The important role of mothers during displacement: Direct and indirect effects of the refugee context on Syrian refugee children's mental health
Abstract
Refugee children are at increased risk for mental health problems, including post-traumatic stress, depression, and externalizing problems. The refugee environment, maternal mental health, and parenting may reduce or exacerbate that risk. This study investigated their direct and indirect associations with child mental health cross-sectionally in a sample of Syrian refugee child–mother dyads in Lebanon in 2017–19. Mediating pathways were tested using structural equation modeling with 1446 dyads (child: Mage = 11.39, 52.1% females) and again 1 year later with 872 (child: Mage = 12.17, 53.1% females) of the original sample. Mediating pathways from the refugee environment through maternal mental health and parenting to child outcomes were detected, emphasizing the importance of a holistic approach to refugee mental health.
Abbreviations
-
- LMIC
-
- low- and middle-income countries
-
- PTSD
-
- post-traumatic stress disorder
The number of forcibly displaced children continues to grow. By the end of 2020, there were an estimated 10.1 million children displaced across borders, but new and ongoing conflicts mean record numbers of families are currently forced to flee (United Nations High Commissioner for Refugees (UNHCR), 2022). As a group, refugee children show increased rates of psychological problems including post-traumatic stress disorder (PTSD), depression, and externalizing behavior problems (Blackmore et al., 2020; Kien et al., 2019), but experiences and psychological outcomes differ widely. One particularly at-risk population is refugee children living in informal settlements in low- and middle-income countries (LMIC). In addition to war and displacement, this group of children often faces continued lack of access to basic resources, stability, and education, and shows more persistent mental health problems than children resettled in higher-income contexts (Müller et al., 2019; Popham et al., 2022).
A considerable amount of work has been carried out in recent years to identify key aspects of the environmental context that affect refugee children's mental health, with a view to understanding what can be done to help (Arakelyan & Ager, 2021; Scharpf et al., 2021). Maternal mental health and parenting are consistently associated with child outcomes (Eltanamly et al., 2021), and aspects of the wider refugee environment (e.g., poverty and postmigration stressors) may also play a role (Zwi et al., 2018). However, these are not independent predictors. Risk and resilience models recognize that developmental processes and adaptation to adversity occur via complex interactions between different contextual and individual factors (Lundberg & Wuermli, 2012; Suárez-Orozco et al., 2018). For example, postmigration stressors might indirectly affect children by worsening caregiver distress (Sim, Bowes, & Gardner, 2018), but there is as of yet no research in LMICs on how specific aspects of the refugee environment affect children via caregivers, particularly with more than one wave of data. In the current paper, we address these points by investigating the direct and indirect effects of the refugee and family environments on child mental health at two separate time points from a sample of Syrian refugee children living in informal tented settlements in Lebanon.
Predictors of child refugee mental health
Syrian refugee children have been exposed to some of the most violent acts of war, including continuous bombing and witnessing the torture and execution of loved ones. Unsurprisingly, previous war exposure is an important predictor of developmental outcomes in refugee children (Çeri & Nasiroğlu, 2018; Gormez et al., 2018). However, the current environment also plays a crucial role (Suárez-Orozco et al., 2018). One of the most consistent findings in refugee research to date is that the family environment, caregiver mental health, and the parent–child relationship are key to refugee children's mental health (Eltanamly et al., 2021; Eruyar et al., 2018; Sangalang et al., 2017). For example, poorer maternal mental health has been linked with worse internalizing and externalizing problems in Syrian refugee children in multiple contexts (Eruyar et al., 2018; Sim, Bowes, & Gardner, 2018). In terms of the caregiver–child relationship, harsh or rejecting parenting (Eltanamly et al., 2021), child abuse and neglect (Karam et al., 2019; Lee et al., 2020), and parent–child conflicts or disagreements (Choi et al., 2008; Sangalang et al., 2017) have been associated with heightened child PTSD, depression, and behavioral problems in a range of refugee environments. Conversely, the family environment can also have positive and protective influences on children. For example, Syrian adolescents in Lebanon highlight their mothers as a source of strength (Nagi et al., 2021), and warm parenting is negatively associated with child symptoms of PTSD and internalizing and externalizing problems (Eltanamly et al., 2021).
Factors outside of the family may also have strong impacts on children, although findings are less consistent. Refugee reports highlight resource constraints, food access, and housing as key stressors (Bermudez et al., 2018; Sim, Fazel, et al., 2018). Indeed, economic status, hunger, and household assets are related to psychological distress in children in refugee camps (Meyer et al., 2017; Zwi et al., 2018), although some studies find no links between the physical environment and refugee child outcomes (Betancourt, Yudron, et al., 2012). Similarly, the wider social environment shows inconsistent associations with child outcomes; community support can play a protective role for children (Zwi et al., 2018) but some studies find no such association (Betancourt, Salhi, et al., 2012).
Interplay between systems
Of the potential risks and resources reviewed above, many are interrelated, and several may affect child outcomes primarily via indirect effects through other predictors. For example, Syrian mothers reported that feeling unable to provide for their children causes them distress, which subsequently causes their children distress (Rizkalla et al., 2020). In this way, lack of resources might affect children indirectly by impacts on maternal mental health or changes in mothering behavior, as in studies of economic hardship in non-refugee populations (Conger et al., 1992). Alternatively, caregivers may also represent important buffers to the effects of environmental stressors. We can conceptualize these mediating pathways using family process and ecological systems models, as well as risk and resilience frameworks specifically for refugee and immigrant youth (Popham et al., 2021; Suárez-Orozco et al., 2018); children are embedded in multiple systems from the more proximal (e.g., family) to the more distal (e.g., national policy), and those distal systems impact a child via effects on the more proximal (Figure 1; Arakelyan & Ager, 2021; Bronfenbrenner, 1979; Conger et al., 1992). There is also interplay between individual features within systems; for example, within the family system, a caregiver's own mental health problems could reduce their ability to provide supportive parenting or to monitor their child's activities (Rizkalla et al., 2020).
Several studies provide empirical evidence that caregiver mental health affects child's mental health via parenting. In fact, some studies suggest that family functioning, including conflict and positive parenting, fully mediates the association between caregiver and child mental health (Sangalang et al., 2017), while others find that a direct effect remains alongside indirect effects (Bryant et al., 2018; Sim, Bowes, & Gardner, 2018). However, the pathway from the wider environment to child outcomes via caregiver mental health and parenting is less studied. Bryant et al. (2018) identified that daily stressors reported by caregivers in their sample of refugees in Australia were related to child outcomes via the pathway through parental PTSD and harsh parenting. In terms of evidence specific to humanitarian settings, Sim and colleagues (Sim, Bowes, & Gardner, 2018) report an association between maternal psychological distress and refugee environment stressors such as access to food, sanitation, and shelter in a sample of Syrian refugee mothers in Lebanon. In the same sample, perceived social support was associated with more maternal psychological resilience and less harsh parenting (Sim et al., 2019). These findings suggest that physical and social support in refugee settings impact the family environment, but do not tell us about potential indirect effects on child mental health. More work is needed to better understand the pathways through which different aspects of the refugee environment and family systems affect child outcomes, particularly in refugees living in LMIC and humanitarian settings, for whom daily stressors will be substantially different from those faced by refugees resettled in high-income countries (Suárez-Orozco et al., 2018).
Effects of time and period of child development
It is also important to consider that the processes influencing child mental health likely change over time. We know that the process of adaptation is highly dynamic (Müller et al., 2019; Popham et al., 2022) and that time since displacement can influence outcomes, in a positive direction in high-income contexts (Müller et al., 2019) compared to a negative direction in some humanitarian contexts (Nasıroğlu et al., 2018). However, research on predictors of refugee child outcomes over time remains limited, and little is known regarding how the relationships between different aspects of the environment and children's mental health might change in the years after displacement. For example, whether some predictors become less important the longer a child has been living in the host country. This gap is particularly prominent in refugee settings in LMICs, in which there are more likely to be significant changes in the environment, including adverse experiences such as settlement raids (Human Rights Watch, 2018).
Age likely plays a crucial role here. Children may be particularly vulnerable to the challenges of war and displacement, due to limited control over their experiences combined with greater susceptibility to the biological and psychological embedding of experiences (Gee & Casey, 2015; Lundberg & Wuermli, 2012). Early experiences of adversity and early levels of functioning can have cascading effects both on future functioning and future exposure to more environmental adversity (Obradović & Hipwell, 2010). Moreover, the developmental period a child is in may affect what contexts they are exposed to (e.g., education, work), how they interact with their environment (e.g., their role within their family), and the interplay between factors such as the parent–child relationship and their mental health. However, findings as to the association between developmental period and mental health in refugee children are inconsistent. Some find that older children are more likely to show PTSD and internalizing symptoms and less likely to show externalizing problems, while others report no association, or find older age to be protective against all symptoms (Ahmad et al., 2015; Betancourt, Yudron, et al., 2012; Braun-Lewensohn & Al-Sayed, 2018; Eruyar et al., 2018; Uysal et al., 2022). These inconsistencies could be explained by the relationship between age and other predictors such as peer problems or war exposure (Eruyar et al., 2018).
The present study
The overall objective of the current study was to address the gaps in the literature reviewed above to understand how specific features of the refugee and family environments impact refugee child mental health via direct and indirect effects, at two separate time points a year apart. We investigated these effects in a unique and novel large sample of Syrian refugee children living in informal refugee settlements in Lebanon, to address the lack of research in this population despite its large number (UNHCR, 2022). Our specific aims were to test two main hypotheses: (1) that maternal mental health mediates the effects of the refugee environment on children, and (2) that the parent–child relationship mediates the effects of maternal mental health on child outcomes. Due to research showing that predictors of child functioning can differ between outcomes (Eruyar et al., 2018), we investigated these pathways separately for symptoms of child PTSD, depression, and externalizing behavior problems, using a combination of exploratory and confirmatory approaches. We determined inclusion of predictors in each outcome model using exploratory analysis, by testing bivariate correlations between hypothesized predictors and child mental health and only including those that correlated significantly. We then parceled items to create indicators for the latent factors of interest, based on exploratory factor analysis. Finally, we structured the models predicting child outcomes according to an ecological systems-based framework (Figure 2) as a confirmatory analysis. Physical and social aspects of the refugee environment were included as separate factors in order to untangle their specific effects on outcomes in mothers and children based on evidence that neighborhood cohesion or sense of community (i.e., the social environment) is distinct from structural disadvantage (i.e., the physical environment) in low-income contexts (Walton, 2016). Unfortunately, we were not able to test the effects of interest over time due to data constraints. This limits us in the conclusions we can make about mediating paths (Maxwell & Cole, 2007). While we use the term “predictors” we cannot examine predictive associations. However, given the novelty of the sample and context, investigating these associations cross-sectionally is a valuable contribution in combination with other refugee research. Finally, we leverage having data from two time points, which is unusual for refugee research in LMICs (Scharpf et al., 2021), to test whether the direct and indirect associations we observe replicate across time by testing separate models at Wave 1 and again at Wave 2 one year later.
METHOD
Study design
We tested the associations between the wider and family environments and child mental health using structural equation modeling (SEM) separately in two waves of data 1 year apart from Syrian refugee child–mother dyads. After identifying predictors of refugee child mental health from the literature, we used an exploratory approach (bivariate correlations and model-fit comparison) to narrow the selection of variables for the final models. We then tested direct predictors of and mediating pathways to child PTSD, depression, and externalizing behavior problems in separate models at Wave 1 and Wave 2. Ethical approval was granted by the Institutional Review Board of the University of Balamand/Saint George University Hospital, 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. Free mental health services were offered to anyone from participating communities, regardless of participation, as an additional service not directly related to the study aims.
Setting and participants
Syrian refugee families living in informal tented settlements in the Beqaa region of Lebanon were recruited using purposive cluster sampling, approaching settlements representing a range of vulnerabilities according to the UNHCR vulnerability index (UNICEF et al., 2017). After receiving agreement from the community leader of each settlement, one child per family from all eligible families was invited to participate. Families were eligible if they had left Syria no more than 4 years ago, the child was aged 8–16 years, and a primary caregiver was available to participate. Where multiple children in one family were eligible, the child with the birthday closest to the recruitment date was selected, to avoid selection bias. This broad age range allowed us to control for the impact of age. We did not recruit children younger than 8 years due to concerns about questionnaire comprehension, and did not recruit children older than 16 as they were more likely to be in full-time work, have married, or no longer live with their family. We recruited families that had left Syria no more than 4 years prior to ensure families had left after the war started and to maximize the likelihood they remembered their war experiences.
Informed consent and assent were given by each caregiver and child, respectively, and families were financially compensated for their time. Participants were provided written, easy-to-read, and verbal versions of the information sheet to account for low literacy levels and were informed about the financial compensation after giving consent to reduce the risk of coercion to participate. 78% of initially approached families were interested and eligible to participate. Of the eligible families who did not consent to participate, 35.2% gave no reason, and 50% did not want or were unable to provide hair and saliva samples. A total of 1600 families (94.8% of those eligible) consented and participated in the first wave of data collection in 2017–2018. A total of 1009 (63%) participated in a second wave 1 year later (2018–2019). Questionnaire data were collected via interviews in settlements by a team of local, trained, native Arabic-speaking interviewers to reduce barriers to participation. Interviews lasted approximately 50–60 min. Where the same caregiver was not available at both Waves 1 and 2, a different caregiver completed the Wave 2 measures. For further details on the sample and recruitment, see the cohort profile (McEwen et al., 2022).
As the majority of participating caregivers were female, we restricted the sample to those child–caregiver dyads in which the mother or another female caregiver (e.g., stepmother and grandmother; their “maternal caregiver”) completed the caregiver questionnaire (94.7%), and was the subject of the child's report on the parent–child relationship (91.5%), in order to investigate the specific relationship between maternal mental health and parenting. We additionally restricted the Wave 2 sample to those dyads with the same participating caregiver at both waves (92.2%), so all Wave 2 participants had Wave 1 data to facilitate comparison across waves.
With these restrictions, and excluding 7 children who participated twice in a single wave, 2 later identified as ineligible, and 17 missing key demographic or symptom data, the final sample included 1446 children (52.1% female; Mage = 11.39, SD = 2.41) and their maternal caregiver (97.2% mother) at Wave 1, and 872 children (53.1% female; Mage = 12.17, SD = 2.30) and their maternal caregiver (98.6% mother) at Wave 2. The majority of children and caregivers were of Syrian nationality (98.5%) and most caregivers reported being Sunni Muslim (97.1%). We did not ask about ethnicity based on feedback that the study sample would more likely identify in terms of their religion than ethnicity. 55.3% of the Wave 1 sample scored above the clinical cut-off (McEwen et al., 2020) for PTSD, 37.5% for depression, and 43.3% for externalizing. At Wave 2, 34.7% scored above the cut-off for PTSD, 26.7% for depression, and 43.5% for externalizing. In total, 79.7% of the Wave 1 and 67% of the Wave 2 samples had clinical levels of symptoms for at least one outcome.
At Wave 1, 46.5% had left Syria in the past 3 years, and the remainder more than 3 years previously. The final Wave 1 sample did not significantly differ from the original sample before exclusions on any of the included measures. Participants with Wave 2 data had slightly higher maternal PTSD compared to the Wave 1 sample (t(1186) = −2.41, p = .016), but the effect size was small (d = −0.13) and there were no other differences between the samples (Table 1). However, average scores on the majority of predictors were significantly different between Wave 1 and Wave 2 (Table 1). For the sake of simplicity, we will refer to all caregivers included in the analyses as the child's “mother” from this point on.
| Measure | Whole-sample baseline scores (n = 1446), M (SD) | Retained sample baseline scores (n = 872), M (SD) | Retained sample follow-up scores (n = 872), M (SD) | Lost to follow-up vs. retained sample baseline scores | Baseline vs. follow-up scores, Whole samplea (retained sample) | ||
|---|---|---|---|---|---|---|---|
| t/χ2 | Effect size (Cohen's d/Somers' d) | t | Effect size (Cohen's d) | ||||
| Child age | 11.39 (2.41) | 11.18 (2.30) | - | 4.19*** | 0.23 | - | - |
| Child gender, % female | 52.1% | 53.1% | - | 0.82 | 0.03b | - | - |
| Child nationality | |||||||
| Syrian | 98.5% | 98.6% | - | - | - | - | - |
| Lebanese | 0.5% | 0.3% | - | - | - | - | - |
| Palestinian | 0.8% | 0.9% | - | - | - | - | - |
| Iraqi | 0.1% | 0.1% | - | - | - | - | - |
| Other | 0.1% | - | - | - | - | - | - |
| Caregiver religion | |||||||
| Sunni | 97.1% | 96.8% | - | - | - | - | - |
| Shia | 0.4% | 0.5% | - | - | - | - | - |
| Other/none/prefer not to say | 2.5% | 2.8% | - | - | - | - | - |
| Time since leaving Syria, % ≤ 3 years | 46.5% | 41.7% | - | 19.28*** | 0.12b | - | - |
| Child war exposure | 9.57 (5.51) | 9.55 (5.48) | - | 0.24 | 0.01 | - | - |
| Child PTSD | 15.70 (12.22) | 15.44 (12.26) | 10.72 (13.01) | 1.01 | 0.05 | 9.90*** (8.55***) | 0.40 (0.29) |
| Child depression | 8.08 (6.99) | 8.01 (7.04) | 6.23 (6.76) | 0.46 | 0.02 | 7.10*** (6.27***) | 0.27 (0.21) |
| Child externalizing | 11.64 (6.32) | 11.70 (6.25) | 11.22 (6.60) | −0.44 | −0.02 | 2.37* (2.19*) | 0.06 (0.07) |
| Child maltreatment | 11.79 (12.09) | 11.80 (12.18) | 8.45 (10.62) | −0.01 | 0.00 | 7.66*** (7.18***) | 0.29 (0.24) |
| Psychological control | 11.21 (2.58) | 11.21 (2.58) | 10.63 (2.16) | −0.11 | −0.01 | 6.47*** (6.05***) | 0.24 (0.20) |
| Parent–child conflict | 6.05 (3.21) | 5.95 (3.16) | 7.26 (3.86) | 1.36 | 0.07 | −8.46*** (−8.25***) | −0.34 (−0.28) |
| Acceptance | 27.22 (4.00) | 27.24 (3.95) | 27.48 (4.30) | −0.25 | −0.01 | −1.62 (−1.49) | −0.06 (−0.05) |
| Family support | 6.01 (0.87) | 6.04 (0.85) | 6.13 (0.89) | −1.63 | −0.09 | −3.33** (−2.36*) | −0.14 (−0.08) |
| Maternal PTSD | 33.54 (17.78) | 34.46 (17.41) | 24.45 (18.13) | −2.41* | −0.13 | 12.95*** (12.92***) | 0.50 (0.44) |
| Maternal depression | 15.32 (6.50) | 15.40 (6.41) | 13.95 (7.39) | −0.57 | −0.03 | 5.20*** (5.18***) | 0.20 (0.18) |
| Maternal anxiety | 8.20 (5.35) | 8.36 (5.28) | 6.88 (5.28) | −1.35 | −0.07 | 6.44*** (6.67***) | 0.25 (0.23) |
| Livelihood | 2.32 (1.04) | 2.34 (1.02) | 2.75 (1.02) | −0.82 | −0.04 | −10.24*** (−8.90***) | −0.41 (−0.30) |
| Basic needs | 3.33 (0.81) | 3.33 (0.82) | 3.59 (0.85) | −0.42 | −0.02 | −8.32*** (−7.03***) | −0.32 (−0.24) |
| Housing | 3.72 (0.80) | 3.75 (0.77) | 3.84 (0.65) | −1.84 | −0.10 | −4.30*** (−2.76**) | −0.16 (−0.09) |
| Service access | 2.95 (0.91) | 2.97 (0.90) | 3.07 (1.05) | −1.03 | −0.06 | −2.84** (−2.46*) | −0.13 (−0.08) |
| Community | 3.81 (0.76) | 3.84 (0.75) | 3.85 (0.82) | −1.92 | −0.10 | −1.26 (−0.30) | −0.05 (−0.01) |
| Collective efficacy | 31.97 (6.48) | 32.02 (6.32) | 31.58 (7.90) | −0.33 | −0.02 | 1.39 (1.43) | 0.05 (0.05) |
- a Partially overlapping samples t-test with Welch's df to compare mean scores at baseline (n = 1446) and follow-up (n = 872) using whole sample at both time points (Derrick, Toher, & White, 2017).
- b Somers' d, all other effect sizes shown are Cohen's d.
- * p < .05;
- ** p < .01;
- *** p < .001.
Measures
All participants were interviewed in their homes in the settlements. Different interviewers conducted the child and mother interviews simultaneously, in different rooms wherever possible to maintain privacy. All measures were in Arabic, piloted in the same population, and modified accordingly to improve comprehension and performance where necessary (McEwen et al., 2022). Some measures were exclusively child or mother reported, while others were reported by both (Table S1; McEwen et al., 2022).
Child mental health outcomes
The outcomes of interest were self-reported PTSD (α = .91; Foa et al., 2001), self-reported depression (α = .88; Faulstich et al., 1986), and mother-reported externalizing behavior problems measured using the externalizing subscale of the Strengths and Difficulties Questionnaire (Goodman, 1997) and additional items related to conduct disorder and oppositional defiant disorder administered separately (α = .80; McEwen et al., 2022). Scales were chosen according to reliability and validity in similar populations and pilot tested with Syrian refugees in Lebanon (McEwen et al., 2022). The CES-DC was abridged to 10 items and minor changes to phrasing (including Arabic dialect) were made to the CPSS and abridged CES-DC based on pilot data (McEwen et al., 2022).
Family environment
We investigated two key aspects of the family environment: maternal mental health and the parent–child relationship (including parenting). Maternal mental health was measured using the PTSD checklist for DSM-5 (α = .93; Blevins et al., 2015), the anxiety subscale of the Depression Anxiety and Stress Scale (α = .85; Henry & Crawford, 2005), and the Center for Epidemiologic Studies Short Depression Scale (α = .84; Radloff, 1977).
Negative parenting was captured with the ISPCAN Child Abuse Screening Tool (α = .86; Runyan et al., 2009) as an index of child maltreatment, in addition to measures of psychological control (α = .64; Barber et al., 2012), and parent–child conflict (α = .73; Barber, 1999). We excluded the final item of the child maltreatment scale (“any private events that have happened to you that you don't want to describe”) because over 95% of children said this did not happen to them.
Supportive parenting was captured using the acceptance subscale of the Child Report of Parent Behavior Inventory (α = .89; Schaefer, 1965) and perceived family social support as measured by the family subscale of the Multidimensional Scale of Perceived Social Support for Arab American Adolescents (α = .68; Ramaswamy et al., 2009).
Refugee environment and war exposure
Mothers reported on the wider refugee environment, as measured by subscales of a newly developed scale (McEwen et al., 2022). The physical environment was captured using the livelihood, basic needs, housing, and service access subscales (α = .83). The wider social environment was captured using the practical and emotional community support subscale combined with a measure of collective efficacy, or the extent to which caregivers feel the neighborhood is close-knit and supportive (α = .78; Sampson et al., 1997).
War exposure was measured using the War Events Questionnaire (WEQ), a 25-item checklist measuring the number of different types of war events experienced (α = .89; Karam et al., 1999). Both the child and their mother completed the questionnaire about the child's experiences. Because self-report can be less reliable in younger children (Oh et al., 2018), child and mother responses were combined such that if either one reported that the child experienced an event, the event was considered to have occurred. As maternal war exposure is likely to have effects on maternal mental health and may interact with their experience of current stressors (Sim, Bowes, & Gardner, 2018), we deemed it important to control maternal war exposure as well as child war exposure in our models. However, we did not have a measure of maternal war exposure, so we used the combined child–mother-reported child exposure measure as a proxy, examining its association with maternal in addition to child mental health. The child war exposure measure likely relates closely to the mother's own exposure, as mothers would be more likely to report on events they were also present for. We excluded six follow-up items that asked whether the child witnessed an event because mothers were not asked those questions. For more measures information, please see Table S1 in the Supplement.
Statistical methods
Exploratory analysis: bivariate correlations
All analyses were conducted in RStudio 2021.09.2 + 382. To determine which of the literature-derived variables to include in our final models, we ran bivariate correlations to investigate the relationship between the three dimensions of child mental health (PTSD, depression, and externalizing) and all other measures, as well as the relationships between the other measures (refugee environment, maternal mental health, and parent–child relationship).
Item parceling
To reduce the complexity and improve the stability of our models, we conducted subset item parceling to create latent factors capturing war exposure, the physical environment, social environment, negative parenting, supportive parenting, and child PTSD, depression, and externalizing. We created three parcels for each latent construct to maximize stability, in line with recommendations from the literature (Little et al., 2022). We used item loadings from exploratory factor analyses as a guide for which parcel to assign each item to, created parcels via sums of items, and then tested the fit of each latent measure at each wave with confirmatory factor analysis to see if it adequately replicated the underlying covariance structure (Hu & Bentler, 1999). Finally, we compared fully unconstrained models with models constrained to be the same over time, to test for temporal invariance. The maternal mental health factor was made up of the maternal PTSD, depression, and anxiety scales, without item parceling. See Table S2 for the final item parcels.
Structural equation models
Using the latent factors, we created three independent SEMs to identify the effects of predictors of interest (identified from the bivariate correlations) on child PTSD, depression, and externalizing behavior problems. These were structured to reflect the different ecological system levels and test our hypotheses that maternal mental health mediates the effects of the refugee environment on children, and that the parent–child relationship mediates the effects of maternal mental health on child mental health (Figure 2). In all models, we controlled for (1) the effects of child age and gender on child mental health and the parent–child relationship, (2) the effects of time since leaving Syria on child and maternal mental health and the refugee environment, and (3) the effects of war exposure on child and caregiver mental health. Models were run with lavaan (Rosseel, 2012) using full-information maximum-likelihood estimation to account for missing data. Mediation was tested using bootstrapping with 1000 replications.
We initially created models incorporating both waves in which we tested indirect effects using cross-lagged paths from Wave 1 to Wave 2 (Little et al., 2007), but saw no evidence of indirect effects over time. This was unsurprising due to the long time between waves (approximately 12 months) during which significant changes in environmental quality occurred (Table 1), possibly as a result of additional events for which we did not have specific measures (e.g., settlement raids; Human Rights Watch, 2018). Moreover, previous tests of cross-lagged paths between several similar predictors and the mental health outcomes at Wave 1 and Wave 2 in the same sample had shown small direct effects (Popham et al., 2022). We therefore ran each model separately for the two waves, to identify whether the effects identified at Wave 1 were replicated 1 year later at Wave 2.
However, to investigate the effects of child development, we conducted a sensitivity analysis, splitting the sample into children aged 8–11 (n = 794) and aged 12–16 (n = 652). This age grouping was chosen so that the groups were as similar in size as possible and based on evidence that Syrian refugee children living in informal settlements in the Beqaa Valley start working at around 11 years of age (Habib et al., 2019). We repeated the Wave 1 models grouped by this categorization to determine whether any observed effects depended on period of development. We did not run a multi-group analysis for Wave 2 as the sample was too small.
RESULTS
Measurement model
The measurement models combining all latent factors (war exposure, wider physical and social environment, maternal mental health, and positive and negative parent–child relationship) for the child PTSD, depression, and externalizing models all adequately fit the data at Wave 1 (CFI = 0.93–0.94; RMSEA = 0.05 (90% CI: 0.04–0.05); SRMR = 0.04) and Wave 2 (CFI = 0.93–0.95; RMSEA = 0.05 (90% CI: 0.04–0.06); SRMR = 0.05). All factor loadings were statistically significant (Figures S1–S3). The models all demonstrated metric invariance across waves according to the criterion that changes in CFI ≤0.01 (Table S4; Cheung & Rensvold, 2009), allowing us to compare covariance between waves.
Post-traumatic stress
According to bivariate correlation matrices, child PTSD symptom scores showed significant correlations with maternal mental health, child maltreatment, psychological control, and parent–child conflict at both waves. Acceptance showed a small correlation with PTSD at Wave 1 only while perceived family support only showed small correlations with PTSD at Wave 2. Finally, PTSD showed small correlations with basic needs and service access at Wave 1 and larger correlations with livelihood, service access, and community at Wave 2 (Table 2).
| 1. | 2. | 3. | 4. | 5. | 6. | 7. | 8. | 9. | 10. | 11. | 12. | 13. | 14. | 15. | 16. | 17. | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1. Child PTSD | - | .39*** | .12*** | .30*** | .15*** | .24*** | −.06 | −.07* | .16*** | .11** | .16*** | −.13*** | .01 | −.01 | −.10** | −.09** | −.01 |
| 2. Child depression | .40*** | - | .09** | .28*** | .26*** | .16*** | −.19*** | −.28*** | .22*** | .18*** | .25*** | −.13*** | −.14*** | −.18*** | .03 | −.09** | −.02 |
| 3. Child externalizing | .06* | .08** | - | .15*** | .12*** | .12*** | −.15*** | −.01 | .33*** | .26*** | .42*** | −.15*** | −.13*** | −.06 | −.21*** | −.15*** | −.15*** |
| 4. Child maltreatment | .35*** | .26*** | .16*** | - | .37*** | .34*** | −.37*** | −.23*** | .15*** | .10** | .16*** | −.06 | .03 | .03 | −.11*** | −.05 | −.03 |
| 5. Psychological control | .22*** | .17*** | .14*** | .33*** | - | .31*** | −.43*** | −.26*** | .09** | .03 | .12*** | −.01 | .01 | −.01 | .00 | −.11** | −.07* |
| 6. Parent–child conflict | .19*** | .14*** | .09*** | .32*** | .30*** | - | −.31*** | −.14*** | .07* | −.07* | .05 | .04 | .00 | −.03 | −.20*** | −.02 | −.07* |
| 7. Acceptance | −.08** | −.19*** | −.12*** | −.20*** | −.36*** | −.20*** | - | .47*** | −.15*** | −.04 | −.13*** | .06 | .03 | .04 | .07 | .07* | .05 |
| 8. Family support | −.05 | −.19*** | −.09** | −.12*** | −.15*** | −.11*** | .33*** | - | −.13*** | −.04 | −.11** | .05 | .02 | .11** | −.03 | .03 | .04 |
| 9. Maternal PTSD | .06* | .22*** | .29*** | .05 | .07** | .01 | −.12*** | −.13*** | - | .57*** | .62*** | −.25*** | −.15*** | −.23*** | −.17*** | −.12*** | −.04 |
| 10. Maternal anxiety | .10*** | .21*** | .21*** | .02 | .00 | .01 | −.05* | −.04 | .58*** | - | .64*** | −.25*** | −.13*** | −.19*** | −.13*** | −.08* | .02 |
| 11. Maternal depression | .17*** | .21*** | .26*** | .15*** | .06* | .03 | −.06* | −.02 | .56*** | .64*** | - | −.27*** | −.22*** | −.24*** | −.17*** | −.24*** | −.16*** |
| 12. Livelihood | −.01 | −.02 | −.01 | −.02 | −.01 | −.04 | −.05 | −.01 | −.04 | −.07* | −.18*** | - | .34*** | .19*** | .22*** | .09** | .00 |
| 13. Basic needs | .05* | −.05 | −.02 | .05 | .03 | .07* | .02 | .02 | −.14*** | −.16*** | −.19*** | .37*** | - | .50*** | .21*** | .18*** | .16*** |
| 14. Housing | .01 | −.07** | .01 | .03 | .04 | .10*** | .01 | −.05 | −.07** | −.13*** | −.15*** | .33*** | .56*** | - | .25*** | .18*** | .13*** |
| 15. Service access | −.05* | −.05 | .04 | −.01 | .04 | −.03 | .00 | .00 | −.07** | −.08** | −.18*** | .32*** | .44*** | .44*** | - | .14*** | .15*** |
| 16. Community | −.01 | −.07** | −.09*** | −.02 | −.01 | .03 | .00 | −.06* | −.07** | −.17*** | −.20*** | .08** | .21*** | .34*** | .27*** | - | .48*** |
| 17. Collective efficacy | .02 | −.08** | −.10*** | .04 | .01 | .08** | .04 | .06* | −.16*** | −.17*** | −.19*** | .15*** | .18*** | .28*** | .17*** | .35*** | - |
- Note: Pearson correlation coefficients. Lower half are correlations at Wave 1, upper half are correlations at Wave 2.
- * p < .05;
- ** p < .01;
- *** p < .001.
The final PTSD SEM (Figure 3a; Tables S3 and S5) retained the physical home environment, the wider social environment, maternal mental health, and the negative parent–child relationship factor. We did not include supportive parenting due to the small and inconsistent associations between the supportive parenting scales and child PTSD. The model showed adequate fit at Wave 1 (χ2[171] = 932.68, p < .001; CFI = 0.90; RMSEA = 0.06 (90% CI: 0.05–0.06); SRMR = 0.06) and Wave 2 (χ2[171] = 656.29, p < .001; CFI = 0.91; RMSEA = 0.06 (90% CI: 0.05–0.06); SRMR = 0.06). War exposure (βw1 = .33, pw1 < .001; βw2 = .17, pw2 = .001) and the negative parent–child relationship factor (βw1 = .39, pw1 < .001; βw2 = .34, pw2 < .001) had direct effects on child PTSD symptoms at both waves. Maternal mental health also had an indirect effect on child PTSD via negative parent–child relationship at both waves (βw1 = .05, pw1 = .002; βw2 = .08, pw2 < .001). The wider social environment had an indirect effect on child PTSD via the pathway through maternal mental health and the negative parent–child relationship factor at both waves (βw1 = −.02, pw1 = .009; βw2 = −.01, pw2 = .017). The wider physical environment had a similar effect on child PTSD via the same pathway (βw2 = −.03, pw2 = .001) at Wave 2 only.
Child age was associated with worse PTSD, psychological control, and parent–child conflict. When the Wave 1 model was run separately for children aged 8–11 and 12–16 (Table S3), negative parenting remained significantly associated with child PTSD for both groups. The indirect effect of maternal mental health on child PTSD via negative parenting (β8–11 = .04, p8–11 = .037; β12–16 = .05, p = .034) also remained significant for both groups. The indirect effect of the social environment on child PTSD via maternal mental health and negative parenting was not significant for either group.
Depression
Depression showed significant bivariate correlations with maternal mental health, child maltreatment, psychological control, parent–child conflict, acceptance, and perceived family support at both waves. In terms of the refugee environment, depression was correlated with housing and community at both waves, with collective efficacy at Wave 1, and with livelihood and basic needs at Wave 2 (Table 2).
The depression SEM (Figure 3b; Tables S3 and S5) retained all predictors and showed adequate fit at Wave 1 (χ2[229] = 1132.39, p < .001; CFI = 0.90; RMSEA = 0.05 (90% CI: 0.05–0.06); SRMR = 0.05) and Wave 2 (χ2[229] = 779.47, p < .001; CFI = 0.91; RMSEA = 0.05 (90% CI: 0.05–0.06); SRMR = 0.06). War exposure (βw1 = .11, pw1 = .002; βw2 = .14, pw2 = .001), maternal mental health (βw1 = .19, pw1 < .001; βw2 = .17, pw2 < .001), and negative parenting (βw1 = .30, pw1 < .001; βw2 = .43, pw2 < .001) had direct effects on child depression at both waves. Supportive parenting had a direct effect on depression at Wave 1 only (βw1 = −.09, pw1 = .006). At both waves, maternal mental health had an indirect effect on depression via negative parenting (βw1 = .04, pw1 = .001; βw2 = .11, pw2 < .001). Maternal mental health also had an indirect effect on child depression via supportive parenting at Wave 1 (βw1 = .01, pw1 = .049). The wider social environment had indirect effects on depression via maternal mental health (βw1 = −.07, pw1 < .001) and via the pathway through both maternal mental health and negative parenting (βw1 = −.01, pw1 = .006) at Wave 1. The physical environment had indirect effects on depression via maternal mental health (βw2 = −.06, pw2 = .003) and the maternal mental health—negative parenting pathway (βw2 = −.04, pw2 = .001) at Wave 2.
Older children scored higher on depression, psychological control, and parent–child conflict, and lower on supportive parenting. When the Wave 1 model was run separately for children aged 8–11 and 12–16 (Table S3), supportive parenting was associated with lower depression for children aged 11 and below only (β = −.16, p = .002). The indirect paths from the social environment to depression via maternal mental health (β8–11 = −.06, p8–11 = .014; β12–16 = −.07, p12–16 = .015) and from maternal mental health via negative parenting (β8–11 = .03, p8–11 = .026; β12–16 = .04, p12–16 = .037) were significant for both groups. The indirect effects of mental health on depression via supportive parenting and from the wider social environment via maternal mental health and negative parenting were not significant for either group.
Externalizing behavior problems
Externalizing behavior problems showed significant bivariate correlations with maternal mental health, child maltreatment, psychological control, parent–child conflict, and maternal acceptance at both waves. Perceived family support correlated with externalizing at Wave 1 only. None of the physical environment measures correlated with externalizing at Wave 1, so we did not test the direct pathway from physical environment to externalizing behavior problems (Table 2).
The final externalizing SEM (Figure 3c; Tables S3 and S5) retained the physical and social environments, maternal mental health, negative parenting, and supportive parenting. It showed acceptable fit at Wave 1 (χ2[230] = 1146.07, p < .001; CFI = 0.89; RMSEA = 0.05 (90% CI: 0.05–0.06); SRMR = 0.05) and Wave 2 (χ2[230] = 808.18, p < .001; CFI = 0.90; RMSEA = 0.05 (90% CI: 0.05–0.06); SRMR = 0.06). Maternal mental health had a direct effect on externalizing at both waves (βw1 = .30, pw1 < .001; βw2 = .42, pw2 < .001). War exposure (βw1 = .08, pw1 = .036), negative parenting (βw1 = .15, pw1 < .001), and supportive parenting (βw1 = −.07, pw1 = .045) had direct effects on externalizing at Wave 1 only. The wider social environment had an indirect effect on externalizing via maternal mental health at both waves (βw1 = −.11, pw1 < .001; βw2 = −.07, pw2 = .003), while the physical environment also had an indirect effect via maternal mental health at Wave 2 only (βw2 = −.15, pw2 < .001). Maternal mental health had an indirect effect on child externalizing via negative parenting (βw1 = .02, pw1 = .010) and the wider social environment had an indirect effect via maternal mental health and negative parenting (βw1 = −.01, pw1 = .019) at Wave 1 only.
Child age was negatively associated with externalizing and maternal acceptance, and positively with psychological control and parent–child conflict. When the Wave 1 model was run separately for children aged 8–11 and 12–16 (Table S3), supportive parenting had a significant direct effect on children aged 11 and below only (β = −.12, p = .032). The indirect effect of social environment via maternal mental health on externalizing (β8–11 = −.14, p8–11 = .001; β12–16 = −.08, p12–16 = .020) remained significant for both groups. War exposure was only a significant predictor of externalizing symptoms (β = .17, p = .006) for children aged 12 and above.
Additional pathways
In addition to the individual SEM results presented above, in all models and at both waves, war exposure and the wider social environment had direct effects on maternal mental health. The physical environment became a significant predictor of maternal mental health at Wave 2. Maternal mental health had direct effects on negative and supportive parenting at both waves. Direct effect estimates are displayed in Figure 3 and indirect effect estimates are in Table S5.
DISCUSSION
The aim of the current study was to explore the complex pathways from the refugee environment through maternal mental health and the parent–child relationship to refugee child mental health. Although we were limited from making conclusions about directionality and mediating pathways (Maxwell & Cole, 2007), we made an important step in understanding the associations between these factors in the refugee environment. We developed upon existing research using a large and unique sample of Syrian refugee children living in informal settlements in Lebanon by investigating paths across multiple ecological systems to child PTSD, depression, and externalizing symptoms, replicating models 1 year apart in the same sample, and controlling for the effects of war exposure. We also tested the effects of child age on these associations.
Our key findings can be summarized in five points: (1) The refugee environment had indirect effects on child mental health via maternal mental health; (2) Maternal mental health had direct effects on child mental health as well as indirect effects via negative parenting; (3) Certain pathways applied to all three outcomes, while others were disorder specific; (4) While key findings were replicated across time, there were some differences between time points; (5) Child age was associated with differences in mental health and the parent–child relationship.
Indirect effects of the refugee environment
In support of our family process-based hypotheses (Bronfenbrenner, 1979; Conger et al., 1992), we found that the wider social and physical environments were for the most part not directly related to child mental health, but were associated with child mental health via a mediating pathway through maternal mental health. Better social environment, as measured by whether caregivers felt the community was close-knit and provided emotional and practical support, and felt neighbors would help discipline children, was associated with better maternal mental health at both waves. In turn, better maternal mental health was associated with fewer child depression and externalizing symptoms at both waves. The indirect effect of the social environment on child depression and externalizing via maternal mental health was replicated at both waves. Better physical environment, including access to necessities such as food and clean drinking water, adequate income and accommodation, and access to services such as healthcare or transportation, was associated with better maternal mental health at Wave 2. The physical environment also had an indirect effect on child depression and externalizing via maternal mental health at Wave 2, but this was not replicated at both time points.
We cannot infer directionality from these results, but the pattern of associations fits with evidence that social support improves mothers' mental health and reduces harsh parenting (Sim et al., 2019) and that daily stressors can affect children via parental PTSD and harsh parenting (Bryant et al., 2018). For the most part, our measures of the refugee environment had no direct effects on child mental health outcomes, supporting the hypothesis that any impacts on children are likely primarily through effects cascading through more proximal systems. This may explain why findings from previous research have been somewhat inconsistent in terms of whether the quality of the refugee environment predicts child outcomes (Arakelyan & Ager, 2021; Scharpf et al., 2021).
Direct and indirect effects of maternal mental health
Focusing on the family environment, we also found support for our second hypothesis at both waves: the parent–child relationship mediated the association between maternal mental health and child outcomes. In particular, worse maternal mental health was associated with increased negative parenting, which in turn was associated with worse child PTSD and depression, in keeping with multiple previous refugee studies (Bryant et al., 2018; Sangalang et al., 2017; Sim, Bowes, & Gardner, 2018). However, in contrast with some previous work (Sangalang et al., 2017), we found that maternal mental health also had a direct association with child depression and externalizing symptoms even when controlling for the effects of parenting, implying that maternal and child mental health are associated in ways not captured by our measures of negative or supportive parenting. One possible explanation for these direct effects could be children witnessing maternal symptoms or distress, which can cause child distress (Rizkalla et al., 2020). It is possible that this has particularly strong effects on children in the current study sample compared to other populations due to their difficult living conditions. The families who participated in this study live in cramped, informal shelters, and the majority of children were not in school; this level of confinement may make it difficult for mothers to mask their own symptoms, thereby exacerbating the effects of maternal mental health on child outcomes (Sim, Fazel, et al., 2018). Alternatively, child mental health may impact maternal mental health, an effect which could be exacerbated by the mother's own trauma and reduced sense of parental efficacy.
Outcome-specific effects
Some of the observed effects only applied to certain child outcomes. For example, the indirect association between maternal and child mental health via negative parenting only replicated at both waves for child PTSD and depression, and not externalizing behavior problems. This is in contrast to previous findings (Bryant et al., 2018). Meanwhile, the association between maternal mental health and child externalizing was much larger compared to the other two child outcomes. This may be due in part to the different informants for the measures; child PTSD and depression were self-reported, as were the parent–child relationship measures, whereas child externalizing and maternal mental health were reported by the mother. The externalizing symptom score may therefore be less strongly associated with the child-reported parent–child relationship measures, and more strongly with maternal mental health, due to sharing the same informant. However, Harman's single-factor test and confirmatory factor analysis (Podsakoff et al., 2003) indicated that common method bias was not an issue in our data. Moreover, externalizing symptoms remained significantly associated with aspects of the parent–child relationship, and did not show consistently lower bivariate correlations with the parent–child relationship measures compared to PTSD and depression, suggesting the observed associations between the parent–child relationship and all three child outcomes are genuine.
Effects of child development
The period of development a child is in likely impacts the associations between the wider environment, family environment, and the child. We found that children aged 12–16 scored higher on PTSD, depression, maternal psychological control, and parent–child conflict compared to children aged 8–11. They also scored lower on externalizing and supportive parenting. War exposure was associated with child externalizing symptoms only in children aged 12–16, while supportive parenting was only associated with child depression and externalizing for those aged 8–11. Perhaps older children are more likely to remember and have been affected by past war exposure, while younger children might be more affected by factors within the family environment. These two groups were relatively small despite us choosing a categorization that made the size of both groups as even as possible, so reduced power could have affected these results. It is also possible that other groupings could have produced different results. Regardless, these results indicate the importance of considering child development in refugee research.
Replication of findings across time
Finally, replicating our models at two time points a year apart allowed us to test the reliability of associations. While several key findings emerged at both waves, our results also showed some differences between time points. The direct association between maternal mental health and child depression and externalizing, the indirect association between the wider social environment and child depression and externalizing via maternal mental health, the association between negative parenting and child depression and PTSD, and the indirect association between maternal mental health and child depression and PTSD via negative parenting all replicated at both waves, providing confidence as to the importance and reliability of these pathways. However, there is also important information to be gained from the differences in models between Wave 1 and Wave 2.
The key difference is the different importance of the physical environment across waves. The physical environment was associated with maternal mental health and indirectly associated with all three child outcomes but at Wave 2 only. This could be due to the cumulative effects of the refugee environment on mothers, whereby the post-migration environment quality becomes more important the longer a family is in a camp (Khamis, 2019). However, the lack of replication of these effects across both waves calls their reliability into question.
Associations within the family environment also fluctuated across waves. A general pattern that emerged from the bivariate correlations is that maternal mental health was more strongly associated with the parent–child relationship at Wave 2. This may indicate a cycle whereby worse maternal mental health leads to a worsening of the parent–child relationship and child mental health, and in turn further worsening of maternal mental health (Popham et al., 2022). Alternatively, improvements in maternal mental health could improve the mother's capacity to parent supportively, creating a positive feedback loop. However, due to the instability across waves and our lack of longitudinal analysis, we cannot make any strong conclusions. In order to understand the processes affecting refugee child mental health, the dynamic nature of family relationships and how they relate to changing environments need to be taken into consideration.
Implications and future directions
The results we have presented here have two key implications. First, as has been emphasized in most refugee research, the family environment and particularly caregiver mental health are strongly related to how a child responds and adapts to the experiences of war and displacement. However, we additionally emphasize here the importance of taking a more holistic view by also considering what impacts caregivers (Suárez-Orozco et al., 2018). Findings suggest that interventions aimed at improving child mental health should not only focus on the child or even the family but also take the wider environment into account. In contrast to previous research (Eltanamly et al., 2021), we found that supportive parenting did not have strong or consistent associations with child mental health in the models, although it was consistently negatively associated with maternal mental health. In such a challenging environment, negative factors including maternal mental health problems, child mental health problems, and strained parent–child relationships might all feed together into a negative cycle that is hard to break. Perhaps interventions are needed that support caregivers by addressing their own mental health problems, reducing external stressors, and enabling an environment in which they have the resources to parent supportively. The extent to which this is possible varies widely; aspects of the social environment could perhaps be targeted using community-level interventions, but aspects of the physical environment, for example, housing quality, require more than psychosocial intervention. Strengthening the coordination between psychosocial and other services may be the most effective way of reducing burden on families. Second, our results suggest that although different dimensions of child functioning are associated with different aspects of the environment, there are considerable commonalities between them. Hence, certain factors, such as the quality of the environment and maternal mental health, could universally help improve refugee children's mental health regardless of their specific pattern of symptoms.
However, several questions remain to be addressed by future research. Although we attempted to develop upon the refugee mental health literature which typically features cross-sectional designs, fully longitudinal designs are needed to test the directionality of the associations we identified, investigate potentially spurious associations, and better understand changes over time. For example, the effects reported here could show the impacts of child mental health on their environment, inverse to our hypotheses (Popham et al., 2022), or could be explained by shared variance with confounding factors not measured. Longitudinal studies are challenging in refugee populations in LMICs and camp settings, as the environment changes quickly, and families tend to move on. Completing multiple assessments closer together in time (e.g., on a monthly basis) would hopefully mitigate both issues.
Strengths and limitations
The current study is characterized by several significant strengths including an ecological systems approach to mental health using two waves of data in a large, challenging to reach sample of Syrian refugees in Lebanon. As the majority of refugees globally reside in LMICs, our findings may also be generalizable to many other groups of refugees. While we cannot infer directionality from our results, two waves of data enabled us to identify which associations were stable over time and infer that there is interplay among the environment, mother, parent–child relationship, and child outcomes.
However, the results should be interpreted with the following limitations in mind. First, as discussed above, our data limited us to assessing indirect effects using cross-sectional analysis. We did not test any effects longitudinally due to being limited to two waves of data, the significant change in all predictors between those two waves, and the relatively small effect sizes of cross-lagged pathways between waves for the predictors of interest and mental health outcomes from previous work in the same sample (Popham et al., 2022). Given these conditions, we did not expect to be able to observe indirect effects across time using the cross-lagged pathway approach. However, cross-sectional tests of mediation are vulnerable to bias or spurious associations (Maxwell & Cole). The models are also fairly complex, containing several measures, which could increase risk of Type 1 errors.
Second, our use of multiple informants (i.e., mother and child) means that associations between measures differ in the degree to which they are confounded by common method variance. This creates some difficulty in interpreting results, as discussed above in regards to externalizing behavior problems. However, our choice to use caregiver-reported externalizing and child-reported depression and PTSD measures was based on careful piloting that demonstrated they were the most reliable and valid measures of these domains. We can also see that our results are not entirely explained by common method variance, as caregiver-reported maternal mental health was strongly related to child-reported parent–child relationship, PTSD, and depression scores, and factor analysis techniques indicated there was not a concerning amount of common method bias. It is also important to note here that our measure of maltreatment related to maltreatment perpetrated by any adults in the home, not just the mother. Therefore, it could be the case that, for example, the father (or another adult) was responsible for violence in the home, which negatively impacted both mother and child. This links back to the need for further longitudinal research, to untangle the directionality of these associations.
Finally, we did not have a direct measure of maternal war exposure. However, as described in the Methods section, it is likely that child and mother's war experiences were highly correlated. The strong association that we observe between child war exposure and maternal mental health supports our theory that it can be considered a proxy measure, and emphasizes the importance of attempting to control for the effects of maternal war exposure when looking at the impacts of other environmental stressors. However, as it is a proxy, we only use that pathway as a control, and have not made any specific conclusions as to the association between war exposure and maternal mental health.
CONCLUSION
Using reports from a large sample of Syrian refugee families living in informal tented settlements in Lebanon at two time points 1 year apart, we identified a range of direct and indirect associations between the refugee and family environments and children's symptoms of PTSD, depression, and externalizing problems. All three outcomes were directly associated with maternal mental health and aspects of the parent–child relationship, and were indirectly associated with characteristics of the refugee environment, the effects of which were mediated through maternal mental health. The effects of maternal mental health on child PTSD and depression were in turn mediated by aspects of the parent–child relationship, particularly negative parenting, indicating a possible cascading of effects from environmental stressors through the family system, in line with family process models. However, we were unable to test the directionality of these effects. Some pathways differed across time points and scores on most measures were substantially different when measured 12 months later, demonstrating the highly dynamic nature of refugee settings in LMICs. Longitudinal research, particularly in less stable humanitarian settings, continues to be a key priority to understanding the reality of refugees' situations, as well as the main factors affecting outcomes.
ACKNOWLEDGMENTS
We warmly thank all participating families for their participation. We thank Patricia Moghames, Stephanie Legoff, Nicolas Puvis, and Zeina Hassan, and all other members of the BIOPATH team (https://www.qmul.ac.uk/sbbs/about-us/our-departments/psychology/global-mental-health/meet-the-team/) for their dedication, hard work, and insights.
FUNDING INFORMATION
The BIOPATH study was funded by the Eunice Shriver National Institute of Child Health & Human Development (R01HD083387). The study was sponsored by Queen Mary University of London (QMUL). The funder and sponsor played no role in study design, in the collection, analysis, or interpretation of data, in the writing of the report, or in the decision to submit the article for publication.
CONFLICT OF INTEREST STATEMENT
Dr. Pluess reports no financial relationships with commercial interests.
Open Research
DATA AVAILABILITY STATEMENT
The data and analytic code necessary to reproduce the analyses presented here are available from the corresponding author upon request, as are the materials necessary to attempt to replicate the findings. The analyses presented here were not preregistered.
