World Health Organization's low‐intensity psychosocial interventions: a systematic review and meta‐analysis of the effects of Problem Management Plus and Step‐by‐Step

Abstract

Many societies have been recently exposed to humanitarian and health emergencies, which have resulted in a large number of people experiencing significant distress and being at risk to develop mental disorders such as depression, anxiety and post‐traumatic stress disorder. The World Health Organization has released a series of scalable psychosocial interventions for people impaired by distress in communities exposed to adversities. Prominent among these is a low‐intensity transdiagnostic psychosocial intervention, Problem Management Plus (PM+), and its digital adaptation Step‐by‐Step (SbS). This systematic review is the first to summarize the available evidence on the effects of PM+ and SbS. Up to March 8, 2023, five databases were searched for randomized controlled trials examining the effects of PM+ or SbS on distress indicators (i.e., general distress; anxiety, depressive or post‐traumatic stress disorder symptoms; functional impairment, self‐identified problems) and positive mental health outcomes (i.e., well‐being, quality of life, social support/relationships). We performed random‐effects multilevel meta‐analyses on standardized mean differences (SMDs) at post‐intervention and short‐term follow‐up assessments. Our search yielded 23 eligible studies, including 5,298 participants. We found a small to medium favorable effect on distress indicators (SMD=–0.45, 95% CI: –0.56 to –0.34) and a small beneficial effect on positive mental health outcomes (SMD=0.31, 95% CI: 0.14‐0.47), which both remained significant at follow‐up assessment and were robust in sensitivity analyses. However, our analyses pointed to substantial between‐study heterogeneity, which was only partially explained by moderators, and the certainty of evidence was very low across all outcomes. These results provide evidence for the effectiveness of PM+ and SbS in reducing distress indicators and promoting positive mental health in populations exposed to adversities, but a larger high‐quality evidence base is needed, as well as research on participant‐level moderators of the effects of these interventions, their suitability for stepped‐care programs, and their cost‐effectiveness.

In recent years, almost all societies have been exposed to an increasing number of crises (e.g., humanitarian and health emergencies), with low‐ and middle‐income countries often being hit harder. This has resulted in a large number of people experiencing significant distress and being at risk to develop mental disorders such as depression, anxiety and post‐traumatic stress disorder (PTSD) ¹ , ² , ³ . Thus, developing and evaluating interventions to prevent and treat mental distress and to promote positive mental health in populations exposed to adversities is recognized as a priority for global health research ⁴ .

The World Health Organization (WHO) has developed a series of scalable psychosocial interventions for adults impaired by distress in communities exposed to adversities, with a special focus on low‐and middle‐income countries ⁵ . In 2015, Problem Management Plus (PM+) has been proposed as a low‐intensity transdiagnostic intervention for adults suffering from mental distress and self‐identified practical problems ⁶ . Being transdiagnostic in nature, PM+ aims at targeting the shared underlying factors of mental disorders (e.g., deficits in stress management, low use of social resources) ⁷ and promoting general strategies relevant for the prevention and treatment of these disorders (e.g., problem management, behavioral activation, use of social support) ⁶ , ⁸ . The five‐session program, with approximately 90 min per session, can be delivered by trained non‐specialist helpers in individual or group face‐to‐face settings.

Step‐by‐Step (SbS) was initially developed as a guided online self‐help version of PM+ ⁹ . However, as the problem management component of PM+ could not be adapted successfully for the online version, SbS specifically focuses on behavioral activation as a core strategy to reduce depressive symptoms. Consequently, SbS is not transdiagnostic as PM+. Behavioral activation, however, is supplemented by other strategies also included in PM+, such as stress management and promotion of social support, which are effective to reduce depressive symptoms. The five‐session online intervention uses a narrative with a customizable character who visits a health professional to seek help for depression. Each session lasts 20 to 30 min and is guided by a trained non‐specialist e‐helper who supports the engagement with self‐help materials.

Since the WHO has released these interventions ⁶ , ⁹ , many trials have been performed to examine their effects in the context of heterogeneous situations of high and prolonged stress (e.g., involuntary displacement ¹⁰ , armed conflicts and war ¹¹ , natural disasters ¹² , health stressors such as the COVID‐19 pandemic ¹³ ). The STRENGTHS project, an international research network which received funding from the European Union, worked on scaling up PM+ programs and examining their effectiveness for refugee populations ¹⁴ .

To date, there is only one individual participant data meta‐analysis on PM+ ¹⁵ , which examined the effects of the intervention on PTSD symptoms, reporting that PM+ reduced re‐experiencing and avoidance, while effects were smaller for hyperarousal. However, this analysis only included three trials and solely focused on PTSD symptoms. A systematic review of the effects of PM+ and SbS, summarizing the whole body of evidence and potentially moderating factors, is still missing. The current study aimed to address this gap.

METHODS

This systematic review adheres to the standards of the Cochrane Collaboration ¹⁶ , and its results are reported in line with the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) ¹⁷ . The review was pre‐registered (ID: CRD42022367698) and the protocol was made available in the Open Science Framework (no. 10.17605/OSF.IO/4Q53C).

Search strategy and data extraction

As PM+ was introduced in 2015, databases were searched from January 1, 2014, with the search being lastly updated on March 8, 2023. Searches were performed in the American Psychological Association (APA) PsycNET (including PsycInfo, PsycArticles, PsycExtra), the Cochrane Central Register of Controlled Trials (CENTRAL), Embase, Scopus and Web of Science.

Search terms comprised two clusters: those related to PM+ and SbS as interventions of interest, and those related to study design. If applicable, we used Cochrane high‐sensitive search filters for the identification of randomized controlled trials ¹⁸ , as well as Medical Subject Headings and Emtree terms (see also supplementary information). Additionally, we checked the reference lists of included studies, reports citing studies included in our review based on Google Scholar citation tracking, and the website of the STRENGTHS project ¹⁹ .

Eligible studies were (cluster) randomized controlled trials (RCTs) examining the effects of PM+ or SbS in stress‐exposed populations of all ages. Studies were eligible if PM+ or SbS was delivered as initially proposed ⁶ , ⁹ or amended with additional components (e.g., targeting alcohol consumption ²⁰ or emotion processing ²¹ ), but ineligible if they examined stepped‐care programs employing PM+/SbS as second step. All comparators were eligible, including waitlist, (enhanced) care‐as‐usual, and active control conditions.

Eligible studies assessed at least one of the following outcomes: distress (i.e., general distress; anxiety, depressive or PTSD symptoms; functional impairment, self‐identified problems); positive mental health (i.e., well‐being, quality of life, social support/relationships); somatic symptoms; family distress/functioning (e.g., child mental distress), and health care costs/use.

After de‐duplication in Zotero, titles, abstracts and full texts were assessed by two reviewers independently in Rayyan ²² . Inter‐rater reliability was almost perfect at title/abstract level (kappa = .97) and substantial at full text level (kappa = .78). At both stages of screening, disagreements were resolved through discussion or by consulting a third reviewer.

We developed a customized data extraction sheet for this review. All descriptive data of eligible primary studies were extracted by one reviewer and checked by a second. Any disagreements were resolved through discussion or consultation of a third reviewer.

Quality appraisal

Two team members independently evaluated the risk of bias of primary studies using the Cochrane risk‐of‐bias tool for randomized trials (RoB 2) ²³ , which assesses the following bias domains: randomization process; deviations from the intended intervention; missing outcome data; outcome measurement; and selection of reported results. For cluster RCTs, we additionally assessed risk of bias due to identification/recruitment of participants. Bias ratings were assessed at single outcome and overall study levels. Judgements could be “low”, “high”, or express “some concerns”.

We examined a potential publication bias statistically by approximating rank correlation tests ²⁴ and using visual inspections of contour‐enhanced funnel plots. Rank correlation tests are available for multilevel models by including sampling error as moderator ²⁵ . If the sampling error significantly predicts effect sizes, this can be interpreted as evidence for a publication bias.

Data synthesis

Eligible studies were summarized narratively and in tabular form. Pairwise meta‐analyses were performed for primary outcomes if more than two effect estimates were available per outcome type (e.g., PTSD symptoms) and assessments were sufficiently homogeneous. For other outcomes, we provided a brief qualitative summary. In cases where data needed for effect size calculation were missing or unclear, primary study authors were contacted by the review team via email (see supplementary information).

Meta‐analyses were performed in R version 4.2.3 ²⁶ using the packages metafor ²⁷ and clubSandwich ²⁸ . All analyses used random‐effects models and maximum likelihood estimations with an inverse variance method. Standardized mean differences (SMDs, Hedges’ g) at post‐intervention and follow‐up assessments were used as effect estimates, and their 95% confidence intervals (CIs) as indicators of significance. SMDs were calculated based on means and standard deviations, with positive SMDs indicating unfavorable intervention effects for distress indicators, but favorable intervention effects for positive mental health outcomes. To account for uncertainty of meta‐analytical findings, we calculated 95% prediction intervals (PIs) as an estimate of the range in which 95% of future observations will fall ²⁹ . In cluster RCTs, effect sizes were corrected for clustering effects ¹⁶ (see supplementary information).

We calculated separate models for distress indicators and positive mental health outcomes, as well as for post‐intervention and follow‐up assessments. Exploratively, we examined the stability of intervention effects between post‐intervention and follow‐up assessments by means of two‐way random‐effects intra‐class correlations (ICCs). We used multivariate multilevel models nesting effect estimates within studies (outer factor) and outcome types (inner factor) ³⁰ . Cluster‐robust tests and CIs were used to account for non‐independent effect estimates.

As little information was available on between‐outcome correlations within studies, covariances were imputed based on a correlation of ρ=0.60, with other correlation estimates being used for sensitivity analyses ³¹ . For each model, we examined whether the use of an unstructured variance‐covariance matrix improved model fit. As this was not the case for any model, symmetric matrices were assumed. Moreover, the specification of multilevel models was examined by calculating profiles of the log‐likelihood, which should show single peaks. In case of evidence for over‐parameterization, standard univariate meta‐analyses were performed. As effects of PM+ and SbS on depressive symptoms were of particular interest, they were examined in additional univariate models for illustrative purposes.

Statistical heterogeneity was assessed using Cochran's Q ³² , with a significant Q indicating the presence of heterogeneity. To quantify the amount of heterogeneity in our analyses, we used the I ² statistic (range: 0‐100%) at single outcome level, with values of 50% and above indicating substantial heterogeneity ¹⁶ .

Due to the substantial heterogeneity in our primary analyses, moderator analyses were performed on distress indicators and positive mental health outcomes (at post‐intervention assessment). For categorical variables (e.g., intervention type) we used subgroup analyses, while meta‐regressions were used for omnibus moderation tests and continuous moderators (e.g., age), with a significant QM test indicating the presence of a moderator effect. All analyses used cluster robust estimations. First, we examined whether intervention effects differed between PM+ delivered in individual settings, PM+ delivered in group settings, and SbS. As we found no evidence for such a difference, additional moderator analyses were performed for all studies. We examined sociodemographic sample characteristics (i.e., age, gender balance per sample), stressor type (i.e., gender‐based violence vs. health stressors vs. humanitarian disasters vs. war or armed conflict), stressor level (i.e., individual vs. collective), duration of intervention (in weeks and minutes), intervention setting (i.e., low‐ or middle‐income vs. high‐income country), and intervention providers (i.e., professionals vs. lay staff) as moderators.

Sensitivity analyses were performed for between‐outcome correlations (ρ=.40, ρ=.80), risk of bias, inclusion of outliers, and context of evaluation (i.e., STRENGTHS project vs. other trials).

The certainty of evidence for specific outcome types (e.g., depressive symptoms) at post‐intervention and follow‐up assessments was evaluated in duplicate using the Grading of Recommendations, Assessment, Development, and Evaluations (GRADE) ³³ .

RESULTS

Search outcome and study characteristics

Our search for primary studies in electronic databases yielded 2,902 eligible records, with 805 duplicates being removed. Of 2,097 records screened at title/abstract level, 97 were assessed at full text level. Five additional eligible records were identified by our searches on websites, citation searching, and Google Scholar citation tracking. Taken together, this resulted in 23 eligible primary studies (from 55 reports) for synthesis (see Figure 1).

Figure 1.

PRISMA flow chart. APA – American Psychological Association

Table 1 presents the characteristics of the 23 studies, comprising 5,298 participants (range of sample sizes: 8 to 680), included in our review ¹⁰ , ¹¹ , ¹² , ¹³ , ³⁴ , ³⁵ , ³⁶ , ³⁷ , ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ , ⁴⁶ , ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ , ⁵² , ⁵³ . These were performed in Pakistan (four studies); Lebanon, Kenya (three studies each); Jordan, The Netherlands, Nepal (two studies each); Australia, Austria, China, Colombia, Switzerland, Turkey and the UK (one study each). Seventeen studies (73.9%) were conducted in low‐to‐middle income countries. Fifteen studies included follow‐up assessments between 3 and 6 months; a longer follow‐up interval of 12 months was only reported for one study ³⁷ . Six studies (26.1%) were performed within the STRENGTHS project and examined effects of PM+ in refugee populations ¹⁰ , ³⁴ , ³⁵ , ³⁷ , ³⁸ , ⁴² , ⁵² .

Table 1.

Characteristics of included studies

Study	Country	Stressor	Population	N	Intervention	Providers	Comparator	Outcomes
Acarturk et al 34	Turkey	Syrian civil war	Syrian refugees (% female: 66.7% in IG, 68.2% in CG)	IG: 24, CG: 22	Group PM+ (+ECAU)	Peer‐/lay‐guided	ECAU	General distress, PTSD symptoms, self‐identified problems
Akhtar et al 35	Jordan	Syrian civil war	Syrian refugees (% female: 68.6% in IG, 72.4% in CG)	IG: 35, CG: 29	Group PM+	Professional guided and/or related educational background	ECAU	Anxiety symptoms, depressive symptoms, general distress, PTSD symptoms, self‐identified problems
Bryant et al 36	Kenya	Gender‐based violence	Women with experience of gender‐based violence	IG: 209, CG: 212	PM+	Peer‐/lay‐guided	ECAU	Functional impairment, general distress, PTSD symptoms, self‐identified problems
Bryant et al 13	Australia	COVID‐19 pandemic	People residing in Australia (% female: 84.2% in IG, 83.3% in CG)	IG: 120, CG: 120	PM+ (adapted to pandemic context)	Professional guided and/or related educational background	ECAU	Anxiety symptoms, depressive symptoms
Bryant et al 37 , 38	Jordan	Syrian civil war	Syrian refugees (% female: 71.1% in IG, 75.2% in CG)	IG: 204, CG: 206	Group PM+	Professional guided and/or related educational background	ECAU	Anxiety symptoms, depressive symptoms, functional impairment, PTSD symptoms, self‐identified problems
Cuijpers et al 39	Lebanon	Humanitarian crises (mixed stressors)	People residing in Lebanon (% female: 72.5% in IG, 67.3% in CG)	IG: 331, CG: 349	SbS	Minimally guided by e‐helpers	ECAU	Anxiety symptoms, depressive symptoms, functional impairment, PTSD symptoms, self‐identified problems, well‐being
Cuijpers et al 40	Lebanon	Syrian civil war	Syrian refugees (% female: 61.7% in IG, CG 55.2% in CG)	IG: 283, CG: 286	SbS	Minimally guided by e‐helpers	ECAU	Anxiety symptoms, depressive symptoms, functional impairment, PTSD symptoms, self‐identified problems, well‐being
Dawson et al 41	Kenya	Urban adversity, gender‐based violence	Women residing in Kenya	IG: 35, CG: 35	PM+	Peer‐/lay‐guided	ECAU	Functional impairment, general distress, PTSD symptoms
De Graaff et al 10	The Netherlands	Syrian civil war	Syrian refugees (% female: 60.0% in IG, 60.0% in CG)	IG: 30, CG: 30	PM+ (+CAU)	Professional guided and/or related educational background	CAU	Anxiety symptoms, depressive symptoms, functional impairment, general distress, PTSD symptoms, self‐identified problems
De Graaff et al 42	The Netherlands	Syrian civil war	Syrian refugees (% female: 29.1% in IG, 47.6% in CG)	IG: 103, CG: 103	PM+	Professional guided and/or related educational background	CAU	Anxiety symptoms, depressive symptoms, functional impairment, general distress, PTSD symptoms, self‐identified problems
Dowrick et al 43	UK	Humanitarian crises (mixed stressors)	Refugees/asylum seekers (% female: 100% in IG, 50.0% in CG)	IG: 4, CG: 4	PM+	Peer‐/lay‐guided	CAU	Anxiety symptoms, depressive symptoms, functional impairment, general distress, PTSD symptoms, self‐identified problems, well‐being
Hamdani et al 44	Pakistan	Humanitarian crises (mixed stressors)	People residing in Pakistan (% female: 64.6% in IG, 68.8% in CG)	IG: 96, CG: 96	PM+	Professional guided and/or related educational background	CAU	Anxiety symptoms, depressive symptoms, functional impairment, general distress, perceived social support, PTSD symptoms, self‐identified problems
Heim et al 45	Lebanon	Humanitarian crises (mixed stressors)	Syrian refugees, people residing in Lebanon (% female: 67.2% in IG, 67.6% in CG)	IG: 67, CG: 71	SbS	Minimally guided by e‐helpers	ECAU	Anxiety symptoms, depressive symptoms, functional impairment, PTSD symptoms, self‐identified problems, well‐being
Jordans et al 46	Nepal	Humanitarian crises (mixed stressors)	People residing in Nepal (% female: 82.3% in IG, 82.0% in CG)	IG: 306, CG: 305	Group PM+	Peer‐/lay‐guided	ECAU	Depressive symptoms, functional impairment, general distress, perceived social support, PTSD symptoms
Khan et al 11	Pakistan	Armed conflicts	Women residing in Pakistan	IG: 59, CG: 60	Group PM+	Peer‐/lay‐guided	ECAU	Anxiety symptoms, depressive symptoms, functional impairment, general distress, PTSD symptoms, self‐identified problems
Knefel et al 47	Austria	Humanitarian crises (mixed stressors) in Afghanistan	Afghan refugees/asylum seekers (% female: 38.5% in IG, 60% in CG)	IG: 49, CG: 39	PM+ (adapted)	Professional guided and/or related educational background	CAU	Anxiety symptoms, depressive symptoms, general distress, PTSD symptoms, quality of life, social relationships, self‐identified problems
Nyongesa et al 48	Kenya	Human immuno‐deficiency virus (HIV)	Young people residing in Kenya (% female: 62.9% in IG, 68.6% in CG)	IG: 35, CG: 35	PM+ (adapted)	Peer‐/lay‐guided	ECAU	Anxiety symptoms, depressive symptoms, perceived social support, quality of life
Perera et al 49	Colombia	Humanitarian crises (mixed stressors) in Venezuela	Venezuelan refugees/migrants (% female: 48.7% in IG, 21.2% in CG)	IG: 40, CG: 39	PM+	Peer‐/lay‐guided	Waitlist	Self‐identified problems, quality of life, social relationships, well‐being
Rahman et al 50	Pakistan	Armed conflict	People residing in Pakistan (% female: 75.0% in IG, 82.8% in CG)	IG: 172, CG: 174	PM+	Peer‐/lay‐guided	ECAU	Anxiety symptoms, depressive symptoms, functional impairment, general distress, PTSD symptoms, self‐identified problems
Rahman et al 51	Pakistan	Armed conflicts	Women residing in Pakistan	IG: 306, CG: 306	Group PM+	Peer‐/lay‐guided	ECAU	Anxiety symptoms, depressive symptoms, functional impairment, general distress, perceived social support, PTSD symptoms, self‐identified problems
Sangraula et al 12	Nepal	Earthquakes	People residing in Nepal (% female: 83% in IG, 84% in CG)	IG: 66, CG: 64	Group PM+	Peer‐/lay‐guided	ECAU	Depressive symptoms, functional impairment, general distress, perceived social support, PTSD symptoms
Spaaij et al 52	Switzerland	Syrian civil war	Syrian refugees/asylum seekers (% female: 45.2% in IG, 57.1% in CG)	IG: 31, CG: 28	PM+	Peer‐/lay‐guided	ECAU	Anxiety symptoms, depressive symptoms, functional impairment, PTSD symptoms
Zhang et at53	China	Multiple myeloma	People living with multiple myeloma (% female: 32.5% in IG, 37.5% in CG)	IG: 40, CG: 40	PM+	Professional guided and/or related educational background	CAU	Anxiety symptoms, depressive symptoms, functional impairment, self‐identified problems

PM+ ‐ Problem Management Plus, SbS ‐ Step‐by‐Step, (E)CAU ‐ (enhanced) care‐as‐usual, PTSD – post‐traumatic stress disorder, IG – intervention group, CG – control group

The mean age of participants was 34.9±8.33 years (range: 21.3‐63.2), and 73% of the participants were women (range: 33‐100%). Most populations were exposed to war or armed conflicts (11 studies), followed by humanitarian crises (seven studies), health stressors (three studies, with one study on COVID‐19), and gender‐based interpersonal violence (two studies).

Thirteen studies (56.5%) reported on PM+ in individual settings, seven (30.4%) examined the group version of PM+, and three (13.0%) investigated the SbS intervention. The duration of the intervention ranged between 5 and 26 weeks (average: 6.4±4.5) and 100 and 750 min (average: 501±156), with SbS being shorter than (group) PM+ (about 100 min). Twenty‐two studies (95.7%) used (enhanced) care‐as‐usual, and one ⁴⁹ employed a waitlist control.

Quality appraisal

Only one study ¹² had an overall low risk of bias rating, while risk of bias was high for the remaining 22 studies (95.6%). The main flaws (some concerns or high risk) were found for outcome measurement (post‐intervention: 95.6%; follow‐up: 100%), selection of reported results (post‐intervention: 43.4%; follow‐up: 32.5%), deviations from the intended intervention (post‐intervention: 35.2%; follow‐up: 32.5%), and missing outcome data (post‐intervention: 30.8%; follow‐up: 24.1%). In most cluster RCTs, identification/recruitment of participants was sufficiently described (see also supplementary information).

Meta‐regression models provided no evidence for an association of standard errors and effect estimates at post‐intervention for distress indicators (QM=1.81, p=0.178) and positive mental health outcomes (QM=3.23, p=0.110). However, the visual inspection of contour‐enhanced funnel plots indicated that more effect estimates fell into the significance border areas of the plots (see supplementary information). At follow‐up assessments, regression models provided evidence for a publication bias in the analysis on distress indicators (QM=4.61, p=0.032), but not for positive mental health outcomes (p≥0.657). The contour‐enhanced funnel plots for distress indicators and well‐being or quality of life suggested that more effect estimates fell into the significance border areas (see supplementary information). Thus, our analyses are potentially impacted by publication bias.

Main analysis

Data preparation

Preliminary analyses pointed to considerably larger effect estimates for positive mental health outcomes (SMD ≥ 2.00) in two studies ⁴⁹ , ⁵³ . As those findings biased our results, the respective effect estimates were excluded from further quantitative synthesis.

Effects at post‐intervention

Twenty‐two studies were included in our analysis on distress indicators (see Table 2 and Figure 2). Across all indicators, we found evidence for a small to moderate favorable effect of PM+/SbS over (enhanced) care‐as‐usual (SMD=–0.45, 95% CI: –0.56 to –0.34), with substantial correlations of outcomes within studies (ρ=0.74). Between‐outcome differences were rather small, but accounted for heterogeneity in effect estimates (QM=18.90, p<0.001).

Table 2.

Results of analyses comparing Problem Management Plus (PM+) and Step‐by‐Step (SbS) with (enhanced) care‐as‐usual at post‐intervention

	n	SMD	95% CI	95% PI	p	Q	p(Q)	I2
Distress indicators	22	–0.45	–0.56, –0.34	–0.87, –0.03	<0.001	206.01	<0.001
Anxiety symptoms	16	–0.51	–0.63, –0.39	–0.92, –0.10	<0.001			58.8
Depressive symptoms	18	–0.46	–0.62, –0.30	–0.88, –0.04	<0.001			61.3
Only (group) PM+	15	–0.46	–0.64, –0.29	–1.02, 0.09	<0.001			71.2
Only SbS	3	–0.60	–0.81, –0.39		<0.001			28.9
Functional impairment	16	–0.36	–0.48, –0.23	–0.76, 0.05	<0.001			62.6
General distress	14	–0.55	–0.68, –0.41	–0.96, –0.14	<0.001			58.2
PTSD symptoms	19	–0.34	–0.47, –0.22	–0.75, 0.06	<0.001			62.5
Self‐identified problems	13	–0.51	–0.70, –0.32	–0.94, –0.08	<0.001			59.4
Positive mental health outcomes	10	0.31	0.14, 0.47	–0.07, –0.69	0.003	18.08	0.080
Well‐being and quality of life	6	0.37	0.15, 0.59		0.005			41.2
Social support/relationships	6	0.26	0.12, 0.40		0.002			39.1

SMD – standardized mean difference, 95% PI – 95% prediction interval, PTSD – post‐traumatic stress disorder

Figure 2.

Forest plots of the meta‐analysis on distress indicators at post‐intervention assessment. Negative estimates indicate an effect favoring PM+/SbS over (enhanced) care‐as‐usual. PTSD – post‐traumatic stress disorder.

For all outcome types, we found evidence for small to moderate favorable effects: SMD=–0.51 (95% CI: –0.63 to –0.39) for anxiety symptoms; SMD=–0.46 (95% CI: –0.62 to –0.30) for depressive symptoms; SMD=–0.36 (95% CI: –0.48 to –0.23) for functional impairment; SMD=–0.55 (95% CI: –0.68 to –0.41) for general distress; SMD=–0.34 (95% CI: –0.47 to –0.22) for PTSD symptoms; and SMD=–0.51 (95% CI: –0.70 to –0.32) for self‐identified problems.

Only the PIs for functional impairment and PTSD symptoms included zero. After accounting for between‐outcome differences, there was still evidence for residual heterogeneity (Q=155.08, p<0.001), which was substantial for all outcome types (58.2 ≤ I² ≥ 62.6). Favorable effects on depressive symptoms were larger for SbS (SMD=–0.60, 95% CI: –0.81 to –0.39) than for (group) PM+ (SMD=–0.46, 95% CI: –0.64 to –0.29).

Ten studies were included in our analysis on positive mental health outcomes (see Table 2 and Figure 3). Those provided evidence for a small favorable effect of PM+/SbS over (enhanced) care‐as‐usual (SMD=0.31, 95% CI: 0.14‐0.47), with only moderate heterogeneity (Q=18.08, p=0.080), and no significant between‐outcome differences (QM=2.04, p=0.191). For well‐being and quality of life, there was evidence for small to moderate favorable effects of PM+/SbS over (enhanced) care‐as‐usual (SMD=0.37, 95% CI: 0.15‐0.59). Effect estimates for social support/relationships were favorable but small (SMD=0.26, 95% CI: 0.12‐0.40). Heterogeneity at single outcome level was moderate (I² ≤ 41.2), and overall non‐significant (Q=15.77, p=0.106).

Figure 3.

Forest plots of the meta‐analysis on positive mental health outcomes at post‐intervention assessment. Positive estimates indicate an effect favoring PM+/SbS over (enhanced) care‐as‐usual.

Based on GRADE, the certainty of evidence was very low for all outcome types (see supplementary information).

Effects at follow‐up assessment

Sixteen studies reported on follow‐up data for distress indicators, finding a small favorable effect of PM+/SbS over (enhanced) care‐as‐usual (SMD=–0.33, 95% CI: –0.46 to –0.21) (see Table 3). Again, between‐outcome differences were small, but accounted for a relevant proportion of between‐study heterogeneity (QM=9.19, p=0.001).

Table 3.

Results of analyses comparing Problem Management Plus (PM+) and Step‐by‐Step (SbS) with (enhanced) care‐as‐usual at follow‐up (3‐6 months)

	n	SMD	95% CI	95% PI	p	Q	p(Q)	I2
Distress indicators	16	–0.33	–0.46, –0.21	–0.77, 0.10	<0.001	163.93	<0.001
Anxiety symptoms	13	–0.40	–0.54, –0.25	–0.84, 0.05	<0.001			64.0
Depressive symptoms	14	–0.36	–0.58, –0.14	–0.83, 0.11	0.005			66.3
Only (group) PM+	11	–0.33	–0.54, –0.13	–0.92, –0.25	0.001			76.7
Only SbS	3	–0.58	–0.76, –0.40		<0.001			0
Functional impairment	13	–0.27	–0.44, –0.10	–0.72, 0.18	0.005			67.5
General distress	9	–0.44	–0.63, –0.25		<0.001			64.6
PTSD symptoms	14	–0.29	–0.47, –0.11	–0.74, 0.17	0.006			67.4
Self‐identified problems	10	–0.27	–0.43, –0.10	–0.72, 0.18	0.005			65.9
Positive mental health outcomes
Well‐being and quality of life	5	0.52	0.35, 0.69		<0.001	1.03	0.906	0
Social support/relationships	4	0.22	0.08, 0.36		0.002	0.77	0.857	0

SMD – standardized mean difference, 95% PI – 95% prediction interval, PTSD – post‐traumatic stress disorder

For all outcome types, we found evidence for small to moderate favorable effects: SMD=–0.40 (95% CI: –0.54 to –0.25) for anxiety symptoms; SMD=–0.36 (95% CI: –0.58 to –0.14) for depressive symptoms; SMD=–0.27 (95% CI: –0.44 to –0.10) for functional impairment; SMD=–0.44 (95% CI: –0.63 to –0.25) for general distress; SMD=–0.29 (95% CI: –0.47 to –0.11) for PTSD symptoms; and SMD=–0.27 (95% CI: –0.43 to –0.10) for self‐identified problems.

After accounting for between‐outcome differences, there was evidence for residual heterogeneity (Q=133.35, p<0.001), which was substantial for all outcome types (64.0 ≤ I² ≥ 67.5). Effect estimates were strongly correlated within studies (ρ=0.78). Beneficial effects on depressive symptoms were larger for SbS (SMD=–0.58, 95% CI: –0.76 to –0.40) than for (group) PM+ (SMD=–0.33, 95% CI: –0.54 to –0.13).

Findings on positive mental health outcomes at follow‐up assessment were reported in eight studies (see Table 3). Separate univariate models were calculated, as multilevel models showed inacceptable fit. In absence of between‐study heterogeneity, there was evidence for small to moderate favorable effects of PM+/SbS over (enhanced) care‐as‐usual for quality of life and well‐being (SMD=0.52, 95% CI: 0.35‐0.69). Effect estimates for social support/relationships were favorable but small (SMD=0.22, 95% CI: 0.08‐0.36).

The GRADE assessment indicated an overall very low certainty of evidence for all outcome types (see supplementary information).

ICCs indicated substantial stability of intervention effects from post‐intervention to follow‐up assessments for both outcome categories: ICC=0.85 (95% CI: 0.77‐0.90) for distress indicators; ICC=0.88 (95% CI: 0.53‐0.98) for positive mental health outcomes.

Moderator analyses

Moderator analyses were performed for distress indicators at both timepoints and for positive mental health outcomes at post‐intervention assessment, as our models indicated no heterogeneity at follow‐up assessment.

At post‐intervention assessment, we did not find evidence for overall differences between PM+, group PM+ and SbS for either distress indicators (QM=1.33, p=0.287) or positive mental health outcomes (QM=2.37, p=0.306). For distress indicators, interventions with longer duration (in weeks) showed larger favorable effects (QM=5.82, p=0.026). However, this finding was mainly driven by one study ⁴⁸ extending the delivery from mostly 5 to 10 weeks, which reported large favorable effects. We found no evidence for other moderator effects (see supplementary information).

At follow‐up assessment, we found no differences between PM+, group PM+ and SbS for distress indicators (QM=2.38, p=0.132). We found a moderator effect of age, with older age being associated with less favorable effects of PM+/SbS (QM=12.24, p=0.004). Interventions with longer duration (in minutes) had smaller favorable effects at follow‐up assessments (QM=7.37, p=0.022). There was no evidence for other moderator effects (see supplementary information).

Sensitivity analyses

Sensitivity analyses on between‐outcome correlations showed that the use of weaker or stronger correlations (ρ=.40, ρ=.80) had no impact on our conclusions (see supplementary information).

To account for a potential impact of risk of bias within studies, we re‐ran all analyses limited to the studies at low risk of bias for the respective bias domain. Neither at post‐intervention nor at follow‐up assessments, results were significantly different (see supplementary information).

Including effect estimates identified as outliers ⁴⁹ , ⁵³ increased the range of CIs and PIs, but did not change our overall conclusions. Studies conducted within the STRENGTHS project did not differ from the others (see supplementary information).

Effects on other outcomes

Seventeen studies examined adverse events, with nine reporting no adverse events during (group) PM+ or SbS, six reporting events unlikely to be related to the intervention (e.g., hospitalization due to physical illness), and two ¹² , ⁴² reporting adverse events in the intervention group (i.e., hospitalizations, suicide ideation), which, however, were equally likely in the (enhanced) care‐as‐usual arm (see supplementary information).

Other outcome categories were only examined in a small number of studies. Two studies ³⁵ , ³⁷ , ³⁸ investigated effects of group PM+ on child mental distress, with heterogeneous results. However, one study ³⁸ found favorable symptom changes in children to be associated with more consistent disciplinary behavior in parents who received group PM+. Two studies ⁴⁶ , ⁴⁷ investigated effects on somatic symptoms, both finding no evidence for favorable effects of (group) PM+. Health care costs and health care utilization were examined in three studies ¹⁰ , ³⁶ , ⁵² , with none of them finding evidence for favorable effects. Hamdani et al ⁵⁴ examined the cost‐effectiveness of PM+ for reducing mood and anxiety disorders in Pakistan, based on data from Rahman et al ⁵⁰ , and found PM+ to be more effective but also more costly than (enhanced) care‐as‐usual.

DISCUSSION

This is the first systematic review and meta‐analysis of the effects of the scalable psychosocial interventions PM+ and SbS, that were developed by the WHO to address an increasing need for mental health care in times of intensified humanitarian crises, especially in low‐ and middle‐income countries. Based on 23 studies, including 5,298 participants, we found evidence for small to moderate favorable effects of these interventions, compared with (enhanced) care‐as‐usual, on distress indicators and positive mental health outcomes. These effects remained significant across all outcome types at short‐term follow‐up 3 to 6 months after the end of the intervention, and were robust in sensitivity analyses.

Even though favorable effects of PM+ and SbS were found consistently across all outcome types, effect estimates were the largest for general distress at both timepoints (SMD=–0.55 to –0.44), which is in line with other transdiagnostic interventions ⁵⁵ and may support the transdiagnostic nature of PM+. We found no evidence for overall differences between PM+, group PM+ and SbS; however, only three studies ³⁹ , ⁴⁰ , ⁴⁵ (with 1,387 participants) delivered SbS. Future reviews will have to examine whether SbS more specifically targets depressive symptoms, as proposed by primary studies on SbS ³⁹ , ⁴⁰ and differences from (group) PM+ at both timepoints in our analyses.

Effects for PTSD symptoms were smaller (SMD=–0.34 to –0.29), which may suggest that more specific interventions (e.g., targeting core symptoms of PTSD such as intrusive memory ⁵⁶ ) might be more suitable to reduce post‐traumatic stress. However, as studies examined heterogeneous stressors (including wars/armed conflicts ³⁴ and health stressors such as the COVID‐19 pandemic ¹³ ), not all stressors may have evoked PTSD symptoms, which may also account for smaller effect sizes.

Based on a smaller number of studies that examined positive mental health outcomes, there was a trend towards lower effect estimates for these outcomes, especially for social support and social relationships. Given the importance of positive mental health ⁵⁷ , future studies on PM+/SbS should include such measures and may answer the question of whether effects on these outcomes emerge over longer time periods, or remain small as many people continue to live under adverse circumstances during PM+/SbS delivery (e.g., in refugee camps ³⁵ ).

Overall, intervention effects were small to moderate and tended to be smaller at follow‐up assessments. Given the high symptom burden after exposure to severe stressors such as humanitarian crises, wars or armed conflicts ¹ , ⁵⁸ , ⁵⁹ , at least a proportion of the affected people will need additional mental health care. In line with this, PM+ and SbS have been proposed as components of stepped‐care approaches that provide effective evidence‐based treatments with the least resources ⁶⁰ . Our review shows that PM+ and SbS have the potential to constitute effective components of such programs. However, future studies will have to examine the combination of PM+ and SbS with less intensive self‐help programs (e.g., Doing What Matters in Times of Stress ⁶¹ by the WHO) and more intensive standard care ⁶⁰ . Together with further research on the cost‐effectiveness of PM+ and SbS, such studies can pave the way for establishing PM+ and SbS as basic interventions for stress‐exposed populations.

The results of this review should be considered in the light of some limitations. First, we found moderate to considerable heterogeneity in all analyses, except for positive mental health outcomes at follow‐up assessment, which could not be fully accounted for by study‐level moderators. Second, we found evidence for a potential impact of publication bias; the overall risk of bias was high; and the certainty of evidence was very low for all outcomes. Our results remained robust in sensitivity analyses, but we cannot exclude that future studies may change effect sizes. Third, although a systematic review was highly needed at this time, the literature search showed that about 23 trials are still ongoing. Based on our sensitivity analyses, we believe that these trials are unlikely to change substantially the scenario we described. However, this systematic review should be updated when a larger high‐quality evidence base becomes available. Fourth, for some moderator levels, only a small number of effect estimates was available, and some analyses provided inconsistent findings (e.g., age, duration of intervention), which need further replication.

Based on all available evidence so far, we conclude that PM+ and SbS are effective programs to reduce distress and promote positive mental health in populations exposed to adversities. Favorable intervention effects remain significant during short‐term follow‐up periods. Future individual participant data meta‐analyses ⁶² may shed light on participant‐level moderators of intervention effects and help to clarify for whom PM+ and SbS are most effective. If further research provides support for the cost‐effectiveness of PM+ and SbS, and their suitability for stepped‐care programs, both WHO interventions can help to reduce the negative mental health consequences of current and future global crises.