in Pakistan showed similar results (Rahman et al., 2019).
In this study, we conducted a pilot RCT to (1) test trial procedures in advance of the definitive RCT; (2) evaluate acceptability and feasibility; and (3) gain a preliminary understanding about the likely effectiveness and cost-effectiveness of PM+ among Syrian refugees with elevated levels of psychological distress in the Netherlands.

Setting
The study was carried out by the Vrije Universiteit Amsterdam (VU) in collaboration with Stichting Nieuw Thuis Rotterdam (SNTR), a non-governmental organization (NGO) providing support with integration, including housing, Dutch language courses and guidance to work to approximately 200 Syrian refugee families with resident status in Rotterdam. The trial was approved by the Research Ethics Review Committee at VU Medical Center, the Netherlands (Protocol ID: NL61361.029.17, 7 September 2017) and prospectively registered online (https:// www.trialregister.nl/trial/6665).

Study design and participants
A single-blind pilot RCT using mixed-methods was conducted from April 2018 to May 2019. The study is part of the larger EU H2020-funded STRENGTHS project, which aims to scale-up brief, psychological interventions among Syrian refugees in Europe and the Middle East (Sijbrandij et al., 2017). CONSORT and CHEERS reporting checklists (Husereau et al., 2013;Eldridge et al., 2016) are appended (Checklist A1 and A2).
Adult Arabic-speaking Syrian refugees (18 years or above) were recruited during language classes and individual home visits by SNTR staff. Inclusion criteria were elevated levels of psychological distress as indicated by a score >15 on the Kessler Psychological Distress Scale (K10) (Kessler et al., 2002) and impaired daily functioning, indicated by a score >16 on the WHO Disability Assessment Schedule 2.0 (WHODAS 2.0;Ustun et al., 2010). Exclusion criteria were acute medical conditions, imminent suicide risk (assessed with the PM+ manual suicidal thoughts interview) (WHO, 2016), expressed acute needs or protection risks, indications of severe mental disorders (e.g. psychotic disorders) or cognitive impairment (e.g. severe intellectual disability; assessed by the PM+ manual observation checklist) (WHO, 2016).

Procedures
Oral and written informed consent (IC) was obtained from all participants before screening. Included participants completed baseline assessment questionnaires and were randomised into PM+/care as usual (CAU) or CAU alone by an independent researcher not involved in the study. CAU comprises all other mental health services available to Syrian refugees in the Netherlands (e.g. community services and non-directive counselling by local NGOs, or referral to specialised PTSD treatment such as narrative exposure therapy). Randomization with block size 6 was performed using software on a 1:1 basis. Participants were phoned on group allocation by an Arabic-speaking team member not involved in assessments. Participants randomised to PM+/CAU were directly contacted by their helper to plan the first session within 1 week after baseline assessment.
Post-and 3-month follow-up assessments were scheduled 1 week after the 5th PM+ session (or 6 weeks after baseline) and 3 months after the 5th PM+ session. Assessments were carried out by two Arabic-speaking assessors who received a 3-day training on questionnaire administration, general interview techniques, common mental disorders, psychological first aid and ethical research conduct. Assessors were blinded to condition and indicated after each post-and follow-up assessment whether group allocation was disclosed to them (i.e. yes PM+/CAU; yes CAU; no).

Problem Management Plus (PM+)
The PM+ intervention has five 90 min sessions, delivered weekly. Four components are introduced by the helper, including a slow breathing exercise, problem-solving strategy, behavioural activation through re-engaging with pleasant and task-oriented activities, and accessing social support. A detailed description of PM+ is available (Dawson et al., 2015). The PM+ manual was translated/culturally adapted for use among Syrian refugees through qualitative study (cf. Applied Mental Health Research Group, 2013) and by using a framework for the cultural adaptation of psychological interventions (Bernal and Sáez-Santiago, 2006;see de Graaff et al., 2020).
The intervention was delivered by eight Arabic-speaking Syrian non-specialist helpers already working as 'connectors' at SNTR. They received 8 days of training followed by weekly face-to-face group supervision by PM+ trainers/supervisors throughout the trial. Training involved education about common mental disorders, 2 A. M. de Graaff et al.
basic counselling skills, delivery of intervention strategies and selfcare (WHO, 2016). Supervision included discussion of individual cases and difficulties experienced by helpers, practice of skills and self-care (WHO, 2016). Helpers had at least high school education, a background in social work, teaching or another related field, and sufficient Dutch or English speaking ability. Trainers/ supervisors were mental health care professionals who underwent 5-day training covering elements of the training of helpers, as well as training and supervision skills (cf. Rahman et al., 2016). A 25% random sample of the audio recordings was independently coded by two research assistants for adequate delivery of PM+ treatment elements (yes/no) through a checklist addressing requisite PM+ components per session (see appended Checklist A3).

Secondary outcome measures
Functional impairment. The WHODAS 2.0 is a well-accepted, validated 12-item instrument to assess health and disability (Ustun et al., 2010). Items are rated on a 1-5 scale (range 12-60).
Self-identified problems. The Psychological Outcomes Profiles (PSYCHLOPS) (Ashworth et al., 2004) questionnaire is a patientgenerated indicator of change after therapy. It covers self-identified problems and function (both free-text fields), and wellbeing scored on a 0-5 scale (range 0-20).
Client Service Receipt Inventory (CSRI). The CSRI (Beecham and Knapp, 1992) was modified for use in Syrian refugees in the Netherlands to self-report health service utilization, receipt of informal family care and participation in employment/ education or other productive use of time over 3 months.

Other measures
Trauma exposure. Life-time traumatic experiences were measured through a 27-item checklist (Schick et al., 2016) adapted for this project. Items were scored 1 (yes) or 0 (no), total range 0-27.
Post-migration stressors. The Post-Migration Living Difficulties checklist (Arabic version) (Silove et al., 1997;Schick et al., 2016) assesses 17 post-migration challenges scored on a 0-4 scale. Items with at least a score of 2 (moderately serious problem) were considered positive responses and summed for analysis (range 0-17).
These measures were used for descriptive analyses.

Quantitative and qualitative analyses
To compare reductions in primary/secondary outcomes between groups across three time points in the intention-to-treat sample (N = 60), we used linear mixed models in RStudio version 3.6.1 (R Core Team, 2018). This method allows the number of observations to vary between participants and handles missing outcome data. The mixed model uses a longitudinal data structure that includes both fixed and random effects. Time (categorical), group (PM+/CAU v. CAU) and interactions between group and time were included as fixed effects in mixed models together with a random intercept and random time effect. Differences in least-squares means (intervention effects) at each time point with 95% confidence intervals were derived. Cohen's d for the effect of the intervention was estimated by calculating the difference between estimated means (corrected for baseline) divided by raw pooled standard deviation. A two-sided p < 0.05 indicated statistical significance. As this study's primary intention was to test feasibility and acceptability, it was not powered to detect significant differences. The reliable change index was used to evaluate whether participants have reliable and clinically significant change scores from baseline to post-and follow-up (Jacobson and Truax, 1991).
The economic analysis was performed from both a healthcare system and a broader perspective, including productivity impacts on participants and their families. Health service utilization and productivity losses were estimated across all time points. PM+ training, supervision resource and delivery costs were obtained from project records. Unit costs were attached to health service utilization using published tariffs used in the Netherlands (Hakkaart-van Roijen et al., 2015). Medication reimbursement rates were obtained from the Netherlands National Health Care Institute (Zorginstituut Nederland, 2020). See Appendix Table A1 for unit costs used. All patient and family productivity losses were valued using age-specific 2018 minimum wage rates (Government of the Netherlands, 2020). All costs are 2018 euros and discounting was not applied given the short study duration. Given the skewed distribution of costs, differences in mean costs were compared between the two groups using bias-corrected and accelerated (BCa) bootstrapping 1000 times. Incremental cost-effectiveness ratios (ICERs) per additional recovery and improvement without recovery at 3-month follow-up were calculated. Statistical uncertainty was explored through bootstrapping 1000 randomly resampled pairs of costs and outcomes. Cost-effectiveness acceptability curves (CEACs) were generated to show the likelihood PM+ is cost-effective at different willingness to pay levels.
Semi-structured interviews were audio recorded and transcribed verbatim in interview language (Arabic, Dutch or English). Arabic transcripts were translated into English by bi-lingual research assistants. Deidentified transcripts were analysed thematically and independently by two researchers (AdG and AW) in NVivo (QSR International Pty Ltd, 2015). Inductive analysis was used to categorise data and elicit themes. Findings were discussed by these researchers and a final coding framework agreed and then applied to all transcripts.

Results
Objective 1: testing trial procedures Recruitment and consent rates Recruitment occurred from April to November 2018. In total, 205 individuals gave permission to SNTR staff to be contacted by the VU research team. Of these, 110 declined, ten were unreachable and five did not attend screening. Eighty-one participants gave IC and were screened for eligibility to participate. Of these, 60 participants were included and randomised into PM+/CAU (n = 30) Epidemiology and Psychiatric Sciences or CAU alone groups (n = 30). Figure 1 presents the CONSORT flow diagram.

Group allocation and blinding
An SNTR team-leader matched PM+ participants to helpers. During the trial, we decided to ask all participants at the screening interview whether they preferred matching with a male or female helper to facilitate the matching procedures; however, female helpers preferred not to be matched to male participants. Blinding was successful in 47% of outcome assessments.
Attendance and follow-up assessments Assessments occurred between May 2018 and March 2019. Twenty-eight (93.3%) of the 30 participants allocated to PM+/ CAU completed all five sessions. Two stopped after sessions 1 and 2 because the spouse did not allow participation and due to lack of time. Three months post-intervention, data were obtained from 54 participants, with 86.7% completing all assessments. Reported reasons for non-attendance across assessments were 'prefers to withdraw' (n = 3), 'lack of time' (n = 3), 'abroad/ unavailable' (n = 1) and 'no approval from spouse' (n = 1). In 23% of outcome assessments, assessors assisted participants by reading questionnaires aloud.

PM+ protocol fidelity assessment
Fifty per cent (n = 15) of PM+/CAU participants agreed to have their sessions audiorecorded. Two research assistants scored 25% (n = 18 tapes) of 72 available audiorecorded sessions after five practice tapes scored by both research assistants (Cohen's κ = .80). Fidelity checks indicated helpers adhered to 76.6% of the PM+ protocol. Helpers completed 27 checklists, reporting 93.7% completion of the PM+ protocol.
Objective 2: experiences with PM+ and barriers to treatment Semi-structured interviews with six PM+ participants, five helpers and five key informants explored experiences with PM+, acceptability and feasibility in addition to barriers and facilitators of  Appendix  Table A2.

Experiences with the PM+ intervention
Interviewees spoke generally positively about the PM+ intervention (Q1.1) and strategies (Q1.2-3). Its therapeutic effect on participant mental health (Q1.5-6), as well as that of helpers (Q1.7), was the dominant reason for holding a positive view. Helpers stressed breaking homework into small feasible steps would improve adherence (Q1.8). According to helpers, participants generally became more motivated and confident about the intervention after their first positive experience in applying a learned strategy. Some participants said they started to forget about strategies after completing PM+ (Q1.9), and several participants and helpers thought PM+ might be too brief (Q1.10-12).
Perceived facilitators of PM+ adherence were related to the accommodation of the programme for participants (e.g. travel expense coverage, flexibility scheduling sessions) and helpers (e.g. time reserved to conduct PM+ sessions and supervision during working hours, ability to decrease caseload) (Q1.13-15).
Overall, participants did not disclose their participation in PM+ for fear of gossip and ridicule (Q1.16). However, a few said they shared their PM+ experiences with family members and practised strategies with them. A challenge to intervention adherence was the 'busy lives' (H1) of some participants.

Views on the helper
Participants and helpers mentioned they quickly established rapport and trust and felt comfortable sharing their stories and problems (Q2.1). Two participants even called their helper a 'friend' (Q2.2).
All interviewees spoke about the importance of PM+ being in Arabic, as it facilitated communication and self-expression (Q2.3-4). Having a similar background was generally found supportive for building rapport, although it was mentioned that it could cause distrust. For example, one participant was initially apprehensive because the helper was Syrian but later found their shared experiences beneficial (Q2.5). Participants appreciated similarities with their helper, but also that the helper was otherwise a stranger (Q2.6-8), creating a confidential and safe environment (Q2.9).
Generally, participants perceived their helper as competent (Q2.10-11). One participant commented that if the helper was a professional, he may have benefitted more (Q2.12).
Information on experiences in training and supervision of helpers is provided in Appendix Table A2.
Objective 3: estimating likely effectiveness and cost-effectiveness Treatment effect Two serious adverse events related to domestic violence were reported to the Research Ethics Review Committee. Table 1 shows demographic characteristics, traumatic events and postmigration stressors. Table 2 presents findings on primary outcomes of anxiety and depression (HSCL-25), and secondary outcomes of functional impairment (WHODAS 2.0), symptoms of PTSD (PCL-5) and self-identified problems (PSYCHLOPS) in PM+/CAU and CAU groups at all time points. There were no significant group differences between participants who were lost to

Epidemiology and Psychiatric Sciences
follow-up v. those retained. The pattern of missing data is presented in Appendix Table A3.

Primary outcomes
Linear mixed models in the intention-to-treat sample for the HSCL-25 total score showed a significant effect of time, moderated by condition (χ 2 (2) = 10.41; p = 0.005). In the PM+/CAU group, overall HSCL-25 scores decreased relative to the CAU group (see Fig. 2 We also found a significant effect of time for HSCL-25 depression, moderated by condition (χ 2 (2) = 8.58; p = 0.013). In the PM+/CAU group, depression decreased relative to the CAU group. Post-hoc contrasts showed a medium effect post-assessment

Reliable change index for symptoms of anxiety and depression
At 3-month follow-up, 14 PM+/CAU participants had a reliable change on HSCL-25 total score, of which three were clinically significant (i.e. recovered). In the CAU group, six participants had a reliable decrease in HSCL-25 scores, while two had a reliable increase in scores (i.e. deteriorated) (Appendix Table A5).    Table 4 summarises results of the exploratory cost-effectiveness analysis. From a health system perspective, PM+/CAU had an ICER of €5047 (95% CI €0-€19 773) per additional recovery achieved. This was €2266 (95% CI €-1070 to €15 930) when productivity losses averted were included. Cost-effectiveness planes in Figs A1-A2 (Appendix) indicate when productivity losses are considered PM+/CAU may even have both better outcomes and lower costs than CAU. The ICER per additional improvement on the HSCL-25 without recovery was €2009 (95% CI €0-€2406). Figs A3-A6 (Appendix) show cost-effectiveness planes and CEACs per additional improvement without recovery.
While no accepted cost-effectiveness threshold for recovery from depression and anxiety exists, CEAC indicate at least an 85% chance that PM+/CAU is cost-effective if funders are willing to pay €10 000 per recovery (Figs A7-A8 in Appendix). (In the Netherlands, €20 000 per additional year lived in full quality health is usually considered cost-effective; Brouwer et al. (2019).)

Discussion
This pilot RCT aimed to test trial procedures, treatment facilitators and barriers, and likely effectiveness and cost-effectiveness of PM+ among Syrian refugees in the Netherlands. Our findings suggest that the adapted PM+ protocol is acceptable for use among Syrian refugees. Participants, helpers and key informants were generally positive about the intervention, including the PM+ strategies, accommodation (e.g. reimbursement of travel expenses) and delivery by Syrian helpers. Training and structural supervision of peer-refugees was perceived feasible and acceptable. These findings are further supported by helpers' adherence to the protocol and low participant drop-out from the intervention. Two adverse events unlikely attributable to the trial or intervention were reported by participants, suggesting that PM+ is a safe intervention. However, some participants may have experienced shame from trial participation.
Although the study was not powered to detect significant differences, depression and anxiety symptoms improved in the PM+/ CAU group relative to the CAU group, with moderate effect sizes. The study also indicated moderate to large improvements in overall psychosocial functioning, PTSD symptoms and self-identified problems. No significant difference in health service utilisation or costs was observed between groups, but overall costs were significantly higher in PM+/CAU due to PM+ implementation costs. Mean intervention costs ultimately are likely to be lower if trainers and helpers can be retained and continue to deliver PM+ to more refugees over a longer time period. Nonetheless, our exploratory economic analysis suggests PM+ has the potential to be costeffective from a health system perspective.
The moderate improvements across a broad range of symptoms are in line with previous PM+ trials (Rahman et al., 2016;Bryant et al., 2017), and support the intervention's transdiagnostic feature (Dawson et al., 2015). One key finding is that, although PM+ is not a trauma-focused intervention, it improved PTSD symptomatology in this war-affected sample. This adds to existing literature indicating PTSD symptoms can be successfully treated with brief, non-trauma-focused interventions (Nidich et al., 2018;Turrini et al., 2019). This is the first study exploring how PM+ can be delivered by peer-refugees in a high-income country. Although refugees are typically exposed to ongoing post-migration stressors, our study showed that effects were retained up to 3 months. A key strength in this study is the mixed-methods design that enabled us to examine both quantitative outcomes in a rigorous RCT, as well as perceptions of various stakeholders about the acceptability and feasibility of PM+. Another strength is the use of the secondary outcome measure PSYCHLOPS, which examines participantgenerated problems, instead of 'Western' mental health constructs. Furthermore, we added the WHODAS measure of overall psychosocial functioning to look beyond mental health and psychosocial problems, something often not included in the evaluation of psychosocial interventions for refugees (Turrini et al., 2019).
Our study also has a number of limitations. First, we failed to interview study drop-outs in our qualitative evaluation, limiting our insights on barriers to trial participation. Second, although treatment effects and cost-effectiveness results are promising, they should be interpreted with caution as no power calculations were carried out. Furthermore, participants were recruited from a foundation established for Syrian families with resident status living in an urban area, and results might be different for those still awaiting completion of their asylum procedure.
A major practical implication of the present pilot RCT is that the study and PM+ procedures can be successfully carried out among Syrian refugees. The observed low drop-out is promising for a definite RCT. Use of Audio-Computer-Assisted Self-Interview software (e.g. Morina et al., 2017) that does not require administration by an assessor may improve blinding.
Longer term follow-up is needed to better assess whether there is an impact on health service utilization and social functioning outcomes such as participation in work and study. With a larger sample size, it will also be possible to estimate changes in quality of life outcomes, using a generic outcome measure such as the Quality Adjusted Life Year (QALY), a metric that has resonance with policy makers. Replication in a fully-powered RCT is needed  Graaff et al., 2020). Other trials on PM+ with different modes of delivery (e.g. group) will be conducted in the larger STRENGTHS project (Sijbrandij et al., 2017). This will strengthen the external validity of trial findings and provide a potential model for scaling up in more and less well-resourced contexts.

Conclusion
This study indicates that the trial procedures and PM+ delivered by peer-refugee, non-specialist helpers are acceptable, feasible and safe. PM+ is likely effective in improving mental health outcomes and psychosocial functioning in Syrian refugees, and potentially cost-effective. A fully-powered, definitve RCT with longer follow-up is needed.
Data. The Vrije Universiteit Amsterdam (VU) will keep a central data repository of all data collected in the STRENGTHS project. The data will be available upon reasonable request to the STRENGTHS consortium. Data access might not be granted to third parties when this would interfere with relevant data protection and legislation in the countries participating in this project and any applicable EU legislation regarding data protection. Interested researchers can contact Dr Marit Sijbrandij at e.m.sijbrandij@vu.nl to initiate the process.
Acknowledgements. We would like to thank the helpers and other staff members of SNTR for their collaboration on implementing PM+; and research team members Jana Uppendahl, Ine Smeets, Shuker Barbour, Yenovk Zakarian and Dima Aldabous for conducting assessments and interviews.
Financial support. This study was supported by Stichting Nieuw Thuis Rotterdam and the STRENGTHS project. The STRENGTHS project is funded under Horizon 2020the Framework Programme for Research andInnovation (2014-2020). The content of this article reflects only the authors' views and the European Community is not liable for any use that may be made of the information contained therein.

Conflict of interest.
None.  For each intervention, report mean values for the main categories of estimated costs and outcomes of interest, as well as mean differences between the comparator groups. If applicable, report incremental cost-effectiveness ratios   In PM+ they will be with Syrian people, from their own culture, from their own language, but not people that they know, not family or so, so they will talk. Theme 3: Experiences of training and supervision The PM+ supervisor described the weekly supervisions as 'hard work' because the helpers had no background in psychology (KI5). Both supervisor and helpers experienced that further practice of PM+ skills during supervision increased helpers' ability to model the strategies to participants and adhere to the protocol (Q3.1-2). Helpers described supervision as a way 'to learn from each other' (H2) and 'improve their approach' (H4) (Q3.3). Communication between supervisors and helpers was in Dutch, but some of the role-play was in Arabic (to practice how PM+ will work with participants).
Individual case monitoring was mainly used to support tailoring PM+ to the individual (Q3.4), and to detect and potentially refer those with possibly more severe mental health problems (e.g. substance abuse) to professional care (Q3.5). Several KIs, however, expressed concerns about the current waiting lists for such professional help.
Recognition of participants' stories was helpful though potentially burdensome for helpers (Q3.6). They explained that hearing about other people's problems could be challenging, but that supervision and talking to colleagues helped them (Q3.7). The supervisor emphasised the importance of self-care during supervision, given the similarity of experiences (emotional burden), and because helpers carried out the sessions alongside their regular work for SNTR (time burden). Both helpers and the supervisor spoke about the importance of knowing 'your limits' as helper. One helper was asked to briefly put PM+ as extra activity in her regular work 'on hold' due to signs of burn-out.  The percentage of missed assessments (unit non-response; Brick and Kalton, 1996) in the study was 7.2%. Eighty-seven per cent (86.7%) of participants completed all three assessments. There were more assessment skips in the CAU group (13% at both post and 3-month follow-up) than in the PM+/CAU group (10% at post and 6.7% at 3-month follow-up). Reported reasons for non-attendance across assessments were 'prefer to withdraw' (n = 3), 'lack of time' (n = 3), 'abroad/unavailable' (n = 1), and 'no approval from spouse' (n = 1).
We used pro-rated, single imputations for outcome measures with missing items. We did not impute unit non-response, as linear mixed-models in R handle missing outcome data (Raudenbush, 2001). HSCL-25 = 25-item Hopkins Symptoms Checklist (range item-mean = 1-4, higher scores indicate elevated anxiety or depression); WHODAS 2.0 = WHO Disability Assessment Schedule 2.0 (range 12-60, higher scores indicate worse functional impairment); PCL-5 = PTSD Checklist for DSM-5 (range 0-80, higher scores indicate greater severity); PSYCHLOPS = Psychological Outcomes Profiles (range 0-20, higher scores indicate poorer outcome); M = mean; S.D. = standard deviation. a Effect sizes are determined by calculating the difference between the estimated means (corrected for baseline) divided by the raw pooled standard deviation. aThe Clinical Significant Change cut-off for the HSCL-25 (total scale) was calculated by subtracting 2 S.D. of the baseline M for the full sample. b The RCI for the HSCL-25 (total score) was calculated using the baseline S.D. for the full sample and baseline Cronbach's α as test-retest reliability coefficient (Jacobson and Truax, 1991). Recovered = clinical significant reliable change; Improved without recovery = no clinical significant reliable change; Deteriorated = reliable change with worsening of symptoms; No change = no reliable change.

22
A. M. de Graaff et al.    Appendix 3: Figures   Fig. A5 Cost-effectiveness acceptability curve: willingness to pay per improvement achieved from PM+/CAU intervention (health system perspective).

Fig. A7
Cost-effectiveness acceptability curve: willingness to pay per recovery achieved from PM+/CAU intervention (health system perspective).

Fig. A6
Cost-effectiveness acceptability curve: willingness to pay per improvement achieved from PM+/CAU intervention (health system perspective).

Fig. A8
Cost-effectiveness acceptability curve: willingness to pay per recovery achieved from PM+/CAU intervention (health system and productivity perspective).

24
A. M. de Graaff et al.