Effectiveness of a stepped‐care programme of WHO psychological interventions in a population of migrants: results from the RESPOND randomized controlled trial

Abstract

Migrant populations – including labour migrants, undocumented migrants, asylum seekers, refugees, internationally displaced persons, and other populations on the move – are exposed to a variety of stressors that affect their mental health. We designed and tested the effectiveness of a stepped‐care programme consisting of two scalable psychological interventions developed by the World Health Organization (WHO) and locally adapted for migrant populations. A parallel‐group randomized controlled trial was conducted in Italy. We recruited migrant adults (≥18 years) with psychological distress (score of at least 16 on the Kessler Psychological Distress Scale, K10). The experimental arm received psychological first aid (PFA) and a stepped‐care programme consisting of two WHO interventions adapted for this population group: first, Doing What Matters in Times of Stress (DWM) and, for participants who still reported significant levels of psychological distress after DWM, Problem Management Plus (PM+). Each intervention lasted 5‐6 weeks and was delivered remotely by lay facilitators. The control arm received PFA and care as usual (CAU). The primary outcome was the change in symptoms of depression and anxiety from baseline to week 21 after randomization, measured by the Patient Health Questionnaire Anxiety and Depression Scale (PHQ‐ADS). Between December 14, 2021 and April 18, 2023, 108 migrants were randomized to the stepped‐care intervention and 109 to CAU. Analysis of the primary outcome revealed that participants receiving the stepped‐care programme showed a greater reduction in anxiety and depression symptoms compared to those receiving CAU (coefficient: –3.460, standard error, SE: 1.050, p=0.001) at week 21. The same difference was observed at week 7 (coefficient: –3.742, SE=1.008, p<0.001) and week 14 (coefficient: –6.381, SE=1.039, p<0.001). The stepped‐care programme was also associated with a greater improvement of depression and anxiety symptoms assessed separately at all timepoints, of post‐traumatic stress disorder symptoms at weeks 14 and 21, and of self‐assessed problems, function and well‐being at all timepoints. No serious adverse events occurred. This study provides evidence supporting the stepped‐care delivery of DWM and PM+ for migrant population groups with elevated distress. As these interventions are low‐intensity, transdiagnostic and task‐shifting, they are highly scalable. Existing evidence‐based guidelines and implementation packages should be updated accordingly.

Migration has been a common phenomenon for centuries. The term “migrant” typically encompasses many different groups: labour migrants, undocumented migrants, asylum seekers, refugees, internally displaced individuals, and other populations on the move ¹ .

According to the International Organization for Migration, there were about 281 million international migrants worldwide in 2021, marking a 27% increase compared to the 221 million migrants in 2010. Migrants represent a total of 3.6% of the world's population. Many migrants arrive in Europe through Italy, with 34,000 new arrivals in 2020 and almost 60,000 in 2021 ² .

Migrant population groups are exposed to various risk factors for mental health conditions. These include discrepancies between expectations and actual achievements, inadequate support systems, challenges in adaptation and acculturation processes, along with financial, administrative and legal hurdles throughout and after the migration trajectory ³ . Among migrants, forcibly displaced people – such as refugees and asylum seekers – face additional severe stressors, including the loss of their homes and possessions, and other traumatic events such as bombings, threats, imprisonment and torture.

In recent years, there has been a growing number of randomized controlled trials (RCTs) exploring the benefits of psychosocial interventions targeting psychological symptoms in migrant populations ⁴ . A systematic review of 52 studies (including 26 RCTs) identified a significant effect of psychological interventions in reducing depression, anxiety and somatization symptoms ⁵ . These results align with those of a scoping review of mixed‐methods studies, which indicated a positive effect of psychosocial interventions on participants’ mental health, mainly through a reduction in depressive symptoms and an improvement in social functioning ⁶ .

However, implementing the above interventions necessitates extensive training, a significant time for delivery, staff members with a background in mental health, and a robust monitoring and supervision framework. Moreover, it involves face‐to‐face individual delivery in most instances. As these characteristics are barriers to implementation, the World Health Organization (WHO) has developed a stress management intervention named Self Help Plus (SH+) ⁷ , and a brief intervention based on cognitive‐behavioral and problem‐solving strategies called Problem Management Plus (PM+) ⁸ . A guided self‐help programme based on the SH+ course, called Doing What Matters in Times of Stress (DWM), is also available and has been adapted to be delivered as a mobile‐supported website ⁹ .

SH+, PM+ and DWM are designed to be scalable, transdiagnostic and task‐shifting. They have been tested as stand‐alone interventions in diverse populations, including health care staff, asylum seekers, refugees, international migrants, involuntarily displaced people, and individuals exposed to armed conflicts, natural disasters, and health stressors such as the COVID‐19 pandemic ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ . Studies have generally found benefits in mental health outcomes, though the effectiveness of interventions may diminish over time ¹⁵ . One study conducted among health care staff with COVID‐19‐related distress combined DWM and PM+ into a stepped‐care programme compatible with fully remote training, delivery and supervision ¹⁶ . The stepped‐care programme was found to be feasible. It resulted in clinically significant reductions in symptoms of anxiety, depression and post‐traumatic stress disorder (PTSD) ¹⁷ , suggesting that it may be potentially beneficial in other populations exposed to adversity.

Against this background, the present study examined the efficacy of DWM and PM+ delivered as a stepped‐care programme in reducing anxiety and depression symptoms in a sample of migrants with elevated psychological distress.

METHODS

Study design

We conducted a parallel‐group RCT in Italy. The trial protocol was published and registered in clinicaltrials.gov (NCT04993534) ¹⁸ . No changes were made to the design after the trial started. The Ethics Committee of the University of Verona approved the project. Written informed consent was mandatory for all participants. In accordance with the Declaration of Helsinki, participants’ confidentiality was preserved, and the contents of the recruitment and follow‐up forms were not disclosed to any third party.

An Ethics and Data Advisory Board monitored the study and provided expert advice on data management and all ethical, legal and societal issues related to the project. The Consolidated Standards of Reporting Social and Psychological intervention Trials (CONSORT‐SPI) statement was followed in reporting trial results ¹⁹ . Participant recruitment occurred from December 14, 2021 to April 18, 2023.

Participants were adult migrants recruited through: a) key stakeholders such as non‐governmental organizations (NGOs) located in Italy; b) other community‐based organizations offering legal and/or social and/or psychosocial support to this vulnerable group; or c) social media and “word of mouth” (i.e., investigators proactively approached local organizations providing social, health and/or legal support to migrant populations, including refugees and asylum seekers, to identify potentially eligible participants).

Interested individuals were informed (in English, Italian or French), using an easily accessible terminology, about the nature and scope of the study. A research assistant explained details of the study and provided study materials. Participants meeting the inclusion criteria were randomized to receive psychological first aid (PFA) combined with the adapted stepped‐care DWM/PM+ intervention, or to receive PFA and care as usual (CAU) alone. After the screening at T0, participants were assessed at baseline before random allocation (T1) (one week after the screening), and after randomization at week 7 (T2), week 14 (T3), and week 21 (T4) (primary endpoint).

Inclusion and exclusion criteria

Participants were included if they met the following criteria: a) aged 18 years or older; b) being a migrant resettled in Italy temporarily or permanently (including labour migrants, undocumented migrants, asylum seekers, refugees, internationally displaced persons, or other persons on the move); c) having elevated levels of psychological distress (score of at least 16 on Kessler Psychological Distress Scale, K10 ²⁰ ); d) sufficient mastery of English, Italian or French (written and spoken); e) oral and written informed consent before entering the study.

Individuals who met the inclusion criteria were excluded from participation if they met any of the following criteria: a) acute medical conditions requiring hospitalization; b) imminent suicide risk or expressed acute needs or safeguarding risks that required immediate follow‐up; c) severe mental disorder (e.g., psychotic disorder); d) severe cognitive impairment (e.g., severe learning difficulties or dementia); e) initiated, stopped or significantly modified psychiatric drug treatment over the previous two months; f) receiving specialized psychological treatment at enrolment (e.g., cognitive‐behavioral therapy, eye movement desensitization and reprocessing); g) planning to permanently move back to their home country before the last quantitative follow‐up assessment (T4).

Randomization and masking

Randomization was coordinated by the WHO Collaborating Centre at the University of Verona. The electronic software Castor Electronic Data Capture (EDC) generated the randomization schedule, employing a variable block randomization method ²¹ . Research team members involved in recruitment could access the web‐based software to randomize each newly enrolled participant, but were not able to access the randomization list and were not aware of the block size. The Castor EDC software allowed random allocation only after the main information on the enrolled participant was entered, upon verification of the inclusion criteria. After random allocation, the software produced a unique identification number for each participant.

Masking participants and research staff was not feasible, due to the nature of the intervention programme. However, the statistician performing the analyses was masked to participant allocation status through pseudo‐blinding using coded groups. The trial statistician was not involved in determining participants’ eligibility, administering the intervention, measuring the outcomes, or entering data.

Experimental and control intervention

All participants initially received a phone call of up to 15‐20 min where intervention helpers provided information on which group they were allocated to, as well as on specific resources and supports they could access following the principles of PFA.

PFA is a WHO‐developed support strategy that involves human, supportive and practical help for individuals who have been affected by humanitarian crises ²² . It consists of a conversation during which the helper provides non‐intrusive practical care and support, assesses needs and concerns, helps people to address basic needs (e.g., information), listens to people without pressuring them to talk, comforts them and helps them calm down; and helps them to connect to information, services and social support ²² , ²³ .

Participants allocated to the control arm received PFA and CAU, which could include community care, social/legal support, and psychoeducation on general distress and personal and community resources.

Participants allocated to the intervention arm received PFA, CAU and the stepped‐care program, which included DWM and – for participants who reported significant levels of psychological distress after DWM (score of at least 16 on the K10 scale) – PM+.

After allocation, participants in the intervention arm were assigned to a helper who provided ongoing support over the phone, assisting with practical exercises and explaining key concepts of DWM. After an initial welcome call, participants received a message with login details to access the DWM course. As a result of the local adaptation process, reported elsewhere ²⁴ , we transformed DWM into a mobile‐friendly website and adapted some content to reflect barriers or stress triggers that might affect migrant populations in Italy.

The DWM course was delivered over a period of 5‐6 weeks, with new modules released every week. Helpers scheduled weekly support calls lasting approximately 15 min each, and provided motivation and support in using DWM. Participants who did not want to receive phone calls could contact their helpers using the messaging system available on the website. The DWM course is based on acceptance and commitment therapy techniques (e.g., acting on values, making room for difficult thoughts and feelings, keeping attention and curiosity), along with audio recordings to support practice ²⁴ . Participants were reminded of the sessions through text messages, in accordance with the WHO manual for delivery of the intervention.

After 5‐7 days from DWM completion (T2), an assessment was made of the criterion for stepping up to PM+, i.e. significant levels of psychological distress as measured by the K10 scale (score of at least 16). The PM+ intervention, culturally and contextually adapted according to WHO protocols, was administered by trained helpers without a formal background in mental health ²⁵ , over a period of 5‐6 weeks.

The PM+ protocol provides five different behavioral strategies: stress management, problem‐solving techniques, behavioral activation, promoting social support, and maintaining the effects. The cultural adaptation of the intervention was conducted through ten online meetings over a 6‐month period between the staff of the University of Verona, WHO officers, and representatives of other sites of the RESPOND Consortium.

Both DWM and PM+ interventions had an online format and were delivered in Italian, English or French. The intervention manuals are available on the WHO website (www.who.int). A detailed description of the interventions delivered during the trial (PFA, CAU, DWM and PM+) is provided in the supplementary information.

Helpers were bilingual (Italian/English or Italian/French) and received training in Italian on PFA, DWM and PM+ according to WHO protocols and manuals ²⁶ . The training was conducted by master trainers based at the University of Verona (clinical psychologists trained by WHO officers and/or experts with long experience in delivering WHO interventions). Details on the training activities are provided in the supplementary information. Intervention supervision was provided for DWM and PM+ helpers by clinical psychologists, who were available to address questions, as well as to provide debriefing after sessions. If necessary, additional training and consultation were available. Fidelity was checked by the intervention supervisor, who was not involved in the delivery of interventions, observed at least 10% of DWM sessions, and listened to at least 10% of recorded PM+ sessions.

Participants in both arms received: a) baseline and follow‐up assessments according to the study schedule, b) information about freely available health and social services, and c) links to community networks providing support for migrant populations.

Measures

Participants completed online questionnaires, using the Castor EDC software ²¹ , at T0 (screening for eligibility); T1 (baseline assessment, before random allocation); T2 (week 7 after randomization); T3 (week 14 after randomization); and T4 (week 21 after randomization).

Screening for eligibility was conducted using the K10. This is a ten‐item self‐report questionnaire to screen broadly for psychological distress experienced in the past 30 days ²⁰ . Each item is scored from 1 (“none of the time”) to 5 (“all of the time”). Scores of the ten items are then summed, yielding a minimum score of 10 and a maximum score of 50. The K10 has robust psychometric properties and strong discriminatory power to distinguish DSM‐IV cases from non‐cases ²⁰ . Suicidality was explored by the “Assessment of suicidal thoughts” risk tool from PM+. The possible presence of a severe mental disorder or cognitive impairment was assessed using the PM+ tool “Impairments possibly due to severe mental, neurological or substance use disorders”.

The primary study outcome was the change in symptoms of depression and anxiety from baseline to week 21 after randomization (T4), measured through the combined sum score of the Patient Health Questionnaire‐9 (PHQ‐9) ²⁷ and the Generalized Anxiety Disorder‐7 (GAD‐7) ²⁸ , previously validated as the Patient Health Questionnaire Anxiety and Depression Scale (PHQ‐ADS) ²⁹ . The scale scores range from 0 to 48, with higher scores indicating higher levels of depression and anxiety symptoms.

Secondary measures included the changes in symptoms of anxiety, depression and PTSD, and self‐assessed problems, function and well‐being, evaluated at all timepoints (T1, T2, T3 and T4). Depression and anxiety symptoms were measured using PHQ‐9 and GAD‐7, respectively. PTSD symptoms were assessed using the eight‐item version of the PTSD Checklist for DSM‐5 (PCL‐5) ³⁰ , which provides scores ranging from 0 to 32, with higher scores indicating higher levels of PTSD symptoms. The instrument is based on the PTSD Checklist ‐ Civilian Version (PCL‐C), a DSM‐IV‐based checklist ³¹ .

Self‐assessed problems, function and well‐being were measured using the Psychological Outcomes Profiles (PSYCHLOPS) ³² , a patient‐generated tool consisting of four questions (two for problems, one for function, and one for well‐being). Participants are asked to give free text responses to the questions. Responses are scored on an ordinal six‐point scale ranging from 0 to 5, producing a maximum score of 20. If both problem questions have been responded, the total score is the sum of the four items. If only the first problem question has been responded, the score of the first question is doubled.

Assessments were completed remotely via secure online links to Castor EDC. Adverse events reported spontaneously by the participants or observed by the research staff were recorded, and any serious adverse events were reported to the Ethics and Data Advisory Board.

Statistical analysis

Based on prior studies on PM+ ³³ , ³⁴ , we aimed to detect a medium effect size (defined as the square root of the ratio of the variance of the tested effect to its error variance) of 0.3 in the PM+ group at T4, based on the primary composite outcome PHQ‐ADS. A power calculation for a repeated measurement design suggested a minimum sample size of N=74 per group (power = 0.95, alpha = 0.05, two‐sided) in order to identify an effect at the time of interest. Assuming an attrition rate of 30%, we aimed to include 212 participants (106 in the DWM/PM+ intervention group and 106 in the control group).

All primary and secondary analyses were performed on an intention‐to‐treat (ITT) basis. The ITT population consisted of all participants randomly assigned to one of the two groups and with data available on at least the baseline assessment. In order to check the robustness of results, all outcomes were additionally analyzed using a per‐protocol (PP) approach that included only DWM participants clicking through all the contents of at least three modules and PM+ participants attending at least four sessions.

We calculated the descriptive statistics (mean with SD for interval‐level variables, number and percentage for categorical variables) at baseline and for the two intervention arms separately. Arms were compared using standardized mean differences (SMDs).

The primary analysis assessed the intervention effect on the average PHQ‐ADS score at each timepoint in the ITT population. To estimate the intervention effect for the timepoints T2, T3 and T4, we employed a linear mixed model for the analysis of PHQ‐ADS, which had time as a fixed effect, baseline measurement of PHQ‐ADS as a covariate, and subject as a random effect. The model was re‐parametrized by constraining the intervention fixed‐effect to be 0, and by including a time‐intervention interaction at T2 as well. In this way, at each timepoint, the intervention effect was measured as the interaction between time (as a categorical variable) and intervention, with its value at T4 being our outcome of interest.

The interaction effects and confidence intervals (CIs) represent the average difference between the two study arms at each timepoint. We used the mean of the values predicted from the model to calculate the estimated average values for the two study arms in case all participants were assigned to the intervention versus the control arm. In addition, a covariate‐adjusted mixed model of the primary outcome was performed by adding covariates showing imbalance at baseline (as measured by a SMD above 0.1 in absolute value). Robust standard errors (SEs) were used in all models.

A secondary analysis of the effect of the intervention on the outcomes was conducted in the PP population, using the same approach as reported above. In addition, a covariate‐adjusted mixed model of primary outcome was performed using this population by adding pre‐specified covariates at baseline (gender, age, whether the person had at least secondary education; prior trauma expressed as replying “Yes” to at least one item from the Brief Trauma Questionnaire ³⁵ ; whether the person had been infected by COVID‐19; and the stressor exposure as measured by the Mainz Inventory of Microstressors ³⁶ ).

No imputations of missing values at the scale level were made, as multilevel models can deal with missing data in case the missing at random assumption holds ³⁷ . If only some items were missing for a particular scale, we used the corrected item mean substitution method (i.e., the item mean across participants weighted by the subject's mean of completed items) ³⁸ , using information from subjects belonging to the same intervention arm for the same follow‐up time (estimated values above the maximum or below the minimum admissible value were set to maximum/minimum). As a sensitivity analysis, the analyses for outcomes with partially imputed scales were repeated by excluding such imputed values. To avoid missing values among categorical predictors, a category “missing value” was included.

A linear mixed model with robust SEs, as mentioned for the primary analysis, was carried out to analyze the following secondary outcomes: changes in depressive symptoms (PHQ‐9), generalized anxiety symptoms (GAD‐7), PTSD symptoms (PCL‐5), self‐assessed problems, function and well‐being (PSYCHLOPS).

Possible interactions between the intervention and specific variables (baseline score on the primary outcome, age, gender, legal status, time since resettlement; whether the person had at least secondary education, was receiving an income, and had ever consulted a mental health professional) were evaluated, by excluding categories with data from less than ten participants. A global test on each variable was implemented and, in case of statistical significance after applying the Benjamini‐Hochberg correction ³⁹ , statistical significance at each timepoint was evaluated for that variable.

Finally, the loss‐to‐follow‐up rate was compared between the two groups using a chi‐square or a Fisher exact test, as appropriate. All analyses were performed using Stata/SE, Release 17.0 ⁴⁰ .

RESULTS

After screening 238 potentially eligible participants, 21 were excluded (19 of them had a level of distress lower than the established cut‐off; one was on an unstable dose of psychotropic medication; and one refused to participate) (see Figure 1). This left 217 individuals who met the inclusion criteria, consented to be randomized by signing a written informed consent form, and were allocated to either the stepped‐care programme (N=108) or CAU (N=109). Only 16.6% of randomized participants were lost to follow‐up. The distribution of participants lost to follow‐up did not differ between the study groups at any timepoint (see supplementary information).

Figure 1.

CONSORT‐SPI flow diagram. K10 – Kessler Psychological Distress Scale

Selected socio‐demographic characteristics of the included participants are shown in Table 1 (see supplementary information for other variables). More than one third of participants were male; the average age was about 36 years in both groups. The majority of participants had at least a secondary education, with almost 40% having an academic education. The country of origin was in Asia/Pacific for 14.8% of them; in Europe or Central Asia for 34.0%; in the Americas or the Caribbean for 26.8%; and in the Middle East or Africa for 24.4%. Most participants were permanent residents in Italy (59.7%), 22.4% had a temporary permit to stay, and 17.9% were refugees or asylum seekers. The reported travel duration to reach Italy was over six months for 16.3% of participants (see Table 1).

Table 1.

Socio‐demographic and clinical characteristics of study participants by treatment allocation

	All participants	Intervention group	Control group	Difference (SE)	SMD
Age (years), mean±SD	35.7±12.5	35.5±13.1	35.9±11.9	–0.414 (1.735)	0.033
Gender (%)
Female	62.7	68.5	56.9	0.116 (0.065)	0.241
Male	36.4	31.5	41.3	–0.098 (0.065)	0.204
Other	0.9	0	1.8	–0.018 (0.013)	0.192
Education level (%)
Up to primary school	12.2	8.3	16.0	–0.077 (0.047)	0.235
Secondary school	48.5	50.0	47.0	0.030 (0.072)	0.060
University	39.3	41.7	37.0	0.047 (0.070)	0.095
Country of origin (%)
Asia/Pacific	14.8	17.5	12.3	0.052 (0.049)	0.146
Europe or Central Asia	34.0	35.9	32.1	0.038 (0.066)	0.081
America or Caribbean	26.8	23.3	30.2	–0.069 (0.061)	0.155
Middle East or Africa	24.4	23.3	25.5	–0.022 (0.060)	0.050
Legal status (%)
Temporary permit	22.4	22.9	22.0	0.009 (0.060)	0.022
Permanent resident	59.7	64.6	55.0	0.096 (0.070)	0.195
Refugee or asylum seeker	17.9	12.5	23.0	–0.105 (0.054)	0.276
Travel duration (%)
Up to six months	83.7	85.4	82.0	0.034 (0.053)	0.092
Over six months	16.3	14.6	18.0	–0.034 (0.053)
Ever consulted a mental health professional (%)
Yes	41.5	38.9	44.0	–0.051 (0.071)	0.102
No	58.5	61.1	56.0	0.051 (0.071)
Having an income (%)
Yes	56.5	60.2	53.1	0.072 (0.072)	0.144
No	43.5	39.8	46.9	–0.072 (0.072)
Measures at baseline, mean±SD
PHQ‐ADS score	19.08±8.69	19.14±8.56	19.02±8.89	–0.119 (1.185)	0.014
PHQ‐9 score	10.05±4.93	9.87±4.82	10.22±5.07	–0.350 (0.672)	0.071
GAD‐7 score	9.03±4.52	9.15±4.65	8.92±4.43	0.231 (0.616)	0.051
PCL‐5 score	11.77±7.23	12.05±7.36	11.49±7.14	0.562 (0.989)	0.075
PSYCHLOPS score	13.304±4.079	13.343±4.276	13.264±3.890	0.079 (0.566)	0.019

SMD – standardized mean difference, SE – standard error, PHQ‐ADS – Patient Health Questionnaire Anxiety and Depression Scale, PHQ‐9– Patient Health Questionnaire‐9, GAD‐7 – Generalized Anxiety Disorder‐7, PCL‐5 – PTSD Checklist for DSM‐5, PSYCHLOPS – Psychological Outcomes Profiles. SMD values in bold prints are those above the threshold for imbalance.

Assessment of more than 10% of DWM and PM+ sessions indicated near‐perfect fidelity. Only in a few cases (<10) DWM calls were longer than the established duration (i.e., 30 min). We identified minor deviations from the PM+ protocol, due to adaptations for cultural aspects, or specific content that did not totally apply to the problems reported by participants. The total supervision time required for all sessions of DWM and PM+ was 3 hours per helper on average (approximately 12 hours in total).

At T2 (week 7), 32.4% (35/108) of participants allocated to the intervention arm versus 21.1% (23/109) allocated to the control condition reported a clinically significant improvement in distress, as shown by a score below the cutoff of 16 at the K10. In the experimental arm, therefore, 35 participants did not step into PM+. The mean K10 value at T2 was 19.49 (SD=5.91) in the intervention arm and 22.94 (SD=8.13) in the control condition. During the study period, apart from the experimental or control intervention, the mental and physical health care received did not differ between the two groups (see supplementary information).

Differences between study conditions on primary and secondary outcome measures are reported in Table 2. The stepped‐care programme led to a significant reduction of anxiety and depression symptoms compared to CAU, as measured by the PHQ‐ADS at T4 (coefficient: –3.460, SE=1.050, p=0.001) (primary outcome). The same was observed at the other timepoints (coefficient: –3.742, SE=1.008, p<0.001 at T2; coefficient: –6.381, SE=1.039, p<0.001 at T3). Figure 2 shows the trend over time in the average values of symptoms of depression and anxiety measured by the PHQ‐ADS in each of the two groups, with their CIs.

Table 2.

Results for primary and secondary outcomes at each timepoint (intention‐to‐treat analysis)

	Intervention	Control
	Estimated average value (SE)	Estimated average value (SE)	Coefficient (SE)	p	Standardized coefficient (SE)
PHQ‐ADS score
T2	12.303 (0.696)	16.045 (0.728)	–3.742 (1.008)	<0.001	–0.414 (0.111)
T3	9.112 (0.631)	15.493 (0.825)	–6.381 (1.039)	<0.001	–0.705 (0.115)
T4 (primary outcome)	10.625 (0.730)	14.085 (0.755)	–3.460 (1.050)	0.001	–0.382 (0.116)
PHQ‐9 score
T2	6.769 (0.390)	8.414 (0.400)	–1.645 (0.560)	0.003	–0.324 (0.110)
T3	5.123 (0.368)	8.300 (0.502)	–3.177 (0.623)	<0.001	–0.625 (0.123)
T4	5.978 (0.418)	7.292 (0.420)	–1.314 (0.593)	0.027	–0.258 (0.117)
GAD‐7 score
T2	5.537 (0.336)	7.621 (0.391)	–2.085 (0.516)	<0.001	–0.471 (0.117)
T3	3.995 (0.308)	7.193 (‘0.423)	–3.198 (0.523)	<0.001	–0.723 (0.118)
T4	4.652 (0.348)	6.783 (0.400)	–2.131 (0.531)	<0.001	–0.482 (0.120)
PCL‐5 score
T2	8.422 (0.630)	10.056 (0.568)	–1.633 (0.849)	0.054	–0.235 (0.122)
T3	6.079 (0.540)	9.592 (0.625)	–3.513 (0.827)	<0.001	–0.506 (0.119)
T4	5.994. (0540)	8.517 (0.538)	–2.523 (0.763)	0.001	–0.363 (0.110)
PSYCHLOPS score
T2	8.085 (0.506)	10.234 (0.529)	–2.149 (0.732)	0.003	0.392 (0.134)
T3	6.379 (0.520)	9.907 (0.572)	–3.528 (0.773)	<0.001	–0.644 (0.141)
T4	5.427 (0.479)	8.995 (0.533)	–3.567 (0.717)	<0.001	–0.651 (0.131)

SE – standard error, PHQ‐ADS – Patient Health Questionnaire Anxiety and Depression Scale, PHQ‐9 – Patient Health Questionnaire‐9, GAD‐7 – Generalized Anxiety Disorder‐7, PCL‐5 – PTSD Checklist for DSM‐5, PSYCHLOPS – Psychological Outcomes Profiles. Bold prints indicate statistically significant differences.

Figure 2.

Average PHQ‐ADS (Patient Health Questionnaire Anxiety and Depression Scale) total score at different timepoints, with confidence intervals, in intervention (solid line) and control (dotted line) arms (intention‐to‐treat population). SE – standard error.

A significant difference was also observed considering depression and anxiety symptoms separately at all timepoints (see Table 2). The stepped‐care programme, compared with CAU, was also associated with larger improvements for PTSD symptoms at T3 (coefficient: –3.513, SE=0.827, p<0.001) and T4 (coefficient: –2.523, SE=0.763, p=0.001), and for self‐assessed problems, function and well‐being at all timepoints (see Table 2).

The results of the ITT analysis were confirmed by the PP analysis (see Table 3). Secondary analyses conducted without any imputations of missing values did not identify any relevant difference with respect to the main analyses (see supplementary information). As the two groups differed on some socio‐demographic variables at baseline, we included these variables in planned regression analyses of the primary outcome, without finding relevant differences in relation to our main analyses (see supplementary information).

Table 3.

Coefficients for primary and secondary outcomes at each timepoint (per‐protocol analysis)

	Coefficient (SE)	p
PHQ‐ADS score
T2	–4.215 (1.039)	<0.001
T3	–6.982 (1.043)	<0.001
T4 (primary outcome)	–4.208 (1.068)	<0.001
PHQ‐9 score
T2	–1.948 (0.609)	0.001
T3	–3.493 (0.612)	<0.001
T4	–1.696 (0.612)	0.006
GAD‐7 score
T2	–2.256 (0.542)	<0.001
T3	–3.491 (0.545)	<0.001
T4	–2.504 (0.545)	<0.001
PCL‐5 score
T2	–1.831 (0.828)	0.027
T3	–4.088 (0.826)	<0.001
T4	–2.930 (0.830)	<0.001
PSYCHLOPS score
T2	–2.606 (0.739)	<0.001
T3	–3.796 (0.776)	<0.001
T4	–4.044 (0.729)	<0.001

SE – standard error, PHQ‐ADS – Patient Health Questionnaire Anxiety and Depression Scale, PHQ‐9 – Patient Health Questionnaire‐9, GAD‐7 – Generalized Anxiety Disorder‐7, PCL‐5 – PTSD Checklist for DSM‐5, PSYCHLOPS – Psychological Outcomes Profiles. Bold prints indicate statistically significant values.

We also tested, as planned, for interactions between intervention allocation and potential moderators (baseline score on the primary outcome, age, gender, legal status, time since resettlement; whether the person had at least secondary education, was receiving an income, and had ever consulted a mental health professional). In mixed models built on our main model, by adding such variables and their interaction with intervention allocation as regressors, only the baseline value of the primary outcome measure was statistically significant, and remained so after the Benjamini‐Hochberg correction (p=0.006, see Table 4). The effect of the intervention on reducing PHQ‐ADS scores was stronger for participants with higher scores at baseline (see supplementary information).

Table 4.

Test for interactions of potential moderators with treatment

	Chi‐square	p	Adjusted p
Baseline score on the primary outcome	17.17	0.0007	0.006
Gender	0.71	0.872	0.885
Age	4.48	0.214	0.506
At least secondary education	7.32	0.292	0.506
Legal status	7.06	0.316	0.506
Time since resettlement	3.65	0.302	0.506
Income	1.85	0.605	0.807
Ever consulted a mental health professional	0.65	0.885	0.885

The reported adjusted p values are those following Benjamini‐Hochberg correction. Bold prints indicate statistically significant values.

We did not identify any serious adverse event. Six adverse events were identified, all of them regarded as unrelated to study participation (one accidental fall, one suicidal thought, two hospitalizations for a medical condition, two bereavement conditions).

DISCUSSION

In a migrant population with elevated psychological distress, a stepped‐care programme combining DWM and PM+ (two WHO‐developed, low‐intensity, task‐shifting psychological interventions) was effective in alleviating anxiety and depressive symptoms.

Efficacy was consistently observed at different timepoints, with coefficients indicating a substantial impact. Improvements were noted in depression and anxiety symptoms separately across all timepoints. Furthermore, the stepped‐care programme showed positive effects on PTSD symptoms and self‐assessed problems, function and well‐being.

Exploratory analyses for heterogeneity did not detect significant interactions between the intervention and potential moderators, except for baseline values of the primary outcome measure. Notably, the intervention effect was more pronounced in participants with higher baseline levels of anxiety and depression, which can be seen as further evidence of its impact. The same finding was reported in a trial of PM+ for Syrian refugees ⁴¹ and in a trial testing stepped‐care DWM/PM+ in health care workers in Spain ¹⁷ . In terms of acceptability, no serious adverse events were detected and very few participants were lost at follow‐up. These findings support the programme's effectiveness and suggest its applicability to migrant populations.

The beneficial effects of the stepped‐care programme may be related to various factors. DWM, based on acceptance and commitment therapy, aims to increase psychological flexibility and improve coping strategies to deal with adversity. As it is self‐administered and only facilitated by trained helpers, it offers the opportunity to practice exercises through an online web/app, and to learn ways of recognizing and managing emotional states ⁹ , ⁴² , ⁴³ . DWM might have encouraged participants to better adapt to fluctuating situational demands, by helping them to find ways of acting in accordance with their values, even in the face of external difficulties and migration‐related stressors ⁴³ . Participants may have acquired and consolidated skills to accommodate and “unhook” from difficult thoughts and feelings, through the integration of mindfulness techniques practiced regularly.

The possibility to move to PM+ for those still experiencing distress after DWM was a practical source of help in identifying and managing problems. PM+ helps people to improve the management of practical (e.g., unemployment, interpersonal conflict, poverty) and psychological (e.g., depression, anxiety, grief, fear, feelings of helplessness) problems. Additionally, PM+ ingredients such as behavioral activation and use/strengthening of social support may have contributed to lowering symptoms of common mental disorders. Psychotherapy research indicates that guided Internet‐based psychological interventions are influenced by social support to a greater extent than in‐person therapy ⁴⁴ , ⁴⁵ . This is attributed to the fact that online interventions heavily depend on self‐motivation and the completion of activities even in the absence of direct or long therapist interaction ⁴⁴ , ⁴⁶ .

Moreover, the interpersonal dynamics with helpers may have exerted a direct and positive impact on outcomes ⁴⁷ . These dynamics are particularly important for migrants, because of the potential lack of robust social support and networks in the country of resettlement. A systematic review of 35 RCTs, which examined 33 mental health interventions delivered through a digital format, found that the effects were larger when interventions were complemented with clinical assistance ⁴⁸ . This underscores the key role of helpers in our trial, especially for PM+.

In addition, the digital format of DWM and PM+ is more flexible than in‐person delivery, and could have contributed to increasing attendance, as highlighted by the low number of participants who did not complete the sessions. This, in turn, could have reinforced the effect of the intervention. In previous RCTs testing SH+ delivered in person and in groups to asylum seekers and refugees, we identified high proportions who did not attend the sessions ¹² , ¹³ . This may reflect the fact that migrants have many competing priorities other than attending mental health‐focused interventions, such as meeting basic needs, securing housing, navigating legal procedures, finding a job, and learning a new language ²⁴ .

We note some limitations of our study. First, we were inclusive in the definition of migrant participants, with the advantage of identifying a large group of distressed people, including asylum seekers, refugees and people in unstable living conditions. However, factors such as the type and number of stressors, barriers in the host country, availability of social support and sheltering centres, and time since resettlement, might have generated heterogeneity in the sample, with potential impact on the intervention's effect ⁴⁹ . Nonetheless, when we tested for interactions between intervention allocation and age, gender, education level, time since resettlement, legal status and receiving an income, we found no signal that the effect of the stepped‐care programme might differ with respect to these factors. Additionally, the main socio‐demographic characteristics of the population group of our trial are aligned with those reported by the International Organization for Migration in relation to international migration flows and migrants resettled in Italy ¹ . All this suggests that the stepped‐care programme has a potential of uptake across migrant populations beyond this trial.

A second limitation is that a double‐blind design was not feasible, and outcome measures were not assessed by masked assessors, but were self‐reported. The use of self‐reports can introduce variability and reduce the reliability of data, and the participants' overall perception of the intervention may influence how they report outcomes, leading to wrong estimates of effects. In the present study, however, this risk was mitigated by a design in which all participants received a supportive intervention, i.e. PFA. It is therefore likely that participants in both arms had similar perceptions of care. The finding that losses to follow‐up were minimal and similarly distributed in the two intervention arms appears to support this consideration. We also observed that the use of social and health care services was similar in the two groups during the study, highlighting a low risk of performance bias.

Third, the study had a relatively short follow‐up period. Therefore, we cannot exclude that the positive effects that we observed would diminish over longer follow‐up periods. Moreover, our study was not specifically designed to test a stepped‐care model against a single intervention. Future studies could usefully examine the stepped‐care model versus PM+ or DWM as stand‐alone interventions.

Overall, these results significantly expand the existing knowledge on the efficacy of psychological interventions in migrant populations, by showing for the first time that low‐intensity, task‐shifting interventions with freely accessible manuals may be implemented as a stepped‐care programme to alleviate anxiety and depression in migrants with elevated distress. Due to these characteristics, these interventions are uniquely suited for implementation in low‐resource settings. Considering that even countries classified as middle‐ or high‐income, such as Italy, may experience significant resource constraints in certain sectors, regions, or for certain populations such as migrants, these interventions may be appropriate for countries at any level of economic development.

Regarding implications for policy makers aiming to scale up these interventions, local adaptation may be a key factor ²⁶ , ⁵⁰ , ⁵¹ . It is important to tailor the stepped‐care programme to the specific needs and characteristics of the target population. The demographics, cultural norms and unique challenges of the community or group for whom the intervention is intended should be carefully assessed ⁵² , ⁵³ , ⁵⁴ . Adaptation may involve translating materials into local languages, considering cultural sensitivities, and incorporating feedback from local stakeholders, to make the intervention accurate, understandable and acceptable.

A second consideration is that the stepped‐care programme may be scaled up in parallel or in series with existing services. In the parallel approach, it is introduced alongside the existing services, creating a parallel track for addressing vulnerability to mental health challenges. By introducing parallel interventions, it may be possible to reach more migrant groups, outside the health care sector, ensuring that a broader spectrum of people can access the support that they require. However, there are some challenges associated with parallel implementation: in particular, it can strain resources, as it necessitates separate funding, staffing and infrastructure. This can lead to inefficiencies or duplication of efforts.

In contrast, the series approach involves introducing the stepped‐care programme sequentially, for example before or after migrants have received the existing social services. One key advantage of implementing in series is resource efficiency. It maximizes the use of existing infrastructure and personnel before introducing new elements, minimizing duplication of resources. However, the series approach may be less adaptable to evolving needs or changing circumstances, and it may not accommodate specific, targeted interventions as effectively as the parallel approach.

The scale‐up of the stepped‐care programme, either in parallel or in series with existing services, needs to be studied using quantitative or mixed approaches, aiming to identify the most cost‐effective implementation strategies.

In conclusion, this study provides evidence supporting the effectiveness of the stepped‐care delivery of DWM and PM+ in migrant population groups with elevated distress. Existing evidence‐based guidelines and implementation packages should be updated accordingly, and applied by various social and health care organizations, to ensure that migrant groups have equitable access to high‐quality mental health care.