Culturally sensitive stepped care for adolescent refugees: efficacy and cost–utility of a multicentric randomized controlled trial

Study design

A longitudinal cluster-randomized controlled trial was conducted on a consecutive refugee sample with depressive symptoms in a rater-blinded setting at four university hospitals across Germany between May 2018 and March 2020. The stepped care approach included four interventions/levels: watchful waiting, smartphone application, group intervention and individual psychotherapy. Allocation to SCM interventions was based on depressive symptom severity (measured by PHQ-9). Intervention periods lasted 3 months for all levels. Figure 1 illustrates the study model and the allocation process to the SCM interventions. Sample size calculations and estimated dropout rates (50%) were only performed for the primary MEHIRA study. Detailed information of the study design (e.g., the randomization, blinding process, sample size calculations and data monitoring) are available in the study protocol [24]. The study design was not changed from the protocol. The study was conducted in accordance with the Declaration of Helsinki [26] and has been approved by the local ethics committees of all participating medical faculties. The study was registered in Clinical Trails.gov (registration number: NCT03109028).

Fig. 1figure 1

Illustration of study design and allocation process to SCM interventions

Recruitment

All study sites used diverse recruitment methods ranging from local school settings, outpatient settings as well as refugee housing and accommodation centers. Study teams presented the study and distributed flyers about the study in the mentioned locations to promote recruitment. No financial compensation was offered. For study inclusion, participants had to be asylum seekers or refugees according to the office of the United Nations High Commissioner for Refugees (UNHCR) definition [27], 14–21 years of age with sufficient language skills in Arabic, Farsi/Dari, English or German. Further, they had to show at least mild depressive symptoms (≥ 5 in the PHQ-9/PHQ-A) and psychological distress (≥ 12 for the items 1–14 or ≥ 5 for item 15 in the Refugee Health Screener (RHS)-15) at screening assessment (T-1). Exclusion criteria were symptoms of a psychotic or degenerative disorder and/or an acute risk of suicidality (≥ 4 on item 10 in the Montgomery–Åsberg Depression Rating Scale—MADRS) [28]. Prior to study begin, all participants were informed about study content, objectives and anonymous processing of the data and gave written informed consent. For participants aged 14–17 years, a written consent from their legal guardian was obtained.

Study procedure

Before the initial trial onset, all participants underwent a screening phase within 4 weeks to 1 day prior to randomization, where affective symptom severity was assessed (T-1). All screenings were conducted by trained psychologists to ensure consistency. After the screening process, participants were cluster randomized to either SCM or TAU at baseline (T0). Randomization was performed by an independent coordinating center for clinical trials in a 1:1 ratio. At baseline, symptomatology and socioeconomic information were assessed by trained psychotherapists via questionnaire and clinical interview. In the next step, participants received either a symptom-tailored SCM intervention or treatment as usual (TAU) for the next 12 weeks, directly followed by post-treatment assessment (T1). Follow-up measures were administered 12 weeks (T2) and 24 weeks (T3) after post-treatment assessment with predefined maximum deviation of 3 weeks.

Interventions

According to development guidelines for culturally sensitive interventions, people with a refugee background contributed to the development of the interventions as advisors, and focus group members as well as within the study in the role of translators and research assistants [29]. The distribution of baseline PHQ scores was used to assign the patients to suitable groups of the SCM. Participants were allocated to one of the following treatment levels according to validated severity classification [30]:

Watchful waiting (level 1: PHQ score 5–9)

Participants with mild depressive symptoms did not receive an active intervention, but contact details were provided in case mental support was needed.

Smartphone application “Balsam” (level 2: PHQ score 10–14)

Participants with moderate depressive symptoms received a smartphone application “Balsam”. This app was recently developed in the context of the MEHIRA project by a group of psychologists and researchers from diverse cultural backgrounds to “help migrants and refugees understand the underlying mechanisms of their stress and arm themselves with the appropriate tools to cope with their struggles” [24]. It contains more than 80 videos, is available in 4 languages (Arabic, Farsi, English and German) and includes 15 different modules covering psychoeducative elements, therapeutic exercises, built-in questionnaires and an emergency support function. Participants were encouraged to use the app daily via push notifications.

Group intervention “START_adapt” (level 3: PHQ score 15–19)

Participants with moderate–severe depressive symptoms were assigned to the “START_adapt” group intervention. This intervention was adapted from the Stress-Traumasymptoms-Arousal-Regulation-Treatment (START), a short cultural sensitive group intervention for adolescents affected by severe trauma and extreme emotional stress [31]. It was designed as a brief and standardized five-session therapeutic program for four to eight refugee minors based on elements of dialectical-behavior therapy (DBT) and trauma-focused cognitive behavioral therapy (TF-CBT). Groups were led by a licensed psychotherapist and a trained assistant. Instructions and information were available in four languages (Arabic, Farsi, English and German) and multiple visual displays were included in the manual. To allow for multi-lingual groups without interpreters, all contents were available by audio via an mp3-player.

Psychotherapy (level 4: PHQ score 20–27)

Participants with severe depressive symptoms received an individual psychological treatment by a licensed psychotherapist. The non-manualized treatment was based on cognitive behavioral approaches. If needed, a translator was provided.

Treatment as usual: TAU

Participants randomized to TAU were allowed to receive all available routine health-care services regardless of symptom severity. Participants were not guided and had to seek help to relieve the symptoms on their own. There were no regulations concerning the institutions, operators or kind of treatments participants received.

Measures

Self-rated questionnaires were available in validated Arabic, Farsi, English and German versions. Rater-based assessments were performed by a trained psychiatrist or psychologist who was blinded to the trial condition, if needed with the assistance of a translator. A comprehensive list of all measures can be found in the study protocol [24]. The following two instruments were used as measures for the clinical outcomes:

Patient Health Questionnaire (PHQ-9/PHQ-A)

The primary outcome, depression symptom severity, was assessed via age-related versions of the PHQ-9. The PHQ-9 is a brief and widely used nine-item self-report instrument based on DSM-IV to assess the frequency of depressive symptoms within the last 2 weeks on a four-point Likert scale [30]. The sum score can range from 0 to 27. Internal consistency of the PHQ-9 is high, with a Cronbach’s alpha of 0.86–0.89 [source]. Adolescents under the age of 18 years filled out the PHQ-A, a slightly modified, highly comparable and well-validated version of the PHQ-9 for adolescents [32].

Child and Adolescent Trauma Screen (CATS)

The secondary outcome was assessed via CATS, a short self-report instrument based on DSM-V to assess PTSD symptoms in children and adolescents [33]. Symptoms are assessed on a four-point Likert scale and can range from 0 to 60. Internal consistency of the symptom screening is high, with a Cronbach’s alpha of 0.88–0.94 [33].

Resource use and costs

Resource use was measured with an adapted version of the Mannheim-Module-Resource-Use (MRV) [35]. We performed the MRV as interview by trained staff from T0 to T3 and asked for the utilization of inpatient and outpatient health services, remedies, counseling and health support services, and medication. Data on resource use were then combined with unit costs. Unit costs were derived from nationally or regionally available data sources. We adopted a health-care system perspective including direct interventions costs as well. We calculated per patient costs for the SCM as a combination of a) recurring or running expenses due to personnel and operating costs and b) one-off costs defined as costs of development for each intervention including tutorial sessions, and preparation of manuals. Costs of SCM were derived via interviews with key persons of depression symptom severity-specific intervention types. We adjusted all prices to the reference year 2019 in euros, but did not discount costs due to the short time horizon of the study.

Determination of utilities

As economic end point for the cost–utility analysis, we considered cost per quality-adjusted life year (QALY). We derived QALYs from the abbreviated World Health Organization Quality of Life questionnaire (WHOQOL-BREF) from T0 to T3 [34] using the method proposed by Salize and Kilian [35] and extrapolated values for a 12-month period.

Statistical analysis

Data analysis was performed using IBM SPSS statistics version 22, SAS version 9.4 (SAS Institute Inc., Cary, North Carolina, USA) and Excel 2016 for Windows. The intention-to-treat (ITT) sample included all randomized participants who provided baseline data of the PHQ. For sensitivity analysis, the per protocol (PP) sample excluded all participants who dropped out at T1, showed missing data at T1, participated in < 50% of the intervention sessions or did not receive the allocated intervention. Differences in sample characteristics between SCM and TAU at baseline were investigated using ANOVA for continuous outcomes, and chi-square tests for categorical variables. In case of significant differences, variables were included as covariates in the main analysis. For the main analyses, we calculated a linear mixed model analysis (LMM) using the ITT sample with time (T0–T1) and group*time–interaction as fixed effects. Intraclass correlation coefficients (ICC) indicated that a nontrivial proportion of total variance was attributed to the study centers (0.12–0.32). Thus, we added study center as random effects to our LMM analyses. To check for robustness of findings, we performed the analyses for the PP sample (sensitivity analysis) and for all time points (follow-up analyses). Tests were two tailed and statistical significance was set at a p value < 0.05. If significant, effect sizes were calculated [36]. To evaluate the significance of change for the individual, response (≥ 50% PHQ-9 reduction at T1) and remission (PHQ-9 < 5 points at T1) rates were calculated for all patients with post-intervention PHQ scores.

Statistical analyses for health-care costs and cost–utility were performed using the ITT sample (base case-scenario). Missing values from T1 to T3 were replaced using the last observation carried forward (LOCF) approach. We applied generalized linear models (GLM) with gamma distributions and identity link function to estimate differences in log-transformed health-care costs between SCM and TAU. We determined the incremental cost-effectiveness ratios (ICER) which represent the additional costs to obtain one additional QALY. The ICER was calculated as the ratio between the differences in mean costs and the differences in mean QALY between SCM and TAU. We considered statistical uncertainty around the point estimate with nonparametric bootstrapping with 10,000 samples. In addition, we performed net monetary benefit (NMB) analyses to consider the likelihood of cost–utility against different willingness to pay thresholds. To check for robustness of findings, we performed sensitivity analyses. Therefore, we varied the conditions of calculating per capita costs of SCM by a) evaluating average costs on the basis of the total number of theoretically assigned participants considering that all intervention types had an expected capacity utilization of 100% (optimal scenario) and b) excluding resource use costs and considering intervention costs alone (on-top scenario).

留言 (0)

沒有登入
gif