Article Text
Abstract
Background Children represent nearly 40% of forcibly displaced populations and are subject to stressors that affect well-being. Little is known about the effects of interventions to enhance psychological resilience in these children, outside clinical settings.
Methods We conducted a systematic review, following Cochrane methods. Eligible studies tested resilience-enhancing interventions outside clinical settings in forcibly displaced children/adolescents. We included longitudinal quantitative studies with comparator conditions irrespective of geographical scope or language. We searched articles published between January 2010 and April 2020 in PubMed, Embase, Cochrane Library, Web of Science, PsycINFO and the WHO’s Global Index Medicus. To standardise effect sizes across the different reported outcomes, we transformed reported mean differences to standardised mean differences using Hedge’s g statistic with associated 95% CI. We pooled data for meta-analysis where appropriate. We used Cochrane tools to assess study risk of bias and used Grading of Recommendations Assessment, Development and Evaluation to determine evidence quality for meta-analysed outcomes.
Results Searches yielded 4829 results. Twenty-three studies met inclusion criteria. Studies reported 18 outcomes measured by 48 different scales; only 1 study explicitly measured resilience. Eight studies were randomised controlled trials; the rest were non-randomised pre–post studies. Interventions were diverse and typically implemented in group settings. Studies reported significant improvement in outcomes pertinent to behavioural problems, coping mechanisms and general well-being but not to caregiver support or psychiatric symptoms. In meta-analysis, resilience was improved (gav=0.194, 95% CI 0.018 to 0.369), but anxiety symptoms and quality of life were not (gav=−0.326, 95% CI −0.782 to 0.131 and gav=0.325, 95% CI −0.027 to 0.678, respectively). Risk of bias varied. Quality of evidence for most graded outcomes was very low.
Conclusions The multiplicity of study designs, intervention types, outcomes and measures incumbered quantifying intervention effectiveness. Future resilience research in this population should use rigorous methods and follow reporting guidelines.
PROSPERO registration number CRD42020177069.
- Systematic review
- Child health
- Health policy
- Mental Health & Psychiatry
- Paediatrics
Data availability statement
All data relevant to the study are included in the article or uploaded as online supplemental information.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
Forcibly displaced persons are a growing population globally and children (ages 0–18) comprise 40% of this group.
Most (80%) of these children experience psychological problems in conjunction with trauma endured before, during and/or after forced displacement.
Clinical interventions can improve outcomes associated with child mental health and well-being. However, forcibly displaced populations may not be able to access clinical settings.
WHAT THIS STUDY ADDS
Studies reported improvement in behavioural problems, coping mechanisms and general well-being but not in caregiver support or psychiatric symptoms.
In meta-analysis, resilience was improved but anxiety symptoms and quality of life were not.
Variation in data collection methods across studies precluded further meta-analysis.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
Non-clinical interventions, including those delivered by lay practitioners, offer scalable methods to improve many resilience outcomes among forcibly displaced children.
Future research would benefit from guidance on reporting and use of standardised measurement scales.
Background
A forcibly displaced population (FDP) is defined by the International Organization for Migration as, ‘Persons or groups of persons who have been forced or obliged to flee or to leave their homes or places of habitual residence, either across an international border or within a state, in particular as a result of or in order to avoid the effects of armed conflict, situations of generalised violence, violations of human rights or natural or human-made disasters’.1
Psychological sequelae are a major effect of forced displacement, an increasingly prevalent experience that disproportionately affects young people. In 2010, 40 million people worldwide were estimated to be forcibly displaced, a figure that nearly doubled to 79.5 million by the end of 2019.2 Around 40% of forcibly displaced individuals in 2019 were below age 18.2
An estimated 80% of forcibly displaced children experience psychological problems.3 4 For these children, premigration traumas of exposure to violence and deprivation are reinforced by extreme hardship experienced during displacement and challenges following arrival in the host location. For example, children may become separated from their caregivers, which increases risk of exposure to sexual and physical violence, poor nutrition and other resource deprivation. Postmigration, children and their families may experience discrimination, impeded access to resources, acculturation challenges and elevated family conflict.5
While resettlement may offer some short-term relief for forcibly displaced children, it is often associated with exposure to a range of other adversities such as discrimination, social marginalisation, economic struggles, language barriers and loss of status. This can be further compounded by a phase of grief, which may cause deep traumas to resurface.5–7 Such stressors become further aggravated among unaccompanied minors, who become vulnerable to additional trauma-inducing events such as child labour, kidnapping or exploitation by drug dealers, human traffickers and militia.4
Given the elicited variability and complexity of trauma-inducing factors in forcibly displaced children, ‘childhood adversities’ in this context not only encompass the commonly cited causes of psychological trauma in childhood (eg, neglect, abuse, household dysfunction) but also the effects of exposure to armed conflicts. This includes but is not limited to family separation, witnessing murders and exposure to bombing and shelling.8 9
Practitioners largely agree that resilience, defined as ‘the ability to maintain stable, healthy psychological and physical functioning despite exposure to trauma,’ is a key lever for mitigating morbidities associated with childhood trauma.6 10 Nevertheless, many of the conceptual dimensions used to describe and measure resilience are still widely debatable, starting with its own definition.11 While some researchers think of resilience as a dynamic lifelong process, others see it as an outcome of different personal traits.10 However, one attribute that most researchers agree on is that resilience does not exist in a dichotomous ‘all or none’ form, but is rather present in individuals to varying degrees and is evidence of a compilation of strengths.12 Resilience is conventionally measured through composite assessments addressing one or more of: cognitive ability, psychological strength, self-esteem, social skills, respect for others, engagement in hobbies, feelings of hope and control, good peer relationships, feelings of safety and/or consistency in behaviour.12 13 Quantifying resilience is inherently challenging as the context of adversity often means that deterioration in well-being is expected and a favourable outcome could potentially be reflected in the absence of a change in related indicators, rather than positive change. This is challenging to prove outside of an randomised controlled trial (RCT). Additionally, where study populations have all experienced trauma at baseline, evidence of improvement is necessary to inform strategies for mitigating the effects of that trauma, even if populations remain at risk for further adversities.
Although effective under optimal conditions, experience illustrates practical shortcomings of clinical settings as venues for resilience-enhancing interventions serving FDP, such as lack of services, too few clinicians proficient in relevant languages, distance between FDP residences and service-delivery locations and cost.14 Research also shows, however, that non-clinical settings can facilitate effective resilience-enhancing interventions that deliver evidence-based programming at accessible venues (eg, school, religious institution), often via trained lay workers, a strategy that simultaneously addresses language and fiscal barriers.14–20
Yet, the best approaches to mitigate childhood adversities and develop resilience in forcibly displaced children remain unclear. Most research has focused on psychopathology rather than factors linked to improved resilience outcomes in children.20 Additionally, most evidence focuses on interventions using credentialed professionals, who are too limited in number to meet the need among FDP, particularly in low-resource settings.14 21 Further, there is a general paucity of research on the effect of childhood adversities on younger children due to logistical and ethical factors, including challenges to obtaining consent and fear, mistrust or suspicion felt by caregivers.4 22 Finally, established models of trauma-informed parenting do not recognise the fact that parents of children experiencing adversities in conflict zones are traumatised themselves. This can render parents unable to meet their basic parenting responsibilities; studies have shown that parents exposed to extreme hardship or those who suffer from mental health conditions such as depression become less emotionally responsive and withdrawn from their children, which can lead to intrusive and abusive parenting.4 23 24 Collectively, these factors result in a knowledge gap around effective, accessible, realistic strategies for enhancing resilience among forcibly displaced children.
Rationale for review
Prior systematic reviews have focused on specific outcomes, settings, intervention types and/or high-income countries.25–30 Although useful, these do not provide actionable information for decision makers in low-resource settings nor a holistic analysis of the global evidence on psychological resilience-enhancing interventions for forcibly displaced children. This review aims to answer the question, ‘among forcibly displaced children and adolescents (ages 3–17 years) or their caregivers, what is the effect of psychological resilience-enhancing interventions offered outside clinical settings (compared with no intervention, an alternative resilience-enhancing intervention or standard of care) in terms of improved resilience or improved resilience-protective factors, as measured with validated scales?’
Methods
We followed Cochrane methods and Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidance.31 32 The study protocol was registered in the international prospective register of systematic reviews.
We developed a search strategy using indexing terms and keywords related to our inclusion criteria based on scoping searches run in PubMed. We searched PubMed, Embase, PsycINFO, Cochrane Library, Web of Science and the WHO Global Index Medicus for studies between 1 January 2010 and 15 May 2020. We excluded studies published prior 2010 because the global refugee crisis started in 2011 and our resources limited our project timeline. See online supplemental appendix 1 for complete search strategies. We used EndNote V.X9 software33 to remove duplicate records.
Supplemental material
Eligible studies were quantitative, interventional, randomised or non-randomised studies with comparators, published in any language, conducted to assess the impact of any intervention designed to develop or enhance resilience (or an associated factor) in forcibly displaced children and adolescents aged between 3 and 17 years at the time of intervention. We defined psychological resilience-enhancing interventions as those that aimed to improve resilience—as defined by investigators—or any of the modifiable factors associated with it as improvement in; psychological strength, self-esteem, social skills and interaction in addition to respect for others, hobbies, feelings of hope, willingness to accept support, feelings of control, good peer relationship, feelings of safety and consistency in behaviour.12 13 Interventions could be directed to children or to their caregivers. We did not restrict geography, country-income level, type of comparator, or follow-up duration, or type outcomes measured.
We excluded studies requiring a threshold of severity of psychological disease in their study populations. We also excluded studies conducted in populations exposed to war but not displaced and those displaced due to natural disasters. We further excluded populations of child soldiers, torture survivors and sexually abused children. The decision to exclude those specific populations was made after consulting a subject matter expert (SG) who considered such trauma and subsequent interventions to be very specific and not generalisable to the target population within the scope of the study. The decision to exclude populations exposed to war but not displaced emerged from the understanding that displacement adds further challenges to the experience of political violence. Those challenges affect both the types of war-related psychological trauma as well as the types of interventions that can be implemented. The complete inclusion and exclusion criteria for this systematic review are summarised in online supplemental appendix 2.
Supplemental material
Eligible outcomes included effect size of change in psychological resilience as defined by authors and measured by recognised, validated scales. We also included secondary outcomes, also measured with recognised, validated scales, measuring change in factors research has shown to be associated with resilience (cognitive ability, psychological strength, self-esteem, social skills, respect for others, hobbies, feelings of hope and control, good peer relationships, feelings of safety and consistency in behaviour).12 13 Measurements could assess resilience enhancement in the short term or long term, which we defined as within or beyond 3 months of the end of the intervention, respectively.
Two authors (AT and SG) reviewed the titles of deduplicated records and excluded clearly irrelevant titles. Both authors then independently reviewed abstracts for remaining records and excluded those they deemed ineligible. Following that, two authors (AT and SG) independently examined the full text of all potentially eligible studies, reconciling disagreement via discussion and/or the arbitrating third author (MM).
We developed and pilot-tested data extraction templates in Microsoft Excel.34 We collected data on study design, setting, sample size and participant demographic characteristics including nationality, intervention characteristics and assessed outcomes. Data extraction was done by two authors (AT and SG) who are fluent in English. One author (AT) extracted data from each study; a second author (SG) verified the extracted data against source documents. All studies were in English except one, which was published in German; data abstraction of this study was done by a faculty colleague fluent in German.
We used two Cochrane instruments to assess the risk of bias in included studies: the Revised Risk of Bias Tool (ROB-2) for RCTs and ROBINS-I for non-randomised studies.31 ROB-2 domains included: selection bias, reporting bias and general sources of bias. ROBINS-I domains included: confounding bias, selection bias, classification bias, bias due to deviation from interventions, missing data bias, measurement bias and reporting bias. For each study, we assigned a rating of high, low or unclear risk of bias for each of the applicable instrument’s domains.
Initial scoping searches determined that relevant studies typically report effects in terms of the mean difference (MD). To standardise effect sizes across different outcomes and measurement scales, we transformed the reported MD to the standardised MD (SMD) using Hedge’s g statistic.35 Analysis was performed by one review author (AT). Where studies had a two-group, pretest and post-test design, we used Morris’s methods (the difference of differences) as a reference for calculating the SMD.36
Where studies reported outcomes as mean scores on assessment instruments, we converted mean scores to SMDs. Where studies reported the significance of difference between baseline and post-intervention scores rather than the actual mean scores, we contacted study authors to obtain additional data in order to calculate SMDs. Where studies reported SE or 95% CIs rather than SD, we back-calculated SDs from reported data points. Where validated scales showed positive effect size with negative scores or vice versa, we inverted these scales to standardise the direction of effect size reporting to present improvement in desirable outcomes with positive scores and increases in undesirable outcomes with negative scores.
Where there was missing information, ambiguity or discrepancies in manuscripts, we conducted additional calculations from study data, where provided; identified/reviewed publications associated from the same study and/or contacted study authors. When none of these strategies resulted in adequate data, we performed descriptive analysis only.
Given the variety of study designs, interventions and outcome types within the scope of the review, we expected methodological, clinical and statistical heterogeneity. We used the χ2 test of homogeneity to assess heterogeneity and the I2 statistic to guide our choice of meta-analytic models.37 We let I2 guide our choice of meta-analysis model: I2<40 (fixed effect), I2 (40%–80%): random-effects mode and explore heterogeneity, and subgroup analysis, and >80% we did not pool data.
Following data extraction but prior to meta-analysis, we grouped similar interventions by type, setting and intensity of intervention to explore potential categorisations of the reported outcome data. We conducted meta-analysis across groups of studies sharing comparable populations, interventions and outcomes, using Open Meta-Analyst software.38 Where the same outcomes were reported by different types of study participants (eg, youth vs caregivers), we used data reported by the group with the greatest number of participants across relevant studies. To address heterogeneity in pooled effect sizes, we conducted subgroup analyses for studies with similar intervention-content domains, participant ages, intervention settings, personnel training levels or intervention intensities.
Where we considered two analytical models for the same comparison, we ran sensitivity analysis to quantify the difference. Where there were stark differences across models, we interpreted the results with caution and recommend further investigation or research.
For anxiety, depression and post-traumatic stress disorder (PTSD), we reported results in the context of statistical significance (based on the CIs) and clinical significance. Clinical significance indicated a change in mean scores relative to the clinical threshold for diagnosis with the relevant measurement scale. We reported a clinical significance value of ‘yes’ where mean scores crossed the clinical threshold to achieve subclinical scores of undesirable outcomes and ‘improved’ where scores improved but did not reach the non-clinical threshold of undesirable outcomes. We reported ‘worsened’ where either: (1) scores were below the threshold for diagnosis before treatment but increased after treatment and crossed the clinical threshold or (2) mean scores became less desirable but did not cross a clinical threshold. Where baseline scores were below the threshold for diagnosis and increased to remain subclinical, we reported a value of ‘no’.
We used the Grading of Recommendations Assessment, Development and Evaluation (GRADE) methodology39 to assess the quality of the overall body of evidence for meta-analysed outcomes. We rated the quality of evidence for each outcome as high, moderate, low or very low for the following domains: inconsistency among study results, indirectness of effect measurement, imprecision of effect estimates and the risk of publication bias.
Patient and public involvement was not a component of this project. As a systematic review, patients were not directly involved in this research. Resource constraints precluded public involvement in study design, execution and dissemination.
Results
Electronic searches yielded 4829 results. After removing 1758 duplicates, we screened 3071 records. 276 articles passed title screening and 87 passed abstract screening. We reviewed the full text of those 87 studies and excluded 64. Twenty-three studies met eligibility criteria and are included in this review (figure 1).
Reasons for study exclusions after full text review were: population was exposed to war but not displaced (n=50); study inclusion criteria required a certain severity of psychiatric symptoms (n=7); children of FDP were born after resettlement (n=3); interventions targeted adult FDPs (n=2); intervention conducted in a clinical setting (n=1) and intervention was not specific (n=1). Online supplemental appendix 3 reports rationale for each excluded study.
Supplemental material
Eight40–47 included studies were RCTs; 1548–62 were non-randomised single-group pre–post studies. Studies were diverse in intervention setting, population age group, intervention type, intervention form and intensity. Studies were conducted in 16 countries, with settings including schools (n=6), refugee camps (n=4), community centres (n=3), units of unaccompanied minors (n=3), homes (n=1), online (n=1), and unspecified or mixed venues (n=5). Ages of the involved children varied, and we categorised age groups into younger than 12 years (n=2), age 12–18 years (n=7), and, broadly, ‘younger than 18 years’ where data were not disaggregated between children and adolescents (n=13) or unspecified (n=1). Intervention content typically involved multiple domains and included psychosocial skills (n=22), family therapy (n=6), parenting skills (n=6) and art therapy (n=10). Interventions were typically implemented in a group setting (n=17) rather than individual sessions (n=3) or mixed individual and group meetings (n=3) and had varying intensities, most commonly between 6 and 9 sessions (n=8) (table 1). Ten studies reported on interventions implemented with professional mental health practitioner, sometimes in conjunction with non-professionals; 12 involved non-professional mental health interventionists (eg, teachers, lay workers).
Outcomes
Studies reported a total of 77 result measurements addressing 18 distinct outcomes assessed with 1 or more of 48 different scales. Ten studies reported separate estimates for the same outcome based on child-reported and caregiver-reported data. Effect sizes for eight results could not be calculated.
To prepare for meta-analysis, we organised outcomes into six categories: resilience (comprised ‘resilience’ (n=1), child psychosocial protective factors (n=1) and family satisfaction (n=1)); coping style (including internalising problems (n=3), externalising problems (n=3) and attention problems (n=1)); behavioural problems of childhood (n=10, no subcategories); psychiatric symptoms (depression (n=7), anxiety (n=4), PTSD (n=9) and general psychopathological symptoms (n=3)); general well-being (including ‘well-being’ (n=3), self-esteem (n=1), optimism (n=2) and quality of life (n=3)); and caregiver support (comprised of caregiver distress (n=5), parenting (n=3) and family communication (n=2)). Online supplemental appendix 4 details outcome definitions and assessment instruments.
Supplemental material
Effects of interventions
Due to a lack of combinable effect sizes and high heterogeneity where outcomes were potentially combinable, we were able to perform meta-analysis only in the behaviour problems of childhood and psychiatric symptoms categories. Table 2 reports pooled estimates; study-level results appear in table 1 and are organised by outcome in online supplemental appendix 5. Online supplemental appendix 6 reports subgroup analyses.
Supplemental material
Supplemental material
Resilience
Child psychosocial protective factors, resilience and family satisfaction outcomes were reported by one study each. The effect of interventions on child psychosocial protective factors reported by children was gav=0.206 (0.027, 0.386), and gav=0.063 (−0.110, 0.237) when reported by caregivers, compared with the control groups.40 The effect on resilience was gav=−0.08 (-0.916, 0.756)53; the effect on family satisfaction was gav=1.789 (1.058, 2.520).53 55 We deemed child psychological protective factors and resilience to be sufficiently similar to combine and generated a pooled estimate of gav=0.194 (0.018, 0.369) with 0% heterogeneity (see figure 2A).
Coping mechanisms
Most data showed favourable change in coping mechanisms. Internalising problems were reported by three studies. Murray reported a significant reduction by both children (gav=−1.600 (−2.123, –1.076)) and caregivers (gav= −1.428 (−1.939, −0.918))56 in pre–post analysis and Annan et al reported non-significant differences among intervention participants (children: gav=0.084 (−0.095, 0.263), caregivers: gav=−0.127 (−0.300, 0.046)) when compared with the control groups.40 Betancourt et al reported an association between caregiver distress and internalising problems (β=4.02, p<0.05), however, the effect size of the intervention on internalising problems could not be estimated.48 The same studies also reported on externalising problems. Annan et al reported an effect size of gav=−0.092 (−0.271, 0.087) by children and gav=−0.22 (−0.395,–0.048) by caregivers when compared with control groups.40 Murray et al reported an effect size of gav=−1.55 (−2.070, –1.030) by children and gav= −1.239 (−1.737, −0.742) by caregivers.56 The effect size for Betancourt et al could not be estimated. The effect of an intervention on the reduction in attention problems was reported only by Annan et al, at gav=−0.275 (−0.449, –0.100) by caregivers and gav= 0.04 (−0.139, 0.219) by children, compared with the control group.40
Behavioral problems of childhood
Six out of the 10 reported effect size estimates showed statistically significant improvements in this category, however, meta-analysis was not appropriate due to the use of differing assessment scales. Additionally, two studies reported significant differences in a desirable direction, but effect sizes could not be calculated.46 49 Significant effect sizes ranged from gav=−5.012 (−5.563, –4.462) to gav= −0.111 (−0.336, 0.114).43 One study reported a undesirable positive effect size (0.059 (−0.410, 0.527)), but it was not significant.53
Psychiatric symptoms
Psychiatric symptoms was the only category in which studies reported statistically significant undesirable effect sizes, . Clinically significant change, in which mean scores crossed a clinical threshold, were reported by at least one study for all three outcomes in this category. Online supplemental appendix 7 reports clinical significance of study findings in context with statistical significance, study design and effect size.
Supplemental material
Because depression was reported by seven studies but with variable designs and intervention types, meta-analysis was inappropriate. Individual study results were promising, with four studies reporting statistically significant improvement; only two of those reported clinically significant improvement, as baseline scores for the other two studies were already above the clinical threshold.50 52 60 61 Two studies reported clinically and statistically significant worsening.43 44
The effect on anxiety was reported by four studies with a pooled estimate of gav=−0.326 (−0.782, 0.131) (figure 2A). Due to high heterogeneity (I2=75.34%), which we attribute to variation in intervention types, intervention intensities, staff training levels and target population ages, we conducted sensitivity analysis for this outcome, with those estimates ranging from gav=−0.534 (−0.818,–0.249) to gav= −0.193 (−0.693, 0.298) (figure 2C).
Meta-analysis was also not possible for PTSD symptoms and evidence was weak overall. While the directionality of results was favourable for most studies, only one study reported clinical and statistically significant improvement.56
General well-being
Meta-analysis was not possible for the well-being outcome, but strong results were observed with effect sizes ranging from gav=0.963 (0.481, 1.444) to 7.821 (7.328, 8.314).41 52 56 The effect on self-esteem was only reported by Foka (gav=1.810 (1.26, 2.36)).52 Two studies showed an effect on optimism, both significant improvements (gav=1.481 (0.958, 2.003), gav=0.755 (0.419, 1.091)).45 52 Meta-analysis was possible for the quality of life outcome, although effect sizes could not be estimated for two studies,42 so our pooled estimate of gav=0.325 (−0.027, 0.678) reflects only two studies (figure 2D).
Caregiver support
Evidence was also weak across the caregiver-support category with only one statistically significant result, which showed improvement in family communication (gav=3.026 (1.885, 4.167)).55
Subgroup analysis
We conducted subgroup analysis for each reported outcome by intervention setting, age group, intervention intensity, intervention content and mental health expertise of interventionists. No conclusive pattern was observed by any of those variables across all categories (see online supplemental appendix 6).
Risk of bias and quality of evidence using grade
Two of the eight RCTs had high overall risk of bias; three were judged to be of concern and three to be at low risk of bias. Of the 15 non-randomised interventional studies, 1 had serious risk of bias, 9 had moderate risk of bias and 5 had low risk of bias (figure 3).
The overall quality of evidence for the three outcomes in which meta-analysis was feasible (pooled estimate of child psychological protective factors and resilience, symptoms of anxiety and quality of life) was ‘very low’. In all three categories, imprecision was the lowest-scored domain. See online supplemental appendix 8. Where outcomes were reported by single studies, evidence certainty was moderate for child psychological protective factors and attention problems, which were reported in an RCT,40 and very low for resilience, family satisfaction and self-esteem, which were reported in observational studies.52 53 55 We were unable to statistically explore risk of reporting bias or generate funnel plots because we identified so few effect sizes per meta-analysed outcome.
Supplemental material
Quality of evidence
Results of the overall level evidence for the three outcomes in which meta-analysis was feasible showed a ‘very low’ quality grade for each of resilience, symptoms of anxiety and quality of life (online supplemental appendix 8).
The reporting quality of this systematic review was ensured by using PRISMA 2009 reporting checklist.
Discussion
To our knowledge, this is the first review to thoroughly assess evidence on the effectiveness of non-clinical, resilience-enhancing interventions targeting forcibly displaced children irrespective of the clinical manifestations of psychological trauma and without geographic limitations. In general, we found that statistically significant improvements were reported by the majority of studies across all outcome categories except for outcomes related to caregiver support and psychiatric symptoms. However, meta-analyses, where possible, found effectiveness of these interventions to be low to moderate and GRADE assessment indicated very low quality of evidence. With such limitations to the evidence, we encourage caution in application to our findings to policy and advocate for further, rigorous research.
The lack of clarity on the effectiveness of the studied interventions can be explained by several factors. Primary studies had design limitations, since randomisation is extremely challenging in humanitarian emergencies.52 Most studies did not include a control group and constraints in resources resulted in variations in intervention formats, durations, follow-up intervals and personnel training. Meta-analysis of SMD had the limitation of combining effect sizes of randomised and non-randomised studies, which provide different quality of evidence. This resulted in heterogeneity and bias in effect-size estimations, which limits their utility in programme implementation. Furthermore, SMD assumes that the differences in SD among studies are due to differences in measurement scales rather than variability among study populations, which is unlikely given the global scope of the review.
Interventions targeting both children and caregivers and involving multiple content domains had greater impacts. However, rigorous comparison and ranking of intervention effectiveness was not possible and the long-term effects interventions could not be determined, nor could implications for global mental health programmes. Jordans et al reached the same conclusions in his review of mental health and psychosocial interventions for children exposed to protracted violence and war in LMICs.63 That review reported weak evidence for comparative assessment of interventions due to methodological (eg, absence of control groups) and geographical limitations. We were unable to identify specific, promising interventions or address our fourth research aim of exploring commonalities among successful interventions to inform the design of universal resilience-enhancing interventions for non-clinical settings. In the absence of stronger evidence, we recommend integrating existing resilience-enhancing interventions with related interventions already recognised as effective, such as trauma-focused cognitive behavioural therapy.64 Service delivery in group settings could provide an effective, lower-cost strategy for low-resource settings.64 65
We expected that most interventions would involve professional mental health personnel but found that fewer than half did. We believe that this is a positive marker for potential scalability, suggesting a paradigm shift in addressing the mental health of children who are exposed to armed conflicts in Low and Middle Income Countries (LMICs). This shift from tertiary prevention at clinical settings to community-based approaches has been widely advocated for.66–68 Within clinical settings, training general practitioners on mental health services in order to integrate them with the primary care delivered to FDP. There may also be an opportunity to leverage the expertise of healthcare professionals within FDP, a strategy that could counteract language and cultural barriers while supporting these professionals’ integration into their host countries, assuming sufficient funding and training were available, as appropriate, and understanding that some providers in the FDP may experience trauma symptoms that prevent them from practising.
Our findings also highlight many of the recognised challenges in mental health research. We observed incongruence between the geographical distribution of study locations (mostly Europe and high-income countries) vs the global distribution of FDPs. This is consistent with the fact that more than 70% of the global burden of mental health comes from LMICs, yet almost 94% of published mental health research in major psychiatric journals is from Europe, North America and Australia.69 70 It is also yet another reason to advocate for further research, to build the body of evidence closely alliged with the actual settings where most forcibly displaced children live.
It was feasible to measure clinical implications of interventions in only a few cases. Many measurement scales were intended to be descriptive rather than diagnostic and the psychometric properties of measurement scales were extremely variable in sensitivity and internal consistency. Some scales did not have hard cut-off points, and their developers advised that the threshold should be set based on the distribution of mean scores and context in which the interventions took place, which were not reported in the studies. Even in cases where clear thresholds were set, some studies reported removal or replacement of items in validated instruments for cultural reasons, rendering the recommended thresholds inapplicable.
The clinical interpretation of effect size estimates of psychological scales is particularly complex. Unlike hypertension, for instance, where any three-point difference in mmHg across the measurement scale imply a clinically significant change, no concrete thresholds are available in mental health research.71 Instead, interpretation should consider difference in mean scores relative to baseline in concert with the clinical thresholds and direction of each measurement tool.
The main limitation of our study reflects limitations in the fields of mental health research in that primary studies must rely on self-reported and parent-reported data. We also note that there is an inherent complexity in the assignment of outcomes to particular categories, although we consulted a subject-area expert prior to making assignments; assessment scales typically cover multiple domains, so symptoms measured to assess anxiety, for example, could potentially reflect depression symptoms as well, although the instrument would report only on anxiety. Finally, we could not draw funnel plots to support our risk-of-publication bias assessments. On the other hand, the main strength of this review is that we did not restrict included studies by region, language or outcome. We also searched six major databases, which provided access to a holistic set of publications.
Our review highlights the need for further resilience research, specifically more rigorous study-design and reporting guidelines. Research guidelines should specify core outcomes and recommended measurement scales. Enhanced efforts should be made to drive mental health research in LMICs, especially among forcibly displaced children since they are a particularly vulnerable and disadvantaged population.
Data availability statement
All data relevant to the study are included in the article or uploaded as online supplemental information.
Ethics statements
Patient consent for publication
Acknowledgments
This systematic review was supported by Dr Mohsen Malekinejad, MD, DrPH, Assistant Professor, Institute for Health Policy Studies; Dr Jess Ghannam PhD, Clinical Professor, Psychiatry, UCSF Weill Institute for Neurosciences; Peggy Tahir, Research and Copyright Librarian, UCSF, Min-Lin Fang, Research Librarian for Nursing and Social and Behavioral Sciences, UCSF and Evans Whitaker, Research Librarian for Medicine and Pharmacy, UCSF.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Handling editor Stephanie M Topp
Contributors AT and SG conducted record/article screening and data extraction. AT performed analysis. All authors contributed to drafting and review of the manuscript. MM provided supervision throughout. AT was responsible for the planning, conduct, and reporting of the work described in the article, SG was a data validator, EB contributed to the manuscript write up and GR and MM were responsible for technical review. AT and MM accepted full responsibility for the conduct of the study, had access to the data, controlled the decision to publish and are the gurantors of the the overall content.
Funding The authors thank the California Healthcare Foundation for their partial support of this study (award G-30906).
Disclaimer This study was conducted as part of a captone project for the Master of Science in Global Health at the University of California San Francisco. Views expressed in this manuscript are those of the authors.
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Author note Subject-mater expertise was provided by Dr Jess Ghannam. Research librarians Peggy Tahir, Min-Lin Fang and Evans Whitaker provided support to the search process.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.