Articles

Efficacy of psychosocial interventions for mental health

outcomes in low-income and middle-income countries:
an umbrella review
Corrado Barbui, Marianna Purgato, Jibril Abdulmalik, Ceren Acarturk, Julian Eaton, Chiara Gastaldon, Oye Gureje, Charlotte Hanlon, Mark Jordans,
Crick Lund, Michela Nosè, Giovanni Ostuzzi, Davide Papola, Federico Tedeschi, Wietse Tol, Giulia Turrini, Vikram Patel, Graham Thornicroft

Summary
Background Mental health conditions are leading causes of disability worldwide. Psychosocial interventions for these Lancet Psychiatry 2020
conditions might have a key role in their treatment, although applicability of findings to poor-resource settings might be Published Online
a challenge. We aimed to evaluate the strength and credibility of evidence generated in low-income and middle-income January 13, 2020
https://doi.org/10.1016/
countries (LMICs) on the efficacy of psychosocial interventions for various mental health outcomes.
S2215-0366(19)30511-5
See Online/Comment
Methods We did an umbrella review of meta-analyses of randomised studies done in LMICs. Literature searches were https://doi.org/10.1016/
done in Medline, Embase, PsychINFO, CINAHL, Cochrane Library, and Epistemonikos from Jan 1, 2010, until May S2215-0366(19)30531-0
31, 2019. Systematic reviews of randomised studies investigating the efficacy of psychosocial interventions for mental WHO Collaborating Centre for
health conditions in LMICs were included. Systematic reviews of promotion, prevention, and protection interventions Research and Training in
were excluded, because the focus was on treatment interventions only. Information on first author, year of publication, Mental Health and Service
Evaluation, Department of
outcomes, number of included studies, and reported summary meta-analytic estimates was extracted from included Neuroscience, Biomedicine and
meta-analyses. Summary effects were recalculated using a common metric and random-effects models. We assessed Movement Sciences, Section of
between-study heterogeneity, predictive intervals, publication bias, small-study effects, and whether the results of the Psychiatry, University of
Verona, Verona, Italy
observed positive studies were more than expected by chance. On the basis of these calculations, strength of
(Prof C Barbui MD,
associations was assessed using quantitative umbrella review criteria, and credibility of evidence using the GRADE M Purgato PhD, C Gastaldon MD,
approach. This study is registered with PROSPERO, number CRD42019135711. M Nosè PhD, G Ostuzzi PhD,
D Papola MD, F Tedeschi PhD,
G Turrini PsyD); WHO
Findings 123 primary studies from ten systematic reviews were included. The evidence on the efficacy of psychosocial
Collaborating Centre for
interventions in adults with depression in humanitarian settings (standardised mean difference 0·87, 95% CI Research and Training in
0·67–1·07; highly suggestive association, GRADE: moderate) and in adults with common mental disorders Mental Health, Neurosciences
(0·49, 0·36–0·62; highly suggestive association, GRADE: moderate) was supported by the most robust evidence. and Substance Abuse,
Department of Psychiatry,
Highly suggestive strength of association was found for psychosocial interventions in adults with schizophrenia for
College of Medicine, University
functional outcomes, in adults with depression, and in adults with post-traumatic stress disorder in humanitarian of Ibadan, Ibadan, Nigeria
settings. In children in humanitarian settings, and in children with disruptive behaviour, psychosocial interventions (J Abdulmalik MSc,
were supported by suggestive evidence of efficacy. Prof O Gureje DSc); Department
of Psychology, Koç University,
Istanbul, Turkey
Interpretation A relatively large amount of evidence suggests the benefit of psychosocial interventions on various (C Acarturk PhD); Centre for
mental health outcomes in LMICs. However, strength of associations and credibility of evidence were quite variable, Global Mental Health, London
depending on the target mental health condition, type of population and setting, and outcome of interest. This varied School of Hygiene and Tropical
Medicine, London, UK
evidence should be considered in the development of clinical, policy, and implementation programmes in LMICs and (J Eaton MRCPsych); Centre for
should prompt further studies to improve the strength and credibility of the evidence base. Global Mental Health
(C Hanlon PhD,
Funding University of Verona. Prof M Jordans PhD,
Prof C Lund PhD,
Prof G Thornicroft PhD) and
Copyright © 2020 Elsevier Ltd. All rights reserved. Centre for Implementation
Science (G Thornicroft),
Introduction and applicability of findings to low-income and middle- Institute of Psychiatry,
Psychology, and Neuroscience,
Psychosocial interventions, broadly defined as non- income countries (LMICs).4,5 Although the generalisability King’s College London, London,
pharmacological interventions focused on psychological issue is theoretically relevant for any type of intervention, UK; WHO Collaborating Centre
or social factors, can improve symptoms, functioning, for psychosocial inter­ ventions there are several for Mental Health Research and
quality of life, and social inclusion when used in the challenges specific to LMICs, including the need for Capacity Building, Department
of Psychiatry, School of
treatment of people with mental health conditions.1 training, fidelity checks, supervision, and monitoring, Medicine and Centre for
Psychosocial interventions also align with the principles concerns about cultural and social acceptability, and Innovative Drug Development
of personal recovery, such as the attainment of a fulfilling considerations of feasibility related to differences in and Therapeutic Trials for
and valued life.2,3 However, most studies assessing the mental health infrastructure and resources. Africa, College of Health
Sciences, Addis Ababa
efficacy of these interventions have been done in high- Psychosocial interventions are typically delivered by University, Addis Ababa,
income countries, raising the issue of generalisability mental health professionals. In LMICs, however, very few Ethiopia (C Hanlon);

www.thelancet.com/psychiatry Published online January 13, 2020 https://doi.org/10.1016/S2215-0366(19)30511-5 1

 Articles

Amsterdam Institute for Social

Science Research, University of Research in context
Amsterdam, Amsterdam,
Netherlands (M Jordans); Evidence before this study and health-care providers. Third, we showed that psychosocial
Alan J Flisher Centre for Public Psychosocial interventions have a key role in the treatment of interventions delivered by providers who are not mental health
Mental Health, Department of mental health conditions, because they might improve professionals are supported by highly suggestive evidence of
Psychiatry and Mental Health,
University of Cape Town,
symptoms, functioning, quality of life, and social inclusion. efficacy. Fourth, for some target mental health conditions and
Cape Town, South Africa Because most efficacy studies on these interventions have been outcome measures, effect sizes are of considerable magnitude,
(C Lund); Department of done in high-income countries, generalisability and suggesting that clinically meaningful results might be obtained.
Mental Health, Johns Hopkins applicability of findings to poor-resource settings is uncertain.
University Bloomberg School Implications of all the available evidence
For this reason, several randomised studies, and subsequently
of Public Health, Baltimore, In contrast with the generic view that an absence of evidence
MD, USA (W Tol PhD); systematic reviews, have examined the efficacy of psychosocial
exists in poor-resource settings, our results could inform
and Department of Global interventions on mental health outcomes in low-income and
Health and Social Medicine, governmental and non-governmental organisations and donors
middle-income countries. However, the available evidence is
Harvard Medical School, willing to implement or fund evidence-based mental health
controversial and fragmented into several reviews focusing on
Boston, MA, USA programmes in low-income and middle-income countries.
(Prof V Patel PhD) different populations, interventions, and outcomes, which
However, between-study heterogeneity, predictive intervals
Correspondence to: makes appraisal of the evidence using a similar metric and
that include the null value, and risk of small-study effects bias
Prof Corrado Barbui, methodological framework difficult.
WHO Collaborating Centre for were the main limitations of the reviewed evidence. On clinical
Research and Training in Mental Added value of this study grounds, these results suggest that developing the capacity of
Health and Service Evaluation, First, we showed how the strength and credibility of evidence non-specialist health-care providers to deliver psychosocial
Department of Neuroscience,
generated in poor-resource settings on the efficacy of interventions might be considered an implementation strategy
Biomedicine and Movement
Sciences, Section of Psychiatry, psychosocial interventions varies, depending on the target supported by a robust amount of evidence generated in
University of Verona, Verona mental health condition, type of population, and outcome of poor-resource settings, as well as in humanitarian settings.
37129, Italy interest. Second, on the basis of a robust methodological More generally, the results of this review could be used to inform
[email protected]
framework, a hierarchy of strength of associations and innovative strategies to build the clinical skills and capabilities of
credibility of evidence was developed to assist policy makers practitioners working in poor-resource settings.

mental health professionals might be available; therefore, broad health-care area and highlight whether the
for feasibility reasons, mental health interventions might evidence base is consistent or contradictory.13
be delivered by non-specialist professionals, including Medline, Embase, PsychINFO, CINAHL, Cochrane
nurses without psychiatric training, lay health workers, Library and Epistemonikos were searched from Jan 1, 2010,
or peer support workers.6,7 Psychosocial interventions until May 31, 2019, to identify up-to-date systematic
delivered by these workers might be less efficacious. reviews. The complete search strategy is provided in the
See Online for appendix Several randomised studies, and subsequently appendix (pp 4–5). No language restrictions were applied.
systematic reviews, have examined the efficacy of Electronic database searches were supplemented by a
psychosocial interventions on mental health outcomes manual search of reference lists from relevant studies.
in LMICs.8,9 However, the available evidence is still The Preferred Reporting Items for Systematic Reviews
controversial and fragmented into several reviews that and Meta-analyses reporting standards were followed to
focus on different populations, interventions, and out­ document the process of systematic review selection.14
comes, which makes appraisal of the evidence using a The selection of potentially relevant systematic reviews
similar metric and methodological framework difficult. was made by inspection of titles and abstracts by
Furthermore, the low quality of evidence affects the two reviewers independently (CB, MP). In case of
credibility of risk estimates, but it has never been formally discrepancies, a third review author (GiT) was involved,
synthesised. The aim of this review of systematic reviews and consensus reached by discussion. When titles and
was to review all available data on psychosocial inter­ abstracts did not provide information on the inclusion
ventions to quantify the efficacy of psychosocial inter­ and exclusion criteria, the full articles were obtained to
ventions for people with mental health conditions in verify eligibility. The full text of potentially included
LMICs . systematic reviews was obtained and carefully appraised
by at least two reviewers. The reference lists of included
Methods articles were analysed for additional items not retrieved
Search strategy and selection criteria by the database searches.
We used an umbrella review methodology to systemati­ Systematic reviews of randomised studies done in
cally review all available reviews on the topic. Umbrella LMICs investigating the efficacy of psychosocial inter­
reviews are systematic overviews of systematic reviews ventions for mental health conditions were included.
and meta-analyses.10–12 This review methodology was Systematic reviews of promotion, prevention, and
chosen because it could provide an overall picture of a protection interventions were excluded, because the focus

2 www.thelancet.com/psychiatry Published online January 13, 2020 https://doi.org/10.1016/S2215-0366(19)30511-5

 Articles

was on treatment interventions only. Psychosocial inter­ Heterogeneity was evaluated with Cochran’s Q statistic22
ventions included any non-pharmacological intervention (statistically significant for p value <0·10) and quantified
focused on psychological or social factors, including, but with the I² metric.23 Egger’s test was used to evaluate
not limited to, individual, family, or group psychological potential publication and small-study effects biases.24,25 A
therapies, education, training, or guidance.15,16 Inter­ p value of 0·10 or less in the regression asymmetry test
ventions with one or multiple components were included. with a more conservative effect in the largest study was
Mental health conditions included any mental health considered evidence for small-study effects bias.
problem along a continuum from mild, time-limited We evaluated the excess significance to examine
psychological distress to chronic, progressive, and severely whether the observed number of studies with statistically
disabling conditions.17 Therefore, both systematic reviews significant results (positive studies, p<0·05) in each
of studies that assessed the presence of a mental health meta-analysis was larger than their expected number.26
condition using a structured psychiatric diagnostic inter­ For each meta-analysis, the expected number was
view and systematic reviews of studies using validated or calculated as the sum of the statistical power estimates
commonly used rating scales were included. Systematic for each study in the meta-analysis. The power of each
reviews that used the World Bank country classifications study was calculated by an algorithm using a non-central
to identify studies done in LMICs were included. t distribution.27 The estimated power depends on the
Only systematic reviews with a quantitative synthesis plausible standardised mean difference. Because the true
of trial results (meta-analysis) were retained. Systematic standardised mean difference for any meta-analysis is
reviews without study-level effect sizes and 95% CIs unknown, we assumed that the most plausible effect is
were excluded. When two systematic reviews presented given by the largest study. Excess significance for each
overlapping datasets on the same comparison, the meta-analysis was claimed at a p value of 0·10 or less.26
systematic review with the largest number of component On the basis of these calculations, we classified the
studies providing study-level effect sizes was retained for strength of each association as “convincing”, “highly
the main analysis, in agreement with umbrella review suggestive”, “suggestive”, or “weak” (appendix p 15).12,28,29
methodology.10 Specifically, meta-analyses were free from biases
(convincing, Class I) if they met the following criteria:
Data analysis p value of less than 10–⁶ based on random effects meta-
From each included systematic review, two investigators analysis, more than 1000 participants, low or moderate
(CB and MP) independently extracted information on between-study heterogeneity (I² <50%), 95% prediction
first author, year of publication, outcomes, number of interval that excluded the null value, and no evidence of
included studies, and reported summary meta-analytic small-study effects and excess significance. Highly
estimates. The following information was extracted suggestive association (Class II) criteria required more
from each primary study: year of publication, popu­ than 1000 participants, highly significant summary
lation (adults, children, or adolescents), mental health associations (p value <10–⁶ by random-effects) and
condition, type of psychosocial intervention, outcomes, 95% prediction interval not including the null value.
type of professionals delivering the intervention, sample
size, and study-specific standardised mean differences
with corresponding 95% CIs.
3393 records identified through database searching
The quality of included systematic reviews was
independently assessed by two reviewers (DP, CG) using
AMSTAR-2 (A Measurement Tool to Assess Systematic 3293 excluded
Reviews), a 16-point assessment tool of the metho­ 1277 duplicates
2016 after title and abstract
dological quality of systematic reviews (appendix p 14).18 assessment
AMSTAR-2 has good inter-rater agree­ ment, test-retest
reliability, and content validity.18
100 full-text articles assessed for eligibility
Summary standardised mean differences with 95% CI
were re-estimated using common metric and random-
effects models because we were expecting high hetero­ 90 full-text articles excluded
geneity.19 In order to produce a pragmatic measure of the 2 wrong population
21 wrong intervention
efficacy of psychosocial interventions, the number 62 wrong study design
needed to treat (NNT) was calculated using the formulae 2 wrong setting
3 reviews on the same topic with
provided by Furukawa and colleagues.20 We also fewer studies than the included one
estimated the 95% prediction interval for the summary
random-effects estimates.21 Prediction intervals further
account for heterogeneity between studies and specify 10 systematic reviews including 19 meta-analyses
the uncertainty for the effect that would be expected in a
new study examining that same research question.21 Figure 1: Study profile

www.thelancet.com/psychiatry Published online January 13, 2020 https://doi.org/10.1016/S2215-0366(19)30511-5 3

 Articles

Population Outcome Studies Random-effects I² Predictive Random-effects Egger’s test Significant studies
(participants) significance interval standardised mean p value
threshold difference of the
reached largest study (95% CI)
Observed Expected p value
Asher et al Adults with Symptoms 7 (862) 0·005 94·7 –1·44 to 3·33 0·22 (–0·04 to 0·48) 0·043 5 5·33 0·68
(2017)2 schizophrenia
De Silva et al Adults with Social 10 (1671) 2·902 × 10–⁶ 89·7 –0·42 to 2·10 0·24 (0·10 to 0·38) 0·013 10 8·31 0·23
(2013)34 schizophrenia functioning
Cuijpers et al Adults with Symptoms 35 (4668) 2·172 × 10–²⁶ 89·7 –0·03 to 2·22 0·62 (0·48 to 0·76) 0·0092 31 28·38 0·39
(2018)35 depression
De Silva et al Adults with Social 12 (4098) 3·639 × 10–⁵ 89·5 –0·35 to 1·28 0·06 (–0·05 to 0·16) 0·0037 5 6·47 0·41
(2013)34 depression functioning
van Ginneken Adults with PTSD Symptoms 3 (223) 0·025 22·1 –2·26 to 2·98 0·28 (–0·17 to 0·73) 0·18 1 1·09 1·00
et al (2013)7
Singla et al Adults with common Symptoms 24 (6703) 8·173 × 10–¹³ 83·2 –0·13 to 1·11 0·72 (0·58 to 0·87) 0·020 16 15·23 0·83
(2017)37 mental disorders
Rahman et al Adults with perinatal Symptoms 14 (16 591) 1·931 × 10–⁵ 79·9 –0·27 to 1·04 0·62 (0·44 to 0·80) 0·99 8 9·39 0·41
(2013)36 common mental
disorders
Purgato et al Adults with PTSD in Symptoms 16 (1272) 8·623 × 10–¹⁴ 77·8 0·02 to 2·11 0·79 (0·54 to 1·04) 0·051 14 12·68 0·55
(2018)38 humanitarian settings
Purgato et al Adults with Symptoms 14 (1254) 1·515 × 10–¹⁷ 55·0 0·26 to 1·47 0·90 (0·65 to 1·15) 0·55 12 10·69 0·54
(2018)38 depression in
humanitarian settings
Purgato et al Adults with anxiety in Symptoms 5 (694) 3·270 × 10–⁹ 48·1 0·03 to 1·44 0·48 (0·24 to 0·72) 0·086 5 4·57 1·00
(2018)38 humanitarian settings
Burkey et al Children with Conduct 26 (6400) 4·366 × 10–⁸ 76·9 –0·24 to 0·99 0·11 (–0·08 to 0·30) 0·0011 15 15·46 0·84
(2018)15 disruptive behaviour problems
van Ginneken Children with PTSD or Symptoms 3 (298) 0·003 78·7 –6·24 to 8·02 1·27 (0·84 to 1·70) 0·24 2 2·14 1·00
et al (2013)7 depression
Purgato et al Children with PTSD in Symptoms 3 (130) 0·052 93·0 –18·3 to 21·4 0·06 (–0·59 to 0·72) 0·21 2 2·05 1·00
(2018)38 humanitarian settings
Purgato et al Children in PTSD 8 (2355) 5·975 × 10–4 80·2 –0·30 to 0·97 0·16 (–0·02 to 0·34) 0·29 4 4·52 0·73
(2018)39 humanitarian settings symptoms
Purgato et al Children in Depressive 10 (2672) 0·468 72·8 –0·47 to 0·58 0·07 (–0·11 to 0·25) 0·15 3 3·61 1·00
(2018)39 humanitarian settings symptoms
Purgato et al Children in Anxiety 7 (1969) 0·701 70·3 –0·49 to 0·56 0·14 (–0·06 to 0·33) 0·50 2 2·81 0·71
(2018)39 humanitarian settings symptoms
Turrini et al Adult and child PTSD 9 (856) 7·802 × 10–4 88·6 –0·84 to 2·53 0·65 (0·43 to 0·86) 0·34 6 5·47 1·00
(2019)16 refugees symptoms
Turrini et al Adult and child Depressive 5 (533) 5·746 × 10–5 92·7 –1·64 to 5·74 0·86 (0·64 to 1·09) 0·13 5 4·65 1·00
(2019)16 refugees symptoms
Turrini et al Adult and child Anxiety 2 (445) 0·075 94·7 ·· 0·45 (0·24 to 0·66) ·· 2 1·99 1·00
(2019)16 refugees symptoms
PTSD=post-traumatic stress disorder.

Table 1: Characteristics, quantitative synthesis, and umbrella review criteria of the 19 meta-analyses comparing psychosocial interventions and inactive controls

Suggestive evidence (Class III) criteria required only produces a credibility of estimate for each outcome and
more than 1000 participants and a p value of 0·001 or less supplies a tabular overview of findings easily under­
by random-effects. Weak association (Class IV) criteria standable for inter­vention participants, policy makers,
required only a p value of 0·05 or less. Associations were research planners, guideline developers, and other
considered non-significant if the p value was more than interested stakeholders.33 Summary of findings tables
0·05. Statistical analyses and power calculations were were developed using the GRADEProGDT app.
done using Stata version 12.0. p values were all two-tailed. This study is registered with PROSPERO, number
In addition to these quantitative criteria, the overall CRD42019135711.
credibility in the estimates was qualitatively assessed
by two reviewers (CB, MP) using the Grading of Role of the funding source
Recommendations, Assessment, Development, and The funder of the study had no role in study design, data
Evaluation (GRADE) method (appendix p 16).30–33 GRADE collection, data analysis, data interpretation, or writing of

4 www.thelancet.com/psychiatry Published online January 13, 2020 https://doi.org/10.1016/S2215-0366(19)30511-5

 Articles

Mental health Outcome Strength of Certainty of Standardised mean

condition association evidence difference (95% CI)
Adults Schizophrenia Symptoms Weak Low 0·94 (0·28 to 1·61)
Adults Schizophrenia Social functioning Highly suggestive Low 0·84 (0·49 to 1·19)
Adults Depression Symptoms Highly suggestive Low 1·09 (0·89 to 1·30)
Adults Depression Social functioning Suggestive Low 0·46 (0·24 to 0·69)
Adults PTSD Symptoms Weak Low 0·36 (0·04 to 0·67)
Adults Common mental disorders Symptoms Highly suggestive Moderate 0·49 (0·36 to 0·62)
Adults Perinatal common mental disorders Symptoms Suggestive Moderate 0·38 (0·21 to 0·56)
Adults, humanitarian PTSD Symptoms Highly suggestive Low 1·07 (0·79 to 1·35)
Adults, humanitarian Depression Symptoms Highly suggestive Moderate 0·87 (0·67 to 1·07)
Adults, humanitarian Anxiety Symptoms Weak Moderate 0·74 (0·49 to 0·98)
Children Disruptive behaviour Conduct problems Suggestive Low 0·37 (0·24 to 0·51)
Children PTSD or depression Symptoms Weak Moderate 0·89 (0·29 to 1·49)
Children, humanitarian PTSD Symptoms No association Very low 1·56 (–0·02 to 3·13)
Children, humanitarian Psychological distress PTSD symptoms Suggestive Moderate 0·33 (0·14 to 0·52)
Children, humanitarian Psychological distress Depressive symptoms No association Moderate 0·06 (–0·10 to 0·22)
Children, humanitarian Psychological distress Anxiety symptoms No association Moderate 0·03 (–0·13 to 0·20)
Adult and child refugees Psychological distress PTSD symptoms Weak Low 0·84 (0·35 to 1·33)
Adult and child refugees Psychological distress Depressive symptoms Weak Low 2·05 (1·05 to 3·05)
Adult and child refugees Psychological distress Anxiety symptoms No association Very low 0·99 (–0·10 to 2·07)

–3 –2 –1 0 1 2 3

Favours inactive Favours psychosocial

controls interventions

Figure 2: Reanalysis of the efficacy of psychosocial interventions in low-income and middle-income countries, with strength of association and certainty of evidence
Grey boxes denote the effect sizes of studies, and the size of each box is proportional to the statistical weight of the included studies. PTSD=post-traumatic stress disorder.

the report. The corresponding author had full access to studies of psychosocial interventions delivered by non-
all the data in the study and had final responsibility for specialist health-care providers and by both types of
the decision to submit for publication. providers, usually non-specialists in differing forms of
collaboration with specialist health-care providers.
Results Of the ten systematic reviews, three were of high
The systematic search yielded 3393 records. After quality according to the AMSTAR-2 scoring system,7,38,39
duplicate removal and inspection of titles and abstracts, one was of moderate quality,15 and six received a low or
100 full-text articles were assessed for eligibility. Ten critically low quality rating (appendix pp 17–19).2,15,34–37
systematic reviews, including 123 primary studies and AMSTAR-2 detected that in five reviews a study protocol
20 417 participants, met the umbrella review inclusion was not available, and study selection criteria were
criteria (figure 1).2,7,15,16,34–39 Details of the reviews excluded unclear. Additionally, the source of funding was reported
and the reasons for exclusion are provided in the appendix by only one review (appendix pp 17–19).
(pp 6–13). From the included systematic reviews, we A total of 15 meta-analyses reported a nominally
extracted information on 19 meta-analyses comparing statistically significant summary effect using random-
psychosocial interventions with inactive controls (table 1). effects models (p≤0·05); however, prediction intervals
In terms of populations, psychosocial interventions were excluded the null value in meta-analyses in adults with
studied in adults with schizophrenia, depression, PTSD, depression, and anxiety in humanitarian settings
common mental disorders, perinatal common mental (table 1). Significant hetero­geneity (I²>50%) was observed
disorders, post-traumatic stress disorder (PTSD), anxiety, in all comparisons, with the exception of the meta-
and in children with disruptive behaviour, PTSD, anxiety, analysis on the efficacy of psychosocial interventions in
depression and psychological distress without a diagnosis. adults with PTSD in the general population, and with
In seven meta-analyses studies were done in humanitarian anxiety in humanitarian settings (table 1). Risk of small-
settings, and three included studies done in refugee study effects bias was observed in eight comparisons,
populations. In terms of outcomes, psychotic, depressive, whereas excess of significance bias was undetected.
anxiety, PTSD, and conduct symptoms were considered, However, some comparisons consisted of few studies
with two comparisons focusing on functioning in adults (table 1), in which case the power of the test would be
with schizophrenia and depression (table 1). reduced.
Of the ten systematic reviews identified, three included None of the 19 meta-analyses had convincing strength
only studies of psychosocial interventions delivered by of association according to quantitative umbrella review
non-specialist health-care providers,7,37,39 and seven pooled criteria, and none scored high with GRADE. However,

www.thelancet.com/psychiatry Published online January 13, 2020 https://doi.org/10.1016/S2215-0366(19)30511-5 5

 Articles

in adults with schizophrenia on functioning outcomes,

Credibility Age Mental health Context Outcome Number
of evidence group condition needed to by the evidence in adults with depression in the general
treat (95% CI) population, and by the evidence in adults with PTSD in
Highly suggestive association humanitarian settings (table 2). NNT for these psycho­
Purgato et al Moderate Adults Depression Humanitarian Symptoms 3·0 (2·5–3·9) social interventions ranged between 2·4 and 5·4 (table 2).
(2018)38 settings The evidence on the efficacy of psychosocial interventions
Singla et al Moderate Adults Common General Symptoms 5·4 (4·2–7·4) in women during the perinatal period was supported by
(2017)37 mental disorders population suggestive evidence (table 2).
De Silva et al Low Adults Schizophrenia General Social 3·1 (2·2–5·4) In children, suggestive strength of association supported
(2013)34 population functioning
the efficacy of psychosocial interventions delivered in
Cuijpers et al Low Adults Depression General Symptoms 2·4 (2·1–2·9) humanitarian settings, and of psychosocial interventions
(2018)35 population
for conduct problems in children with disruptive behaviour
Purgato et al Low Adults PTSD Humanitarian Symptoms 2·5 (2·0–3·3)
(2018)38 settings (table 2). NNT for these psycho­logical interventions ranged
Suggestive association between 7·0 and 8·0 (table 2). All other psychosocial
Rahman et al Moderate Adults Common Perinatal Symptoms 7·0 (4·7–13·2) interventions were supported by weak evidence, and four
(2013)36 mental disorders comparisons did not show any association (table 2).
Purgato et al Moderate Children Psychological Humanitarian PTSD 8·0 (5·0–19·4) The evidence on the efficacy of cognitive behavioural
(2018)39 distress settings symptoms therapy in adults with depression, and the evidence on
De Silva et al Low Adults Depression General Social 5·7 (3·8–11·2) the efficacy of interpersonal therapy in adults with
(2013)34 population functioning
common mental disorders, was supported by the most
Burkey et al Low Children Disruptive General Conduct 7·1 (5·2–11·3) robust evidence, followed by the evidence on the efficacy
(2018)15 behaviour population problems
of other psychological interventions and multicom­ponent
Weak association
collaborative care (table 3; appendix pp 63–81). Suggestive
Purgato et al Moderate Adults Anxiety Humanitarian Symptoms 3·5 (2·7–5·3)
(2018)38 settings strength of association supported the efficacy of psycho­
van Ginneken Moderate Children PTSD or General Symptoms 2·9 (1·9–9·2)
education and parenting education programmes for
et al (2013)7 depression population women with common mental disorders in the perinatal
Asher et al Low Adults Schizophrenia General Symptoms 2·8 (1·8–9·7) period, whereas cognitive behavioural therapy for adults
(20)2 population with PTSD in humanitarian settings was supported
Turrini et al Low Adults and Psychological Refugees PTSD 3·1 (2·0–7·6) by highly suggestive strength of asso­ ciation (table 3;
(2019)16 children distress symptoms appendix pp 63–81).
Turrini et al Low Adults and Psychological Refugees Depressive 1·6 (1·4–2·5) In children, suggestive evidence was found for group-
(2019)16 children distress symptoms
based, focused psychosocial interventions delivered in
van Ginneken Low Adults PTSD General Symptoms 7·5 (3·9–63·2)
et al (2013)7 population humanitarian settings. All other interventions were
No association
supported by weak strength of association or did not
Purgato et al Moderate Children Psychological Humanitarian Depressive ··
show any association (table 3).
(2018)39 distress settings symptoms
Purgato et al Moderate Children Psychological Humanitarian Anxiety ·· Discussion
(2018)39 distress settings symptoms This umbrella review included 19 meta-analyses of
Purgato et al Very low Children PTSD Humanitarian Symptoms ·· studies done in LMICs assessing the efficacy of psycho­
(2018)38 settings social interventions for a range of mental health
Turrini et al Very low Adults and Psychological Refugees Anxiety ·· outcomes. Overall, available experimental evidence
(2019)16 children distress symptoms
suggests that psychosocial interventions might have a
PTSD=post-traumatic stress disorder. clinically relevant effect, although strength of associations
Table 2: Ranking of the 19 meta-analyses comparing psychosocial interventions and inactive controls by and credibility of evidence were variable. We found
strength of association and credibility of evidence that between-study heterogeneity, prediction intervals
including the null value, and risk of small-study effects
bias were the main factors bringing down the overall
Strength of association was highly suggestive for confidence in the evidence.
five meta-analyses, suggestive for four, and weak for six In adults, highly suggestive evidence supported
(figure 2). According to GRADE, the credibility of psychosocial interventions in schizophrenia, depression,
evidence was moderate for eight meta-analyses, low for common mental disorders considered as a group, and in
nine, and very low for two (figure 2; appendix pp 20–24). adults with depression and PTSD in humanitarian
The evidence on the efficacy of psychosocial inter­ settings. The credibility of evidence ranged from moderate
ventions in adults with depression in humanitarian to low, indicating that the credibility in the estimate for
settings, and in adults with common mental disorders, some comparisons was not optimal. In children,
was supported by the most robust evidence, followed by psychosocial interventions were supported by at least
the evidence on the efficacy of psychosocial interventions suggestive evidence in conduct disorders and for PTSD

6 www.thelancet.com/psychiatry Published online January 13, 2020 https://doi.org/10.1016/S2215-0366(19)30511-5

 Articles

outcomes among children in humanitarian settings. In that a fifth to a third of women from LMICs experience
most of these comparisons, effect sizes were of symptoms of perinatal depression,54–57 and we found
considerable magnitude in adults and children, with low suggestive evidence of efficacy for these interventions.
corresponding NNT values, indicating that clinically
meaningful results might be obtained in addition to
statistical significance. This finding expands previous data2 Standardised mean I² More than Strength of
difference (95% CI) 1000 participants association
showing that community-based psychosocial interventions
Adults with schizophrenia—symptoms2
might have a strong effect on symptom severity and
Psychoeducation 0·91 (0·33 to 1·50) 54·3 No Weak
functioning in schizophrenia and depression, and that
Psychosocial rehabilitation 0·01 (–0·42 to 0·43) 81·0 No No association
psychological interventions are effective in depression and
in people with common mental disorders in LMICs;35,37 Case management 1·63 (0·96 to 2·29) 80·2 No Weak

however, on the basis of umbrella review and GRADE Adults with schizophrenia—social functioning34

criteria, none of these effect sizes reached the maximum Psychoeducation 1·15 (0·05 to 2·25) 95·1 No Weak
of the ratings in terms of strength of association and Multicomponent structured 0·33 (0·10 to 0·55) 0·0 No Weak
psychosocial interventions
evidence credibility.
Art therapy 0·71 (0·31 to 1·12) ·· No Weak
The findings of this review might have policy and
Multicomponent 0·33 (0·10 to 0·55) 0·0 No Weak
practice implications. From a policy perspective, the community care
availability of a substantial body of experimental evidence Adults with depression—symptoms35
generated in LMICs is a major finding and should be
Interpersonal therapy 1·25 (0·96 to 1·54) 12·4 No Weak
emphasised to contrast the generic view that evidence is
Cognitive behavioural 1·16 (0·89 to 1·43) 92·0 Yes Highly suggestive
absent in poor-resource settings.8,40 This umbrella review therapy
showed that for psychosocial interventions the amount Other psychotherapies 0·94 (0·89 to 1·30) 73·3 Yes Suggestive
of evidence generated in LMICs is relatively large and Adults with depression—social functioning34
viable for implementation initiatives. We argue that this Interpersonal therapy 0·84 (0·39 to 1·29) 67·5 No Weak
message is of paramount relevance for governmental and Problem solving 0·10 (–0·15 to 0·35) ·· No No association
non-governmental organisations, and donors, willing to Morita therapy 0·66 (0·26 to 1·05) ·· No Weak
implement or fund mental health programmes in LMICs. Multicomponent 0·35 (0·11 to 0·59) 89·0 Yes Suggestive
These results might also be used to inform clinical collaborative care
practice. Linking evidence with practice remains Adults with PTSD—symptoms7
challenging,9,37,41–47 but the evidence from this review might Psychological interventions 0·22 (–0·10 to 0·54) 0·0 No No association
give some practical suggestions. For example, because Narrative exposure therapy 0·72 (0·18 to 1·26) ·· No Weak
most psychosocial interventions were delivered by non- Adults with common mental disorders—symptoms37
specialist health-care providers, alone or in collaboration Psychoeducation 0·36 (0·13 to 0·58) 50·2 No Weak
with specialist providers, developing the capacity of non- Psychosocial interventions 0·25 (0·14 to 0·36) 0·0 Yes Suggestive
specialist providers can be considered an implementation Cognitive behavioural therapy 0·67 (0·37 to 0·97) 89·5 Yes Suggestive
strategy supported by a robust amount of evidence Interpersonal therapy 0·80 (0·57 to 1·03) 75·1 Yes Highly suggestive
generated in LMICs.7,37,48 This approach could be crucial in Problem solving 0·64 (0·36 to 0·62) 0·0 No Weak
ensuring that such interventions are sustainable, ethical, Adults with perinatal common mental disorders36
and of sufficient quality.49,50 Depending on local culture
Parenting education 0·19 (0·09 to 0·30) 0·0 Yes Suggestive
and traditions, programmes aimed at fostering task-
Psychoeducation 0·36 (0·21 to 0·51) 9·6 Yes Suggestive
shifting initiatives between non-specialist mental health
Multimodal cognitive 0·94 (0·21 to 0·56) 93·0 Yes Weak
providers and informal community care providers could behavioural therapy
be developed to improve pathways to mental health Adults with PTSD in humanitarian settings38
care.51,52 A second suggestion is a more responsive Eye movement desensitisation 2·01 (1·52 to 2·51) 2·4 No Weak
approach to the broad mental health needs of populations and reprocessing
affected by humanitarian crises in LMICs, because we Cognitive behavioural therapy 0·85 (0·58 to 1·13) 70·0 Yes Highly suggestive
were able to show that psychosocial interventions for Interpersonal therapy 1·45 (0·44 to 2·47) ·· No Weak
adults with depression and PTSD, and for children in Thought field therapy 1·27 (0·91 to 1·63) ·· No Weak
humanitarian settings, are supported by at least Adults with depression in humanitarian settings38
suggestive evidence. Further research needs to confirm Eye movement desensitisation 1·44 (0·99 to 1·88) 0·0 No Weak
whether evidence-based psychosocial interventions can and reprocessing
be safely and sustainably implemented in contexts where Cognitive behavioural therapy 0·81 (0·60 to 1·02) 45·6 No Weak
stressors are ongoing, because most of the included Interpersonal therapy 0·84 (0·08 to 1·60) 58·9 No Weak
studies were done in the aftermath of humanitarian Adults with anxiety in humanitarian settings38
crises.53 A third suggestion is the inclusion of psychosocial Cognitive behavioural therapy 0·74 (0·49 to 0·98) 48·1 No Weak
interventions for women in the perinatal period in mental (Table 3 continues on next page)
health programmes, because epi­demiological data show

www.thelancet.com/psychiatry Published online January 13, 2020 https://doi.org/10.1016/S2215-0366(19)30511-5 7

 Articles

The main limitations of this umbrella review are those

Standardised mean I² More than Strength of
difference (95% CI) 1000 participants association of the systematic reviews included and, in turn, the
limitations of the primary studies. The most frequently
(Continued from previous page)
reported review shortcomings, detected by AMSTAR-2,
Children with disruptive behaviour15
were absence of a review protocol describing review
Child-focused cognitive 0·79 (0·03 to 1·56) 86·3 No Weak
methods before the review was done, of a thorough
behavioural therapy
discussion of between-study heterogeneity, and of
Child-focused interpersonal 0·04 (–0·23 to 0·31) 44·2 No No association
therapy information on funding. According to umbrella review
Child-focused psychosocial 0·39 (0·15 to 0·63) 64·7 Yes Weak and GRADE criteria, publication bias could not be
interventions excluded for some comparisons, and issues in properly
Child-focused social skills 0·25 (–1·06 to 1·56) 97·0 No No association masking outcome assessment were detected. We were
training also unable to ascertain whether included meta-analyses
Parent-focused psychosocial 0·39 (0·15 to 0·64) 70·0 Yes Weak included publications in languages other than English.
interventions
These limitations decreased the strength of associations
Parent-focused cognitive 0·68 (0·22 to 1·14) ·· No Weak
behavioural therapy
and credibility of evidence.
Additional limitations were related to the umbrella
Multi-component 0·18 (–0·13 to 0·50) 74·7 No No association
psychosocial interventions review methodology, because this approach is based on
Classroom-based 0·49 (0·28 to 0·71) 0·0 No Weak statistical reanalysis of meta-analyses. By definition,
psychosocial interventions umbrella reviews include only systematic reviews that
Children with PTSD or depression7 applied a quantitative approach to data presentation,
ERASE-Stress 1·27 (0·84 to 1·70) ·· No Weak whereas systematic reviews providing qualitative descrip­
Narrative exposure therapy 0·24 (–0·29 to 0·78) ·· No No association tions of the included studies, without applying meta-
Group psychotherapy 1·12 (0·64 to 1·60) ·· No Weak analytic techniques, are excluded. For example, systematic
Children with PTSD in humanitarian settings38 reviews that assessed the efficacy of psycho­social inter­
Cognitive behavioural 1·56 (–0·02 to 3·13) 93·0 No No association ventions in reducing intimate partner violence58 or in
therapy reducing psychological distress in people with HIV59 were
Children with psychological distress in humanitarian settings39 excluded, because no meta-analysis was done. However,
Group-based focused 0·33 (0·14 to 0·52) 80·2 Yes Suggestive the absence of a meta-analytical approach is typically
psychosocial interventions— motivated by scarcity of sufficient and homo­ geneous
PTSD symptoms
experimental evidence, which therefore does not reach the
Group-based focused 0·06 (–0·10 to 0·22) 72·8 Yes No association
psychosocial interventions—
minimum clinical and methodological require­ ments
depressive symptoms needed to be meta-analysed. Another limitation is that we
Group-based focused 0·03 (–0·13 to 0·20) 70·3 Yes No association did not analyse whether the efficacy of psycho­ social
psychosocial interventions— interventions is moderated by length of follow-up, type of
anxiety symptoms
inactive control condition, type of provider, or by other
Adult and child refugees with psychological distrress16 clinical, social, or context-related variables.60,61 Analysis of
PTSD symptoms these variables was not feasible owing to the nature of the
Eye movement 2·04 (1·56 to 2·51) 11·2 No Weak primary data. In terms of interventions, stratifying the
desensitisation and
reprocessing analysis by type of psycho­social intervention inevitably
Psychosocial interventions 0·69 (0·51 to 0·88) 0·0 No Weak
decreased the power of the analysis, but provided clinical
Narrative exposure therapy 0·00 (–0·24 to 0·25) 0·0 No No association
insights. For example, cognitive behavioural interventions
Depressive symptoms
for depression and PTSD and interpersonal therapy for
Psychosocial interventions 2·96 (0·27 to 5·66) 96·0 No Weak
adults with common mental disorders were supported by
the most robust evidence, whereas very few evaluations
Eye movement 1·46 (1·07 to 1·86) 0·0 No Weak
desensitisation and of interventions had social components. In terms of
reprocessing outcomes, we were able to reanalyse functional outcomes,
Anxiety symptoms in addition to sympto­matic improvement, for just two
Eye movement 1·56 (1·10 to 2·01) ·· No Weak comparisons, which is unexpected in view of the emphasis
desensitisation and on general wellbeing, functioning, and quality of life in
reprocessing
studies involving people with mental disorders.62 No meta-
Psychosocial interventions 0·45 (0·24 to 0·66) ·· No Weak
analyses were found on rare pragmatic outcomes, such as
PTSD=post-traumatic stress disorder. overall mortality, suicide, or deliberate self-harm.63–65 A
Table 3: Reanalysis comparing psychosocial interventions and inactive controls by type of psychosocial general limitation is that we did not consider promotion,
intervention prevention, or protection interventions.66,67
In view of the variability in the strength of associations
and credibility of the evidence, action is required to
support further research efforts in LMICs in diverse

8 www.thelancet.com/psychiatry Published online January 13, 2020 https://doi.org/10.1016/S2215-0366(19)30511-5

 Articles

settings. Several initiatives have aimed to improve the Acknowledgments

capacity of LMIC research centres to conduct high quality CB, MP, GiT, and FT receive support from the European Commission
(grant 779255). GT is supported by the National Institute for Health
trials. These efforts need to be sustained and expanded to Research (NIHR) Collaboration for Leadership in Applied Health
ensure rigorous evidence generation can take place, led by Research and Care South London and by the NIHR Applied Research
partners in LMICs, to address research questions that Centre at King’s College London NHS Foundation Trust and the NIHR
are relevant to low-resource settings.68,69 In terms of Applied Research Unit. GT receives support from the National Institute
of Mental Health of the National Institutes of Health (award
populations, this review suggests a need for further R01MH100470). GT is supported by the UK Medical Research Council
studies involving children and adolescents, especially in through the Emilia (MR/S001255/1) and Indigo Partnership (MR/
humanitarian settings, and adults and children with a R023697/1) awards. CH receives support from AMARI as part of the
migrant background. A focus on mental health along DELTAS Africa Initiative [DEL-15–01]. GT, CH, and CL are funded by the
NIHR Global Health Research Unit on Health System Strengthening in
a continuum from mild psychological distress to Sub-Saharan Africa, King’s College London (GHRU 16/136/54) using
severely disabling conditions, as suggested by the Lancet UK aid from the UK Government. The views expressed in this
Commission on global mental health and sustainable publication are those of the authors and not necessarily those of the
development,17 seems an appropriate and feasible NIHR or the Department of Health and Social Care.

approach, although we note a scarcity of evidence for References

1 Dua T, Barbui C, Clark N, et al. Evidence-based guidelines for
specific diagnostic conditions, such as bipolar disorder. In mental, neurological, and substance use disorders in low- and
terms of interventions, considering the number and middle-income countries: summary of WHO recommendations.
diversity of available psychosocial interventions, future PLoS Med 2011; 8: e1001122.
2 Asher L, Patel V, De Silva MJ. Community-based psychosocial
research efforts should be directed to ascertain which interventions for people with schizophrenia in low and
delivery method would be more feasible and sustainable, middle-income countries: systematic review and meta-analysis.
assessing whether brief, basic, group, and non-specialist- BMC Psychiatry 2017; 17: 355.
delivered versions of existing evidence-based psychosocial 3 Mutiso VN, Pike K, Musyimi CW, et al. Feasibility of WHO
mhGAP-intervention guide in reducing experienced discrimination
interventions could be an affordable and scalable in people with mental disorders: a pilot study in a rural Kenyan
alternative, for example.64,70–73 In terms of outcomes, setting. Epidemiol Psychiatr Sci 2019; 28: 156–67.
assessment of the long-term effectiveness of these 4 Tol WA, Barbui C, Bisson J, et al. World Health organization
guidelines for management of acute stress, PTSD, and bereavement:
interventions would be relevant, including functional and key challenges on the road ahead. PLoS Med 2014; 11: e1001769.
quality of life measures, because they have been seldom 5 Tol WA, Barbui C, Galappatti A, et al. Mental health and
considered by the studies included in this umbrella psychosocial support in humanitarian settings: linking practice and
research. Lancet 2011; 378: 1581–91.
review. More generally, research activity needs to be more 6 Keynejad RC, Dua T, Barbui C, Thornicroft G. WHO Mental Health
sensitive to questions and concerns arising from Gap Action Programme (mhGAP) Intervention Guide: a systematic
implementation activities, and implementation activities review of evidence from low and middle-income countries.
Evid Based Ment Health 2018; 21: 30–34.
need to optimise the uptake of research findings into 7 van Ginneken N, Tharyan P, Lewin S, et al. Non-specialist health
practice. worker interventions for the care of mental, neurological and
Given the pressing need for evidence-based answers substance-abuse disorders in low- and middle-income countries.
Cochrane Database Syst Rev 2013; 11: CD009149.
for people with mental health conditions, and in view of
8 Patel V. From delivery science to discovery science: realising the full
the data on the efficacy of psychosocial interventions in potential of global mental health. Epidemiol Psychiatr Sci 2016;
adults and children, we argue that these forms of 25: 499–502.
interventions should be made routinely available to 9 Hanlon C. Next steps for meeting the needs of people with severe
mental illness in low- and middle-income countries.
distressed adults and children in LMICs, recognising Epidemiol Psychiatr Sci 2017; 26: 348–54.
that the feasibility and sustainability of psychosocial 10 Ioannidis JP. Integration of evidence from multiple meta-analyses:
interventions, especially in the long term, might be a a primer on umbrella reviews, treatment networks and multiple
treatments meta-analyses. CMAJ 2009; 181: 488–93.
challenge.74–77 In general, because psychosocial inter­ 11 Ioannidis J. Next-generation systematic reviews: prospective
ventions are valued by service users as a complement to meta-analysis, individual-level data, networks and umbrella reviews.
pharmacological treatment,79 their involvement in this Br J Sports Med 2017; 51: 1456–58.
implementation process could contribute to mental 12 Papatheodorou S. Umbrella reviews: what they are and why we
need them. Eur J Epidemiol 2019; 34: 543–46.
health system strengthening. 13 Solmi M, Correll CU, Carvalho AF, Ioannidis JPA. The role of
Contributors meta-analyses and umbrella reviews in assessing the harms of
CB and MP designed the study. CB drafted the manuscript. CG, MN, psychotropic medications: beyond qualitative synthesis.
Epidemiol Psychiatr Sci 2018; 27: 537–42.
GO, DP, FT, and GiT contributed to the database preparation and double
check. CB, MP, and GO did data analyses. CG and DP assessed the 14 Moher D, Liberati A, Tetzlaff J, Altman DG. Preferred reporting
items for systematic reviews and meta-analyses: the PRISMA
quality of meta-analyses using AMSTAR-2. JA, CA, JE, OG, CH, MJ, CL,
statement. BMJ 2009; 339: b2535.
WT, VP, and GrT critically revised the manuscript. All authors
15 Burkey MD, Hosein M, Morton I, et al. Psychosocial interventions
commented on and approved the draft and final manuscripts.
for disruptive behaviour problems in children in low- and middle-
Declaration of interests income countries: a systematic review and meta-analysis.
We declare no competing interests. J Child Psychol Psychiatry 2018; 59: 982–93.
16 Turrini G, Purgato M, Acarturk C, et al. Efficacy and acceptability of
Data sharing psychosocial interventions in asylum seekers and refugees:
All data included in this umbrella review were extracted from publicly systematic review and meta-analysis. Epidemiol Psychiatr Sci 2019;
available systematic reviews. 28: 376–88.

www.thelancet.com/psychiatry Published online January 13, 2020 https://doi.org/10.1016/S2215-0366(19)30511-5 9

 Articles

17 Patel V, Saxena S, Lund C, et al. The Lancet Commission on global 42 Tol WA, Purgato M, Bass JK, Galappatti A, Eaton W. Mental health
mental health and sustainable development. Lancet 2018; and psychosocial support in humanitarian settings: a public mental
392: 1553–98. health perspective. Epidemiol Psychiatr Sci 2015; 24: 484–94.
18 Shea BJ, Reeves BC, Wells G, et al. AMSTAR 2: a critical appraisal tool 43 Lund C, Alem A, Schneider M, et al. Generating evidence to narrow
for systematic reviews that include randomised or non-randomised the treatment gap for mental disorders in sub-Saharan Africa:
studies of healthcare interventions, or both. BMJ 2017; 358: j4008. rationale, overview and methods of AFFIRM. Epidemiol Psychiatr Sci
19 DerSimonian R, Laird N. Meta-analysis in clinical trials. 2015; 24: 233–40.
Control Clin Trials 1986; 7: 177–88. 44 Docherty M, Shaw K, Goulding L, et al. Evidence-based guideline
20 Furukawa TA. From effect size into number needed to treat. Lancet implementation in low and middle income countries: lessons for
1999; 353: 1680. mental health care. Int J Ment Health Syst 2017; 11: 8.
21 Higgins JP, Thompson SG, Spiegelhalter DJ. A re-evaluation of 45 Chisholm D, Heslin M, Docrat S, et al. Scaling-up services for
random-effects meta-analysis. J R Stat Soc Ser A Stat Soc 2009; psychosis, depression and epilepsy in sub-Saharan Africa and
172: 137–59. South Asia: development and application of a mental health systems
22 Cochran WG. The combination of estimates from different planning tool (OneHealth). Epidemiol Psychiatr Sci 2017; 26: 234–44.
experiments. Biometrics 1954; 10: 101–29. 46 Eaton J, Ryan G. Making Universal Health Coverage a reality:
23 Higgins JPT, Green S. Cochrane handbook for systematic reviews of bridging the gap between Global Mental Health and practical
interventions. Version 5.1.0. The Cochrane Collaboration, 2011. integration into local health systems. Epidemiol Psychiatr Sci 2017;
https://handbook-5-1.cochrane.org/ (accessed Aug 14, 2019) 26: 245–47.
24 Sterne JA, Sutton AJ, Ioannidis JP, et al. Recommendations for 47 Votruba N, Ziemann A, Grant J, Thornicroft G. A systematic review
examining and interpreting funnel plot asymmetry in meta-analyses of frameworks for the interrelationships of mental health evidence
of randomised controlled trials. BMJ 2011; 343: d4002. and policy in low- and middle-income countries.
Health Res Policy Syst 2018; 16: 85.
25 Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-
analysis detected by a simple, graphical test. BMJ 1997; 315: 629–34. 48 Padmanathan P, De Silva MJ. The acceptability and feasibility of
task-sharing for mental healthcare in low and middle income
26 Ioannidis JP, Trikalinos TA. An exploratory test for an excess of
countries: a systematic review. Soc Sci Med 2013; 97: 82–86.
significant findings. Clin Trials 2007; 4: 245–53.
49 Andrews L. Non-specialist health worker interventions for the care
27 Lubin JH, Gail MH. On power and sample size for studying features
of mental, neurological, and substance-abuse disorders in low- and
of the relative odds of disease. Am J Epidemiol 1990; 131: 552–66.
middle-income countries. Issues Ment Health Nurs 2016;
28 Dragioti E, Karathanos V, Gerdle B, Evangelou E. Does 37: 131–32.
psychotherapy work? An umbrella review of meta-analyses of
50 Hanlon C, Tesfaye M, Wondimagegn D, Shibre T. Ethical and
randomized controlled trials. Acta Psychiatr Scand 2017; 136: 236–46.
professional challenges in mental health care in low- and
29 Machado MO, Veronese N, Sanches M, et al. The association of middle-income countries. Int Rev Psychiatry 2010; 22: 245–51.
depression and all-cause and cause-specific mortality: an umbrella
51 Burns JK, Tomita A. Traditional and religious healers in the pathway
review of systematic reviews and meta-analyses. BMC Med 2018;
to care for people with mental disorders in Africa: a systematic
16: 112.
review and meta-analysis. Soc Psychiatry Psychiatr Epidemiol 2015;
30 Guyatt GH, Juniper EF, Walter SD, Griffith LE, Goldstein RS. 50: 867–77.
Interpreting treatment effects in randomised trials. BMJ 1998;
52 Nguyen T, Holton S, Tran T, Fisher J. Informal mental health
316: 690–93.
interventions for people with severe mental illness in low and lower
31 Guyatt GH, Oxman AD, Kunz R, Vist GE, Falck-Ytter Y, middle-income countries: a systematic review of effectiveness.
Schünemann HJ. What is “quality of evidence” and why is it Int J Soc Psychiatry 2019; 65: 194–206.
important to clinicians? BMJ 2008; 336: 995–98.
53 Morina N, Malek M, Nickerson A, Bryant RA. Meta-analysis of
32 Guyatt G, Oxman AD, Akl EA, et al. GRADE guidelines: interventions for posttraumatic stress disorder and depression in
1. Introduction-GRADE evidence profiles and summary of findings adult survivors of mass violence in low- and middle-income
tables. J Clin Epidemiol 2011; 64: 383–94. countries. Depress Anxiety 2017; 34: 679–91.
33 Barbui C, Dua T, van Ommeren M, et al. Challenges in developing 54 Fellmeth G, Fazel M, Plugge E. Migration and perinatal mental
evidence-based recommendations using the GRADE approach: health in women from low- and middle-income countries:
the case of mental, neurological, and substance use disorders. a systematic review and meta-analysis. BJOG 2017; 124: 742–52.
PLoS Med 2010; 7: e1000322.
55 Fisher J, Cabral de Mello M, Patel V, et al. Prevalence and
34 De Silva MJ, Cooper S, Li HL, Lund C, Patel V. Effect of determinants of common perinatal mental disorders in women in
psychosocial interventions on social functioning in depression and low- and lower-middle-income countries: a systematic review.
schizophrenia: meta-analysis. Br J Psychiatry 2013; 202: 253–60. Bull World Health Organ 2012; 90: 139G–49G.
35 Cuijpers P, Karyotaki E, Reijnders M, Purgato M, Barbui C. 56 Gelaye B, Rondon MB, Araya R, Williams MA. Epidemiology of
Psychotherapies for depression in low- and middle-income maternal depression, risk factors, and child outcomes in
countries: a meta-analysis. World Psychiatry 2018; 17: 90–101. low-income and middle-income countries. Lancet Psychiatry 2016;
36 Rahman A, Fisher J, Bower P, et al. Interventions for common 3: 973–82.
perinatal mental disorders in women in low- and middle-income 57 Munodawafa M, Mall S, Lund C, Schneider M. Process evaluations
countries: a systematic review and meta-analysis. of task sharing interventions for perinatal depression in low and
Bull World Health Organ 2013; 91: 593–601I. middle income countries (LMIC): a systematic review and
37 Singla DR, Kohrt BA, Murray LK, Anand A, Chorpita BF, Patel V. qualitative meta-synthesis. BMC Health Serv Res 2018; 18: 205.
Psychological treatments for the world: lessons from low- and 58 Tol WA, Murray SM, Lund C, et al. Can mental health treatments
middle-income countries. Annu Rev Clin Psychol 2017; 13: 149–81. help prevent or reduce intimate partner violence in low- and
38 Purgato M, Gastaldon C, Papola D, van Ommeren M, Barbui C, middle-income countries? A systematic review.
Tol WA. Psychological therapies for the treatment of mental BMC Womens Health 2019; 19: 34.
disorders in low- and middle-income countries affected by 59 Sikkema KJ, Dennis AC, Watt MH, Choi KW, Yemeke TT, Joska JA.
humanitarian crises. Cochrane Database Syst Rev 2018; 7: CD011849. Improving mental health among people living with HIV: a review of
39 Purgato M, Gross AL, Betancourt T, et al. Focused psychosocial intervention trials in low- and middle-income countries.
interventions for children in low-resource humanitarian settings: Glob Ment Health (Camb) 2015; 2: 2.
a systematic review and individual participant data meta-analysis. 60 Cristea IA. The waiting list is an inadequate benchmark for
Lancet Glob Health 2018; 6: e390–400. estimating the effectiveness of psychotherapy for depression.
40 Eaton J, McCay L, Semrau M, et al. Scale up of services for mental Epidemiol Psychiatr Sci 2019; 28: 278–79.
health in low-income and middle-income countries. Lancet 2011; 61 Kohrt BA, Asher L, Bhardwaj A, et al. The role of communities in
378: 1592–603. mental health care in low- and middle-income countries:
41 Flisher AJ, Lund C, Funk M, et al. Mental health policy development a meta-review of components and competencies. Int J Environ Res
and implementation in four African countries. J Health Psychol 2007; Public Health 2018; 15: E1279.
12: 505–16.

10 www.thelancet.com/psychiatry Published online January 13, 2020 https://doi.org/10.1016/S2215-0366(19)30511-5

 Articles

62 Nevarez-Flores AG, Sanderson K, Breslin M, Carr VJ, Morgan VA, 70 Bockting CLH, Williams AD, Carswell K, Grech AE. The potential
Neil AL. Systematic review of global functioning and quality of life of low-intensity and online interventions for depression in low- and
in people with psychotic disorders. Epidemiol Psychiatr Sci 2019; middle-income countries. Glob Ment Health (Camb) 2016; 3: e25.
28: 31–44. 71 Einfeld SL, Stancliffe RJ, Gray KM, et al. Interventions provided by
63 Brandão DJ, Fontenelle LF, da Silva SA, Menezes PR, parents for children with intellectual disabilities in low and middle
Pastor-Valero M. Depression and excess mortality in the elderly income countries. J Appl Res Intellect Disabil 2012; 25: 135–42.
living in low- and middle-income countries: systematic review and 72 Fazel M, Patel V, Thomas S, Tol W. Mental health interventions in
meta-analysis. Int J Geriatr Psychiatry 2019; 34: 22–30. schools in low-income and middle-income countries.
64 Arjadi R, Nauta MH, Chowdhary N, Bockting CLH. A systematic Lancet Psychiatry 2014; 1: 388–98.
review of online interventions for mental health in low and middle 73 Naslund JA, Aschbrenner KA, Araya R, et al. Digital technology for
income countries: a neglected field. Glob Ment Health (Camb) 2015; treating and preventing mental disorders in low-income and
2: e12. middle-income countries: a narrative review of the literature.
65 Bantjes J, Iemmi V, Coast E, et al. Poverty and suicide research in Lancet Psychiatry 2017; 4: 486–500.
low- and middle-income countries: systematic mapping of literature 74 Petersen I, Lund C, Stein DJ. Optimizing mental health services in
published in English and a proposed research agenda. low-income and middle-income countries. Curr Opin Psychiatry 2011;
Glob Ment Health (Camb) 2016; 3: e32. 24: 318–23.
66 Petersen I, Evans-Lacko S, Semrau M, et al. Promotion, prevention 75 Brooke-Sumner C, Petersen I, Asher L, Mall S, Egbe CO, Lund C.
and protection: interventions at the population- and community- Systematic review of feasibility and acceptability of psychosocial
levels for mental, neurological and substance use disorders in interventions for schizophrenia in low and middle income
low- and middle-income countries. Int J Ment Health Syst 2016; countries. BMC Psychiatry 2015; 15: 19.
10: 30. 76 Murray LK, Jordans MJ. Rethinking the service delivery system of
67 Purgato M, Tol WA, Bass JK. An ecological model for refugee mental psychological interventions in low and middle income countries.
health: implications for research. Epidemiol Psychiatr Sci 2017; BMC Psychiatry 2016; 16: 234.
26: 139–41. 77 Thornicroft G, Ahuja S, Barber S, et al. Integrated care for people
68 Schneider M, Sorsdahl K, Mayston R, et al. Developing mental with long-term mental and physical health conditions in low-income
health research in sub-Saharan Africa: capacity building in the and middle-income countries. Lancet Psychiatry 2019; 6: 174–86.
AFFIRM project. Glob Ment Health (Camb) 2016; 3: e33. 78 Semrau M, Lempp H, Keynejad R, et al. Service user and caregiver
69 da Silva ATC, Hanlon C, Susser E, et al. Enhancing mental health involvement in mental health system strengthening in low- and
research capacity: emerging voices from the National Institute of middle-income countries: systematic review. BMC Health Serv Res
Mental Health (NIMH) global hubs. Int J Ment Health Syst 2019; 2016; 16: 79.
13: 21.

www.thelancet.com/psychiatry Published online January 13, 2020 https://doi.org/10.1016/S2215-0366(19)30511-5 11