Types of studies
Any studies that allocated participants or clusters of participants by a random method and included a no-intervention standard care or wait list condition will be included. The study selection process will be reported in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines . There will be no restrictions on publication type, status, language, or date.
Types of participants and settings
Children of both sexes (aged 0–18 years) who have been exposed to a traumatic event in the context of a humanitarian crisis, and residing in a LMIC. The review will not be restricted to participants meeting specific psychiatric diagnostic categories (as we are interested in selective preventive interventions), but will include participants who have been exposed to traumatic events related to humanitarian crises (e.g., displacement, exposure to violence, or major losses). Traumatic events and humanitarian settings will be defined according to the criteria used in a recent systematic review and meta-analysis of mental health and psychosocial support interventions in LMICs . Humanitarian settings will include countries that have had a humanitarian crisis, including any natural or technological disasters and political violence (armed conflicts and wars). LMICs will be selected using the World Bank criteria . Interventions delivered in any setting will be accepted, including healthcare facilities, refugee camps, schools, communities, survivors’ homes, and detention facilities.
Types of interventions
Interventions will include selective preventive psychosocial interventions compared to no intervention or waiting list. Selective preventive interventions will be defined according to the ecodevelopmental model of prevention proposed by Weisz . According to this model, selective prevention targets subgroups of children with specific risk factors (for example children with psychological distress); who are not already experiencing a disorder .
Types of outcome measures
PTSD symptoms (measured with the Child Posttraumatic Symptom Scale , Clinician Administered PTSD Scale for Children and Adolescents  (or with any other commonly used rating scales).
Anxiety (measured with the Screen for Anxiety Related Emotional Disorders (SCARED-5) , or with any other commonly used rating scales).
Depression (measured with the Children Depression Rating Scale , Depression Self-Rating Scale (DSRS) , or with any other commonly used rating scales).
Hope (measured with the Children’s Hope Scale (CHS)  or with any other commonly used rating scales).
Coping (measured with the Kidcope  or with any other commonly used rating scales).
Social support (measured with the Social Support Inventory Scheme  or with any other commonly used rating scales).
Functioning (measured with the Child Function Impairment Measure  or with any other commonly used rating scales).
Loss to follow-up.
Search methods for identification of studies
The search method will start with an update of the search strategy performed by Tol et al. in a systematic review of RCTs on mental health and psychosocial interventions in LMIC humanitarian settings in 2011 . The Cochrane Collaboration Trials Registers of the Developmental Psychosocial and Learning Problems Group , the Depression, Anxiety and Neurosis Group , and the Schizophrenia Group  will be searched to identify other relevant studies. These registers are compiled by systematic searches of major databases, hand searches, and conference proceedings. Medline, Embase, PsycInfo, CENTRAL, and PILOTS will be searched as well.
These sources will be supplemented by searching reference lists of relevant papers and previous systematic reviews and relevant specialist websites.
Moreover, the following regional databases will be hand searched for additional studies not retrieved in the electronic database and for citations of unpublished reports: AfricaBib databases, African Index Medicus, AFROLIB Database, Biomedicina Croatica, Chinese Medicine Premier, East View Information Service, EurasiaHealth, Hellenic Ph.D. Dissertations Thesis, HERDIN NeON Database, Hrcak, Index Medicus for the Western Pacific, Indian Citation Index, IndMED, InfoMED, IranMedex, KoreaMed, LILACS, Magyar Orvosi Bibliográfia, MedCarib, Medical Bibliography Hippocrates, Medical databases (Russia), Panteleimon, Turk MEDLINE, Türk Tıp Veri Tabanı, and University of Zagreb Medical School Repository.
Material downloaded from electronic sources will include author, institution, or publication journal details. As an initial step, on the basis of titles and abstracts, two reviewers (MP and WT) working independently will select potentially relevant studies. Studies rated as possible candidates by either of the two review authors will be added to a preliminary list and their full texts will be retrieved.
No blinding to the names of authors, institutions, and journal of publication will take place. We will resolve any further disagreements by consensus with a third member of the review team (CB).
Data collection, transfer, and management
The principal investigator of each included trial will be contacted by email to ask cooperation on the project and to confirm study eligibility. We will prepare a standard template in which we will introduce our research group and the objectives of this work, and we will list the individual and study level information requested from study authors. Attached we will forward the study protocol of this IPD meta-analysis. Principal investigators will be asked to contribute to the IPD meta-analysis with the original datasets. Study level information will include study protocol, published papers, and unpublished or additional material. Individual level information will include socio-demographic and clinical characteristics, primary and secondary outcome measures, date of randomization, and date of follow-up.
All study data will be entered in a computerized password-protected database, only accessed by named study staff and stored by the Department of Mental Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, United States. The computer will be placed in a secure location protected by a key (physical protection). All study data will be used only for the purposes stated in this Study Protocol, and will not be forwarded to third parties.
The following documents will be requested for each of the included studies:
– Study protocol;
– Study questionnaires;
– Clinical study report (if available);
– List of publications.
We will assign a unique number to each of the included studies. Study level variables will include: author(s), institution or journal of publication, year of publication, country, setting, duration of the study, type of intervention, recruitment period, primary and secondary outcomes, methodological issues. Risk of bias and quality information will additionally be extracted (see below).
For study level information that will not be provided by the investigators, a reviewer (MP) will extract information from the study protocol, if available, and from published paper(s).
We will assign each child a unique identification number and we will create a dataset from each study with one row per child. Individual level information will include:
Unique identification number (code), age, study, country, date of assessment, time;
– Socio-demographic characteristics: family characteristics, caregivers occupation, religion, displacement experience, household size;
– Exposure to traumatic events: past and current exposure;
– Symptoms: PTSD symptoms, depressive symptoms, anxiety symptoms;
– Resilience (hope, coping, social support);
– Type of psychosocial intervention (duration, frequency, clinical features).
Assessment of risk of bias
Knowledge of risk of bias is crucial in interpreting the results of individual studies and in identifying differences between studies that may determine a high level of heterogeneity in the results . The risk of bias of each study will be assessed by two reviewers (MP and CB) using the Cochrane Collaboration risk of bias tool . This comprises a judgment and a support for the judgment for each specific feature of the study (sequence generation, allocation concealment, blinding, incomplete outcome data assessment, selective outcome reporting and other source of bias/locally validated diagnostic instrument). The judgment for each entry involves assessing the risk of bias as ‘low risk’, as ‘high risk’, or as ‘unclear risk’, with the last category indicating either lack of information or uncertainty over the potential for bias .
Descriptive statistics will be used to summarize baseline, clinical, and socio-demographic characteristics of participants in each study, and for the pooled sample. We will describe the characteristics of included studies, the number of studies contributing with data to the IPD meta-analysis, and the types of outcomes provided by each study. The number of excluded studies together with the reasons for exclusion will be also reported (following the PRISMA statement). Results tables, graphs, and statistical syntax will be discussed by the review authors during meetings and/or telephone conferences.
The sample size available for each specific outcome will depend on the completeness and quality of the data from the RCTs.
Individual level analysis
Two general approaches to IPD meta-analysis have been suggested [37, 38]. The first is the ‘one-stage analysis’. This combines all the IPD from all studies to perform a single analysis. This can be in a ‘mega-trial’ analysis, where distinctions between studies are ignored and the data are analyzed as if they belong to a single trial. This approach is feasible even if outcome summary scores differ; common items among scales can be used to link outcomes from multiple studies to a common metric in pooled data. The second is the ‘two-stage approach’. Here, studies are analyzed separately, and then summary statistics combined using meta-analysis techniques. In the current IPD meta-analysis, both these approaches will be employed.
For dichotomous outcomes, relative risk (RR) with a 95% confidence interval (CI) will be calculated based on a random effects model, as this takes into account differences between studies . Continuous scores from different outcome scales will be analyzed using weighted or standardized mean differences with a 95% CI. A random-effects-model will be employed, as this takes into account any differences between studies . Assorted graphical tools and examination of heterogeneity (the I2 statistic) will be used to investigate the possibility of statistical heterogeneity among studies. We will interpret the I2 estimate as indicating the presence of high levels of heterogeneity if it is greater than or equal to 0.50 . Data from included studies will be entered into a funnel plot. Funnel plot will be visually inspected in order to detect the presence of publication bias . We will adjust for confounding variables in regression models of outcomes on predictors. We will select potentially confounding variables using a priori assumptions and exploratory data analysis.
We will analyze mediation using structural equations modeling to implement a one-step procedure that simultaneously models relationships between exposure and mediator, mediator and outcome, and exposure and outcome . The test of mediation will be based on a test of the distribution of the product of z-scores for the associations between the exposure and mediator and the mediator and outcome ; this approach has been demonstrated in simulation studies to have superior power than other approaches for our particular case . Potential effect modification will be examined by including terms for exposure, suspected moderator, and the interaction between the two, and testing the significance of the interaction.
A sensitivity analysis excluding studies that failed to provide individual-level information will be carried types of exposure to traumatic events;out. Additionally, subgroup analyses will be conducted grouping studies according to:
types of exposure to traumatic events;
levels of family and social support;
The Principal Investigator of each included trial, as well as other investigators with a specific interest for this project, will be asked to be co-authors of any publication(s). These investigators will be asked to provide feedback on data analysis and interpretation, as well as on the writing of the publication(s).
The present project does not involve primary data collection from humans, as it will be based on secondary analyses of already collected de-identified datasets. Only studies that have received formal approval from an ethical committee and followed acceptable ethical standards will be included in the present work. Further institutional review board approval will be obtained at the Johns Hopkins Bloomberg School of Public Health. Additionally, national and international regulations on patient privacy will be followed.