Main

According to the most recent statistics, more than a hundred million people are currently fleeing wars, violence, persecution and human rights violations, almost half of whom are children.1 Most refugees live in low- and middle-income countries, with a considerable number living in camps and informal settlements1,2. As well as exposure to war-related events, many face ongoing adversities, including violence, insufficient access to basic resources, poor-quality accommodation, compromised livelihood, lack of job opportunities, child labour and limited access to education. These cumulative stressors contribute to a substantially elevated risk of mental health problems2.

Accurate epidemiological data from refugee children living in vulnerable contexts are critically important to plan and deliver mental health services. However, most of the available data are from children living in high-income countries and less-challenging settings3,4,5. Furthermore, there is substantial variation in prevalence rates of common mental disorders across studies6. Reliance on cross-sectional designs, non-probabilistic sampling, transcultural measurement errors and the use of screening measures rather than clinical interviews contributes to this heterogeneity3,7,8. The prevalence of mental disorders in refugee children probably also depends on a broad range of individual, familial and socioecological factors that are often not systematically measured in epidemiological studies2.

To address these gaps in the literature, we report on children’s mental health in a cohort of Syrian refugees living in informal tented settlements in Lebanon. Lebanon is a middle-income country with a population of around 6 million people, which accommodates an estimated 1.5 million Syrian refugees, approximately half of whom are children9,10. Syrian refugees in Lebanon face significant challenges, including lack of legal residency, extreme poverty, food insecurity and accommodation that is overcrowded or below humanitarian standards11. This cohort is thus considerably more representative of global refugee populations than samples settled in high-income countries.

We aimed to accurately estimate the prevalence of mental health problems and their patterns of co-occurrence and to identify the main predictors among refugee children in a cohort representative of refugee populations living in informal settlements in Lebanon. We report on prevalence of depression, anxiety disorders, post-traumatic stress disorder (PTSD) and conduct/oppositional defiant disorder (CD/ODD) in children and adolescents. To tackle methodological factors that potentially bias prevalence estimates, we (1) employed settlement-based probabilistic sampling, (2) collected longitudinal data and (3) used culturally adapted structured clinical interviews and locally validated screening tools for mental health problems. Further examination of psychopathological presentations was performed by evaluating concurrent comorbidity. Finally, we estimated the effect of a range of risk factors on children’s mental health problems, including exposure to war, maltreatment, caregiver psychopathology and the immediate refugee environment (for example, housing conditions and social environment).

Evidence to date

Two systematic reviews and meta-analyses were published in 2020 (one on high-income countries, covering up until 2019, and the other on low- and middle-income countries, covering 2003–2018), providing pooled prevalence estimates of mental health problems in refugee and asylum-seeking children3,5. We conducted a search on PubMed for more-recent literature, covering the period from 1 January 2018 to 10 September 2021. The search strategy used MeSH terms and keywords to identify studies about child or adolescent (‘child*’, ‘adolesc*’) refugees or asylum seekers (‘refugee*’, ‘asylum-seek*’) and reporting on mental illness, diagnosis or trauma (for example, ‘mental health’, ‘mental*’, ‘mental disorder*’, ‘depress*’, ‘anxiety*’, ‘phobi*’, ‘emotional disorder*’, ‘trauma’, ‘PTSD’). A total of N = 1,102 abstracts were reviewed, identifying 46 papers for review. Nineteen empirical studies were of sufficiently high quality (N > 50, sampling strategy likely to result in representative sample and outcomes measured using structured or semi-structured clinical interviews or established questionnaires) and reported prevalence of common mental disorders. Eleven studies reported on Syrian refugees. Pooled prevalence estimates from previous meta-analyses showed elevated prevalence of PTSD (22.7–52.0%), depression (13.8–28.0%) and anxiety (15.8–32.0%) and lower rates of externalizing behaviour problems (ODD, 1.7%; attention-deficit/hyperactivity disorder (ADHD), 8.6%). Higher prevalence was found using questionnaires compared with diagnostic interviews. More-recent literature suggested higher prevalence of PTSD and depression, but this was probably driven by a preponderance of studies (18/19) using questionnaires rather than clinical interviews.

A systematic review of socioecological factors contributing to risk and protection of the mental health of refugee children was published in 20214. Risk factors included cumulative exposure to war-related events, family separation, parental mental health problems, negative parenting, lower socioeconomic status, child labour, perceived discrimination, bullying and cumulative exposure to daily stressors. However, there is little previous research in low- and middle-income countries and in camps.

Contribution of this work

This study addresses the bias towards high-income countries by providing data from a large cohort of Syrian refugees living in informal settlements in Lebanon. These data are more representative of the majority of refugees globally and particularly those living in camps. We used probabilistic sampling, culturally adapted clinical interviews and locally validated symptom scales to provide more-accurate prevalence estimates for a range of mental health problems. We also examined comorbidity to better guide service planning. Finally, we estimated the effect sizes of a range of psychosocial and environmental risk factors to provide a clearer picture of their relative importance.

Results

Sample

At baseline, data were available from n =1,591 children (52.5% female, mean (s.d.) age = 11.44 (2.44) yr) and their primary caregivers (Table 1). The sample at 1 yr follow-up (n = 1,000) was broadly representative of the baseline sample, although children who participated were slightly younger, more likely to participate with their mother, to have left Syria ≥3 yr ago, to be registered with UNHCR and to have access to education. They were less likely to be from the most vulnerable localities, and caregivers were less likely to be working and less likely to have very low literacy. However, all these differences were small12. A subsample selected for clinical interview (n = 134) was also broadly representative of the baseline cohort, although participating children were slightly younger, had higher levels of externalizing behaviour problems, were more likely to be from the most vulnerable localities and were more likely to attend school. Other than the bias towards more-vulnerable localities, the differences were small12. See Table 1 and Supplementary Table 3 for details.

Table 1 Description of sample
Full size table

Prevalence of mental disorders

Point prevalence estimates for mental disorders based on clinical interview (n = 134) are presented in Table 2. Mental disorders were common: 39.6% of children met criteria for PTSD, 26.9% for either conduct disorder (CD) or oppositional defiant disorder (ODD), 20.1% for major depressive disorder and 47.8% for any anxiety disorder, with separation anxiety, agoraphobia, social phobia and specific phobia being most prevalent. Importantly, more than half of the children (57.5%) met the criteria for any common mental disorder. Diagnosis of ADHD applied in 5.2% of children. Prevalence was higher in males than in females for externalizing disorders. There was a trend towards obsessive compulsive disorder being more prevalent in children >11 yr and towards depression and separation anxiety being more common in children who had left Syria ≥3 yr ago.

Table 2 Prevalence estimates for common mental disorders derived from clinical interview
Full size table

Point prevalence estimates for common mental disorders based on symptom scales in the entire cohort are presented in Table 3. Adjusted estimates (adjusted for both the false-positive rate and false-negative rate) were on average 66.2% of the magnitude of raw estimates and consistent with prevalence based on clinical interview. Externalizing behaviour problems were more common in males, while anxiety disorders were more common in females; no other group differences reached significance.

Table 3 Prevalence estimates for common mental disorders in BIOPATH cohort at baseline and 1 yr follow-up
Full size table

Comorbidity

We focus on comorbidity patterns based on clinical interview data here (Fig. 1), but results for the whole sample based on symptom scales and separated by age, gender and time since leaving Syria are presented in Supplementary Figs. 1–4. Of those children who had a diagnosis, 79.2% met criteria for more than one disorder. Depressive and anxiety disorders were significantly associated (odds ratio (OR): 21.8 [95% confidence interval (CI): 4.9–97.4], P < 0.001). This mixed depressive–anxiety presentation was observed in 4.5% of children while isolated depression was observed in 0.7% and an isolated anxiety disorder was observed in 3.7%. An isolated PTSD diagnosis was ascribed to 6.0% of children while concurrent presentations of PTSD with anxiety, with mixed depression–anxiety or with combined depression, anxiety and CD/ODD ranged from 6.0% to 10.4% each; 33.6% of children had a diagnosis of PTSD with anxiety, depression and/or CD/ODD. PTSD and anxiety diagnoses were significantly associated (OR: 12.3 [95% CI: 5.2–28.7], P < 0.001) as were PTSD and mixed depression–anxiety (OR: 4.3 [95% CI: 1.8–10.4], P = 0.01). Similar patterns of concurrent comorbidity emerged in the whole sample at baseline and follow-up (Supplementary Fig. 1).

Fig. 1: Venn diagram showing co-occurrence of PTSD, major depressive disorder, anxiety disorders and CD/ODD in subsample with clinical interview data (n = 134).
figure 1

Odds ratios, with 95% CIs, were calculated to evaluate the degree of association between each pair of co-occurring disorders, and Bonferroni correction for multiple comparisons were used. *P < 0.05 after correction for multiple comparisons. PTSD versus depression (OR: 3.3 [95% CI: 1.4–8.1], Bonferroni-corrected P value: 0.09); PTSD versus anxiety (OR: 12.3 [95% CI: 5.2-28.7], P = 2.7 × 10−9); PTSD versus CD/ODD (OR: 4.7 [95% CI: 2.1–10.8], P = 0.002); depression versus anxiety (OR; 21.8 [95% CI: 4.9–97.4], P = 1.3 × 10−6); depression versus CD/ODD (OR: 2.8 [95% CI: 1.1–6.7], P = 0.32); anxiety versus CD/ODD (OR: 16.5 [95% CI: 5.4–50.7], P = 3.9 × 10−8); PTSD and anxiety versus depression (OR: 4.3 [95% CI: 1.8–10.4], P = 0.01); PTSD and anxiety versus CD/ODD (OR: 5.8 [95% CI: 2.5–13.2], P = 4.7 × 10−4); PTSD and depression versus CD/ODD (OR: 3.7 [95% CI: 1.3–10.7], P = 0.18 ); anxiety and depression versus CD/ODD (OR: 2.6 [95% CI: 1.1–6.5], P = 0.28); anxiety and depression and CD/ODD versus PTSD (OR: 4.6 [95% CI: 1.2–18.3], P = 0.28).

Full size image

Predictors of mental health symptoms

Table 4 shows effect sizes and significance of associations between risk factors and children’s depression, anxiety, PTSD and CD/ODD symptoms over time based on linear mixed-effects models (LMMs). Depression, anxiety and PTSD symptoms all showed a significant decrease between baseline and follow-up. The quality of the refugee environment was associated with all but anxiety symptoms, with effect sizes (Cohen’s d = –0.20 to –0.28) roughly an order of magnitude larger than other risk factors. The refugee environment was consistently the strongest predictor across a range of sensitivity analyses, suggesting that this was a robust effect (Supplementary Information section 6). Moreover, specific aspects of the environment contributed differently to mental health symptoms. For example, poor housing quality was specifically associated with children’s depression symptoms, lack of access to services was associated with children’s PTSD and CD/ODD symptoms, and a hostile and unsupportive community environment was associated with children’s depression, PTSD and CD/ODD symptoms.

Table 4 Risk factors for mental health problems in the whole longitudinal sample (N = 1,591, 2,595 observations)
Full size table

Depression and PTSD symptoms were positively associated with age while CD/ODD symptoms were negatively associated. As expected, anxiety symptoms were more pronounced in girls while CD/ODD symptoms were more pronounced in boys.

The level of exposure to war-related events was associated with all outcomes, with the greatest effect size (d = 0.05) for PTSD followed by depression (d = 0.03) symptoms. Child-reported maltreatment was also associated with all outcomes, with strongest effects for PTSD (d = 0.03) and depression (d = 0.02). Conflict between child and caregiver had strongest effects for PTSD (d = 0.05), followed by depression (d = 0.03) and CD/ODD (d = 0.02).

Caregiver mental health was associated with all children’s mental health symptoms. Caregiver depression and anxiety symptoms were particularly associated with CD/ODD (d = 0.04 and 0.04, respectively) and depressive symptoms (d = 0.03 and 0.04). Caregiver impulsivity was associated with CD/ODD (d = 0.03) and depressive (d = 0.02) symptoms in children. Although to a lesser extent, caregiver PTSD was associated with all child mental health outcomes.

Discussion

We report prevalence estimates of common mental disorders, their comorbidity profile and the main predictors of symptoms in a large cohort of Syrian refugee children living in informal settlements in Lebanon. Importantly, previously identified sources of spurious variability on those estimates8 were explicitly addressed by (1) using probabilistic sampling in (2) a prospective design and (3) using transcultural validated screening measures with (4) cohort-specific cut-offs derived from a subsample of families with whom a structured clinical interview and culturally sensitive diagnostic procedures were conducted. A range of individual, familial and socioenvironmental risk factors were assessed.

A high prevalence of common mental disorders was observed, but with some reduction in depression, anxiety and PTSD symptoms over time. Externalizing behaviour problems, however, were more stable. The observation of decreasing prevalence of disorders associated with acute distress, particularly PTSD and anxiety disorders, but maintenance of the prevalence of externalizing behaviour problems seems consistent with a change in symptom manifestation due to greater time since exposure to an acute stressor such as war while being continually exposed to the chronic stressors of displacement. However, longitudinal investigations with more than two timepoints and over a longer period will be necessary to better understand the relationship between changing symptom manifestation and differential exposure to acute and chronic stressors in refugee children.

Prevalence was high even after accounting for the expected false-positive rates resulting from screening tools. The difference between unadjusted and adjusted prevalence estimates—adjusted estimates were, on average, a third lower than raw estimates—confirms that the use of screening tools probably results in problems such as over-reporting of mental health problems13,14,15,16,17,18. Prevalence of PTSD was similar to that reported in one recent meta-analysis3 but higher than another5. However, the one recent study that was most comparable to ours—conducting clinical interviews with Syrian refugee children aged 7–18 yr in Lebanon and Jordan—found a similar prevalence of PTSD (45.6%)19. Prevalence of depression was broadly consistent with previous reports, but prevalence of anxiety was higher in our study3,5. The high level of adversity faced by families in informal settlements in Lebanon11, including daily stressors and events such as army raids on settlements20, may contribute to a higher burden of mental health problems compared with those resettled in high-income countries2,21. For example, the high prevalence of specific phobia reflected the unsafe nature of settlements: triggers included dogs and snakes (which come into tents) as well as fire (fires causing fatalities are not uncommon).

Importantly, as Ventevogel and Faiz have argued, results of population surveys on mental health symptoms in war-exposed and forcibly displaced populations might be better understood as general indicators of psychosocial distress rather than evidence of mental disorders18. Although we corrected prevalence estimates on the basis of data from clinical interviews, our results still reflect heightened prevalence rates of mental disorder. But these elevated prevalence rates should be interpreted in light of the challenging context of refugee settings that children are living in. Relying exclusively on categorical psychiatric diagnoses that emphasise a focus on individual psychopathology can be misleading in humanitarian settings and may reinforce stigma and direct efforts to ineffective interventions22,23. Consequently, comprehensive and systemic treatment approaches will probably benefit refugee children more than interventions focused only on the individual child.

A consistent pattern of concurrent comorbidities was observed, and the majority of children with mental health problems met criteria for more than one disorder. Comorbidity with PTSD was particularly prevalent, suggesting a substantial contribution of complex trauma-related presentations to the higher prevalence of mental disorders in refugee children. This may reflect the multiple and ongoing challenges and stressors that many children face, including violence, family separation, bereavement, child labour, maltreatment and daily stressors in camps13. The high degree of concurrent comorbidity reinforces the potential limits of a purely categorical approach to diagnosis in humanitarian settings. In line with this, thematic analysis of supervision notes from our clinical interviews suggests that apparent comorbidity could often be better characterized as a broader traumatic and adjustment reaction to the violence and displacement that children experienced20. Hence, a broader diagnostic formulation that links experience, pre-existing difficulties and symptomatology may be more helpful to inform treatment while also reducing the risk of stigma associated with categorical diagnoses.

Our results on the predictors of mental health symptoms confirm known associations of mental health problems in refugee children with exposure to war, maltreatment and caregiver mental health problems2. However, a gap in evidence highlighted in a recent review is the paucity of research on social and material factors of displacement2. We observed a consistent impact of the quality of the refugee environment on children’s mental health, which was an order of magnitude greater than individual and family-level factors. This is consistent with ecological systems models that emphasise a stronger effect of concurrent daily-life stressors compared with previous exposure to catastrophic events in conflict and post-conflict settings and among refugees and asylum seekers21,24. We found that an unsupportive and hostile environment was consistently associated with children’s mental health symptoms, in line with research in asylum-seeking adults showing that poor social integration is more strongly associated with depression and PTSD than traumatic events25. Such ongoing chronic stressors in the social ecology may explain the modest effect sizes of mental health interventions for children in post-conflict settings22. Hence, our results further support an emerging consensus in global mental health arguing for a multi-layered mental health care model, including assistance with issues such as housing, health and education and broader interventions and policy change aimed at fostering livelihoods, permitting employment and facilitating social support networks21.

More specifically, Syrian families living in informal settlements in Lebanon tend to experience difficulty with access to health care, work and education26,27,28, and Syrian refugee children have been reported to experience widespread discrimination, harassment and social isolation29. Refugee boys are particularly vulnerable to physical violence and exploitation as child labourers29 while girls experience gender-related violence and forced marriage or child labour30,31.

In a recent qualitative study with Syrian refugee families in Lebanon26, caregiver well-being was found to be intimately tied to parents’ financial situation. Economic hardship in war-torn environments and forcibly displaced populations has previously been shown to significantly impact parents’ mental health and, consequently, to lead to harsh parenting, familial conflict, violence and neglect32,33. Growing evidence also supports the crucial influence of caregiver mental health on the well-being and mental health of war-affected children34,35,36,37. Our results regarding caregivers’ symptoms of common mental disorders predicting children’s symptoms are in line with these findings and further support the need for interventions targeting caregiver mental health.

Importantly, our findings emphasise the need to address structural challenges (including access to health care, education and work) that probably affect both caregivers’ and children’s mental health30,38. Multi-level and cross-sectoral interventions and policies that reduce structural stressors, as captured here with the Perceived Refugee Environment Index (PREI), as well as family-level stressors (for example, caregiver’s mental health symptoms, conflict, maltreatment and neglect), will be of paramount importance to effectively promote the mental health of refugee children34. Qualitative research on the perspectives of Syrian refugee parents highlights the important role of economic and social daily stressors during displacement30. Hence, policies and programmes that aim to remove structural barriers to refugee families’ economic security will probably impact caregivers’ well-being, resulting in improved parenting quality and better child mental health outcomes34.

Despite the many strengths of the study, there are some limitations that should be considered. Biases in estimates induced by an inflated false-positive rate on the MINI Kid structured clinical interview cannot be fully ruled out as the instrument has not been specifically validated for Syrian refugee children. We attempted to address this by assigning diagnoses after supervision with an experienced clinical psychologist, critically reflecting on cultural and contextual factors that might influence the expression and diagnosis of mental disorders20. Interestingly, prevalence of ADHD was not significantly higher than seen in other populations39, providing some reassurance that the diagnostic approach per se was not artificially inflating prevalence. Mental health symptom scales were broadly found to be reliable and valid in this sample, although the SCARED fell below conventional standards for diagnostic accuracy and there was poor discrimination between anxiety cases and non-cases40. While we could address this in prevalence estimates by adjusting for the rate of false positives and negatives, this was potentially problematic for comorbidity and risk-factor analyses and may explain why few significant predictors of anxiety symptoms emerged. Finally, the sample that completed clinical interviews was not perfectly representative of the cohort.

In conclusion, Syrian refugee children living in informal settlements in Lebanon show a high level of mental health problems with substantial comorbidity, whether measured through clinical interview or validated symptom scales, and we identified a range of relevant risk factors with clear policy implications. Importantly, the finding that the effects of the current refugee environment are substantially greater than other factors highlights the need to focus on concurrent social determinants of mental health41. Policy change and interventions that reduce the social inequities that make it difficult for people to live dignified lives are required, in addition to scaling up quality mental health services, to address the mental health needs of refugee children41.

Methods

Ethics statement

This study complied with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. Ethical approval was granted by the Institutional Review Board of the University of Balamand/Saint George Hospital University Medical Center, Lebanon (ref: IRB/O/024-16/1815). The study was also reviewed by the Lebanese National Consultative Committee on Ethics and approved by the Ministry of Public Health. The linked clinical trial was granted ethical approval by the Institutional Review Board of the American University of Beirut (ref: SBS-2018-0582) and the Ministry of Public Health. The sponsor, Queen Mary University of London, reviewed the study for compliance with all relevant legal and regulatory requirements.

In all cases, written informed consent was first obtained from caregivers, and then written assent was obtained from children after verbal explanation by trained staff supported by a written information sheet. Financial compensation was offered to families for their time.

Mental health services were offered to anyone from participating communities; to reduce the risk of perceived pressure to participate, service access was not dependent on study participation. Services were provided by an international non-governmental organization that delivers primary care and mental health services in Lebanon, either as part of their standard services or through the linked clinical trial.

Sample

Data are from the Biological Pathways of Risk and Resilience in Syrian Refugee Children (BIOPATH) study, a cohort of Syrian refugees living in informal settlements in the Beqaa region of Lebanon12. The cohort was established in 2017, recruiting families who had left Syria within four years, had a child reported to be 8–16 years old and had the primary caregiver (typically the mother) available to participate. A purposive cluster-sampling approach was used to select localities with varying levels of vulnerability, and n = 77 settlements were visited. No statistical methods were used to pre-determine sample size for the analyses reported in this paper; however, the cohort sample size was calculated to ensure sufficient power to conduct analyses the cohort was specifically designed for (for example, DNA methylation analyses, not reported here). Within each settlement, all households were approached, and all eligible families were invited to participate. Recruitment and baseline data collection were completed between October 2017 and January 2018. A total of N = 2,282 families were approached, n = 1,591 (69.9%) of whom were eligible, provided consent and had valid data. Follow-up was completed one year later, between October 2018 and January 2019 (mean [s.d.] = 51.55 [1.84] weeks) and valid data were collected from n = 1,000 families (62.8% of the cohort). A subsample (n = 134, 8.4%) completed a clinical interview approximately six months later (mean [s.d.] = 27.03 [7.17] weeks), either as part of a sub-study looking at the reliability and validity of mental health questionnaires (n = 101) or as part of a linked clinical trial (n = 33). See (ref. 12) for full description of the cohort.

Procedure

At baseline and follow-up, families were visited at home and interviewed in Arabic by trained local research staff. One child from each family was selected using pre-defined criteria (child within age range whose birthday was closest to the interview date). The child and primary caregiver were interviewed simultaneously but separately by different interviewers; interviews took approximately 50–60 min. Clinical interviews in the subsample were completed at home or in a local clinic: children aged ≤12 yr were generally interviewed with one or both caregivers whereas older children were interviewed alone (depending on the preference of child and caregiver). In nearly all cases, the clinical interview was conducted primarily with the child with additional information sought from the caregiver. The visit for the clinical interview took approximately two hours. Recruitment data (including contact details) were collected and managed using REDCap electronic data-capture tools hosted at Queen Mary University of London42, and pseudonymized interview data were entered into Qualtrics (Qualtrics, version 2017–2019) via electronic tablet at the time of the interview.

Measures

Clinical assessment

The MINI International Neuropsychiatric Interview for Children and Adolescents (MINI Kid 6.0 (Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV)))43 was used to gather information about symptoms of mental disorders in the subsample. The MINI Kid 6.0 was previously translated into Arabic for Lebanon via a standard process of forward and back translation and review by the MINI Kid author and local experts. Additional questions were drafted by an experienced clinical psychologist (T.B.) and used at the same appointment to gain sufficient information to assign DSM-5 diagnoses. All cases were reviewed in clinical supervision with T.B. and diagnoses agreed by consensus, taking into account contextual, cultural and linguistic factors that might impact the diagnostic process. A Clinical Global Impression–severity score (CGI-s) was assigned, capturing functional impairment and distress resulting from symptoms (range 1–7). Case criteria were (1) meeting DSM-5 criteria for mental disorder and (2) CGI-s score ≥4, indicating moderate to severe functional impairment and/or distress. See Supplementary Table 1 and (ref. 20) for details and Supplementary Information section 1 for interrater reliability.

Mental health symptoms

Locally adapted and validated self-report and caregiver-report questionnaires were used to assess mental health symptoms in the full cohort at baseline and follow-up. Questionnaires were administered via interview, and visual aids were used to facilitate response. Mental health problems were measured using the Center for Epidemiological Studies Depression Scale for Children (CES-DC, abridged, self-report), the Screen for Child Anxiety Related Emotional Disorders (SCARED, abridged, self-report) and the Child PTSD Symptom Scale (CPSS, self-report), and externalizing behaviour disorders were measured using the Strengths and Difficulties Questionnaire (SDQ, licensed Arabic version, caregiver report) and a separate set of items reflecting CD and ODD criteria. See Supplementary Table 1 for a detailed description of measures and (ref. 40) for information about validity in this cohort.

Risk factors

War exposure was measured using the War Events Questionnaire, a 25-item checklist of war events reported at baseline. To take multiple raters into account44, child and caregiver responses were combined such that if either one reported that the child experienced an event, the event was considered to have occurred. The Perceived Refugee Environment Index (PREI) is a multidimensional measure, developed for this study, to assess the quality of the refugee environment. Nine subscales—Livelihood, Basic Needs, Housing, Family Environment, Learning Environment, Access to Services, Community Environment, Working Situation, and Future Mobility—and a total PREI score were used. Child maltreatment was measured using the International Society for the Prevention of Child Abuse and Neglect (ISPCAN) Child Abuse Screening Tool (ICAST, abridged). Caregiver–child conflict was measured using the Parent–Adolescent Conflict scale. Caregiver mental health was measured using the PTSD Checklist for DSM-5 (PCL-5), the anxiety subscale of the Depression, Anxiety and Stress Scale (DASS-21), the Center for Epidemiologic Studies Short Depression Scale (CES-D 10) and the Abbreviated Barratt Impulsiveness Scale (ABIS). See Supplementary Table 1 for detailed description of measures.

Data analysis

Point prevalence estimation

Cohort-specific cut-offs for each of the symptom scales were calculated using a subsample of the cohort (n = 119) with clinical interview data and contemporaneous questionnaire data. Each questionnaire was compared with DSM-5 diagnosis of the relevant disorder(s), and receiver operating characteristics (ROC) curve analysis was used to summarize overall diagnostic accuracy. The ROC curve was used to select a cut-off, and sensitivity, specificity, positive predictive value and negative predictive value were calculated at the cut-off (Supplementary Table 2).

Point prevalence of DSM-5 mental disorders was estimated in the subsample with clinical interview data and in the full cohort for depression, anxiety, PTSD and externalizing behaviour disorders using mental health symptom scales and cohort-specific cut-offs. Raw prevalence was calculated as the proportion of children scoring above cut-off on each scale. Prevalence was then adjusted for the proportion of false positives (based on the positive predictive value) and false negatives (based on the negative predictive value):

$$\begin{array}{l}{{{\mathrm{Adjusted}}}}\,{{{\mathrm{prevalence}}}} = \left( {{{{\mathrm{raw}}}}\,{{{\mathrm{prevalence}}}} \times {{{\mathrm{PPV}}}}} \right) \\+ \left( {1 - {{{\mathrm{raw}}}}\,{{{\mathrm{prevalence}}}} \times \left( {1-{{{\mathrm{NPV}}}}} \right)} \right)\end{array}$$

Adjusted prevalence was calculated for the whole sample and then separately by gender, age (≤11 yr, >11 yr) and time since leaving Syria (<3 yr, ≥3 yr). The two-proportion z test was used to test for differences between groups. Prevalence estimation analysis was performed in SPSS (v.28.0.0.0 (190)) and Microsoft Excel 2019 (v.16.01).

Comorbidity

The rates of co-occurrence of disorders were determined by calculating the frequency of overlapping diagnoses assigned by clinical interview for the subsample and the cohort-specific thresholds for symptom scales in the whole sample at baseline and follow-up. Venn diagrams depicting co-occurrence patterns and respective prevalence rates were constructed. Odds ratios, with 95% CIs, were calculated to evaluate the degree of association between each pair of co-occurring disorders. Patterns of co-occurrence were also investigated separately by gender, age (≤11 yr, >11 yr) and time since leaving Syria (<3 yr, ≥3 yr). Comorbidity analyses were performed in R Studio 2022.02.2 (R v.4.2.2).

Predictors of mental health symptoms

To describe the associations between risk factors and mental health symptoms over time, LMMs were constructed. The total scores for children’s depressive, anxiety, PTSD and externalizing symptoms assessed at baseline and follow-up were considered as independent outcomes. The LMMs were adjusted for subjects clustered by settlement (with random intercepts estimated for participants nested in settlements). This modelling strategy allowed us to describe associations across time while accommodating the complex covariance and unbalanced data structure45.

Effect sizes with 95% CIs were calculated for each risk factor on each continuous symptom scale score. First, four multivariate LMMs were constructed, one with each symptom scale score as the outcome, with child age, gender and time since leaving Syria as predictors. These models allowed us to quantify the specific relations between demographic factors and outcomes, controlling for the other demographic factors. Following this, independent LMMs were constructed for each risk factor and outcome, including the demographic factors as covariates. In the models constructed to quantify the association between caregiver mental health and child’s mental health symptoms, caregiver’s age and time since leaving Syria were included as covariates. Sensitivity analyses including child age, gender and time since leaving Syria as covariates were also carried out for these models, as well as comprehensive sensitivity analyses for all models (including only cases with follow-up data; only cases with complete data; in subgroups defined by gender, age and time since leaving Syria) to ensure that results were robust. These are presented alongside detailed description of the models in Supplementary Information section 6. Predictors analyses were performed in R Studio 2022.02.2 (R v.4.2.2).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.