Abstract
Parental violence poses a considerable, yet mitigable risk for the mental health and well-being of refugee children living in resource-poor refugee camps. However, little is known about potential risk factors for parental violence in these settings. Using an ecological systems perspective and a multi-informant approach, we investigated ontogenic (parental childhood experiences of violence), microsystem (parents’ and children’s psychopathology) and exosystem (families’ monthly household income) risk factors for child-directed parental violence in a sample of 226 Burundian families living in refugee camps in Tanzania. Data were collected through individual structured clinical interviews with mothers, fathers, and children. In the child-report path model [χ2 (6) = 7.752, p = .257, comparative fit index (CFI) = 0.99, root mean square error of approximation (RMSEA) = 0.036 (p = .562)], children’s posttraumatic stress disorder (PTSD) symptoms, externalizing symptoms and paternal PTSD symptoms were positively associated with violence by both parents. Maternal psychosocial impairment was positively associated with child-reported paternal violence. In the parent-report path model [χ2 (6) = 7.789, p = .254, CFI = 0.97, RMSEA = 0.036 (p = .535)], children’s externalizing problems as well as a lower monthly household income were positively related to maternal violence. Each parent’s childhood victimization was positively linked to their use of violence against children. Maternal psychosocial impairment and paternal alcohol abuse were positively associated with paternal violence. Child and paternal psychopathology, maternal psychosocial impairment, parents’ childhood victimization, and families’ socioeconomic status may be important targets for prevention and intervention approaches aiming to reduce parental violence against refugee children living in camps.
According to the latest report of the United Nations High Commissioner of Refugees (UNHCR), the number of refugees in the world has recently exceeded 26 million with almost half of them being children and adolescents (UNHCR, 2019). About 85% of all refugees fled to neighboring low- and middle-income countries, where they often live in refugee camps. While many refugees have endured conflict-related violence in their homes and dangerous flights, the precarious living conditions in camps pose an additional risk to their mental health (Miller & Rasmussen, 2010; Reed et al., 2012). Accordingly, high prevalence rates for mental health problems, including posttraumatic stress disorder (PTSD), depression, and anxiety, have been reported among camp-based refugee children and adults (Riley et al., 2017; Vossoughi et al., 2018).
Another risk factor for conflict-affected children’s mental health and well-being that has so far received only limited attention in research and interventions is violence and maltreatment within the family (Catani, 2018; Miller & Jordans, 2016), though it may be a risk factor that is particularly responsive to intervention. Findings in high-income countries have stressed that children who experience violence by parents are at an increased risk for a range of mental health problems (Gilbert et al., 2009). Therefore, it is essential to identify factors that are associated with violent parenting among refugee families living in camps.
The ecological integration model (Belsky, 1980) poses a suitable framework to investigate potential risk factors on different levels of families’ social ecology, which are embedded in and interact with each other. Such a framework has been used to conceptualize risk factors for violent parenting in nonrefugee (Stith et al., 2009) and refugee families in high-income countries (Timshel et al., 2017). The focus of this study lies on the ontogenic development and the microsystem, as these levels are most proximal to the context in which parent-against-child violence occurs, and mediate the impact of factors on the exo- and macrosystem (Belsky, 1980).
Ontogenic development refers to the caregivers’ experiences across the lifespan preceding the individuals becoming parents, which in turn shape the understanding of their own parenting role and influence family interactions in the microsystem. According to the “cycle of violence” hypothesis, parents who experienced violence in their family of origin will be more likely to behave violently toward their own children (Widom, 1989; Widom et al., 2008). Although a systematic review generally found support for this hypothesis, the evidence was largely based on methodologically weak studies (Thornberry et al., 2012). In particular, the use of parents’ reports for both their childhood experiences and their own perpetration is problematic. However, a study in postconflict Uganda found that maternal childhood experiences of violence were associated with increased maltreatment as reported by both mothers and children (Saile et al., 2014).
The microsystem comprises characteristics of the individual family members and all interactions within the family system (Belsky, 1980). In refugee families, psychopathology stemming from prior exposure to armed conflict and to ongoing daily hardships may be particularly relevant (Sim et al., 2018; Timshel et al., 2017). In line with a systemic view, children`s own psychopathology may increase their risk of experiencing violent parenting (Catani, 2018). For example, longitudinal studies have demonstrated reciprocal associations between parental use of physical violence and children’s externalizing problems (MacKenzie et al., 2014; Wang & Liu, 2018). Children’s PTSD symptoms, as the most direct psychological consequence of their exposure to war trauma, have not been investigated as potential risk factors, but it can be expected that PTSD-related symptoms, such as irritability, anger and concentration problems, may elicit violent parenting behavior (Catani, 2018; Timshel et al., 2017).
Regarding parental psychopathology, parental PTSD symptom severity has been shown to be associated with an increased abuse potential in refugee and other conflict-affected families (Montgomery et al., 2019; Timshel et al., 2017). PTSD-related alterations in arousal and reactivity of the adult seem to play an important role in interactions with the child and appear to be linked to hostile and insensitive parenting (van Ee et al., 2016). Meta-analytic evidence points to associations between both maternal and paternal depression with negative parenting behaviors, characterized by coercive behaviors, irritability and hostility towards the child (Lovejoy et al., 2000; Wilson & Durbin, 2010). Finally, studies conducted in postconflict settings found a strong link between paternal alcohol abuse and their increased use of violence against children (Catani et al., 2008a; Saile et al., 2014; Sriskandarajah et al., 2015). However, when investigating associations between parents’ psychopathology and aspects of their parental functioning, it is important to not only rely on parents’ self-reports, but also to consider the child’s perspective, as parents’ symptoms may affect the way they view themselves and their parenting behavior (van Ee et al., 2016).
Psychopathology is often associated with significant impairments in psychosocial functioning, which refers to a person’s capacity to fulfill different roles in the family and the society (Brissos et al., 2011). For example, studies with US military families have shown that veteran fathers’ PTSD symptoms affected children`s mental health through mothers’ secondary PTSD symptoms (Herzog et al., 2011) and also increased mothers’ parenting stress (Blow et al., 2013). Depression in one parent has been shown to impair the marital relationship, which in turn affected children’s adjustment (Cummings et al., 2005). One parent’s distress also undermines their capacity to function as a coparent in effectively supporting the spouse’s parenting efforts (Feinberg, 2003). Therefore, from a family systems perspective, one family member’s psychopathology will not only impact on the direct relationships between the affected member and the other members, but also the interactions between other family members (Riggs & Riggs, 2011). Notwithstanding, the psychopathology of the respective coparent has not yet been considered as potential risk factor for the other parent’s perpetration of violence against children in refugee families.
The individual and the family microsystem are embedded in the exosystem, which subsumes social structures and circumstances, such as poverty and community support, and the macrosystem, which represent cultural values and belief systems (Belsky, 1980). Low socioeconomic status has been singled out as a potent risk factor for child maltreatment in an umbrella synthesis of meta-analyses (van IJzendoorn et al., 2020). In the case of refugee camps, where formal income-generating activities are often prohibited and most people rely almost exclusively on aid programs, the potential role of socioeconomic status remains unclear.
The present study aimed to identify factors associated with violence by mothers and fathers in refugee families living in refugee camps using an ecological systems framework. In order to overcome important limitations of previous research, the study applied a multi-informant approach considering both children’s and parents’ reports of violence and included characteristics of the coparent as potential associated factors. The sample consisted of Burundian refugee families living in refugee camps in Tanzania close to the Burundian border. Burundi has a history of extreme ethnic and political violence which has spurred multiple waves of refugees. After the most recent eruption of political violence in 2015, over 400,000 Burundians fled to neighboring countries, primarily Tanzania (UNHCR, 2018).
We hypothesized that (1) higher levels of children’s psychopathology (externalizing problems and PTSD symptoms), parental childhood experiences of violence, and parental psychopathology (PTSD and depressive symptoms, psychosocial impairment, alcohol abuse [only for fathers]) would be associated with increased violence by the respective parent according to child- and parent-report. We also hypothesized that (2) higher levels of psychopathology of the respective coparent would be associated with increased use of violence by mothers and fathers in the child- and parent-report.
Methods
Setting and Sampling
From February to May 2018, the study was conducted with Burundian refugee families in three refugee camps situated in the Kigoma Region of Tanzania, close to the border to Burundi (UNHCR, 2018). Families were recruited using a combined systematic and random sampling approach. All three camps are subdivided into 10-12 camp zones. After randomly selecting two zones in each camp, a sampling direction was randomly determined by spinning a pen in the centers of the chosen zones. The family in every sixth house or tent in this direction was invited to be interviewed on the following day. A family was defined as a triad consisting of mother or female caregiver, father or male caregiver and the oldest child in primary school age (7-15 years). This wide age range was chosen in order to be able to generalize across a large group of minors and to take into account the lived realities of a variety of families living in the camps.
Participants
Sociodemographic Data of the Refugee Families.
Note. aCaregivers’ responses refer to the highest level achieved. bInformation is averaged across parents. cThe number of children’s siblings was used as an indicator.
Procedure
Structured clinical interviews were conducted individually with each family member by Tanzanian masters-level psychologists and research assistants from the respective refugee communities within the facilities of cooperating nongovernmental organizations (NGO) within the refugee camps. Research assistants had participated in a one-week training in handling the study instruments, understanding underlying psychological concepts, and adhering to general guidelines for performing clinical interviews.
Upon arrival, the entire family was informed about the procedure and purpose of the study, the anonymization and confidentiality of their data, and their right to withdraw from the study at any time without facing any consequences. Informed consent was obtained by the parents and all children aged at least 11 years by written signature or fingerprint. For children below this age, consent was given by the caregivers on their child’s behalf. All but two families consented to participate in the study. The interviews lasted between one and two hours and were held in Kirundi, the local language, by the local research assistants or in Swahili by the Tanzanian psychologists, who were supported by local interpreters. In order to increase children’s and adolescents’ common understanding and interpretation of interview questions, they were interviewed by psychologists with great experience in research and counselling with youth of different ages. Moreover, the structured clinical interview format allowed interviewers to reformulate and explain questions in an age-appropriate way. Each family received a compensation of 8 USD for the time and efforts they invested to participate in the study. The study protocol was approved by the ethics committee of the University of Zurich (No. 2017.10.2) and the Tanzanian National Institute for Medical Research (No. NIMR/HQ/R.8a/Vol.IX/ 2632). All necessary permits to conduct research in Tanzania and the refugee camps had been granted by the Tanzanian Commission for Science and Technology and the Tanzanian Ministry of Home Affairs. Other aspects of this extensive investigation have been published elsewhere (Scharpf et al., 2019, 2021).
Measures
Study instruments were either available in Kiswahili or translated from English to Kiswahili according to recommended procedures in transcultural research using blind-back translation (Brislin et al., 1973). The applicability of mental health concepts and terms in Burundian culture and appropriate translations into Kirundi were evaluated in focus group discussions. A pilot-assessment of eight families in the first camp and subsequent adjustments further supported the applicability of the measures.
Sociodemographic data.
Sociodemographic information was collected from all participants (e.g., age, gender, and educational level). Both mothers and fathers were also asked about the monthly household income and their reports were averaged.
Current violence against the child (parent and child report).
Children answered the 27-item and parents the short form (10-item) of the Parent-Child Conflict Tactics Scale (CTSPC; Straus et al., 1998) including subscales on emotional (five and four items respectively) and physical violence (thirteen and four items respectively). Emotional violence is conceptualized as “verbal and symbolic acts by the parent that cause psychological pain or fear on the part of the child” (example items: “I swore or cursed at him/her”; “Mother called me dumb or lazy or some other name like that”), while physical violence comprises parental acts of lower severity, i.e. corporal punishment (example item: “I spanked him/her on the bottom with my bare hand”) and higher severity, i.e. physical abuse (example item: “Father beat me up by hitting me over and over as hard as he could”) (Straus et al., 1998). The CTSPC has been successfully used to assess parental violence in East-Africa (Nkuba et al., 2018) and has shown good psychometric properties (Sierau et al., 2018). The frequency of a given act of violence in the past year was coded on a seven-point Likert scale (0 = never, 1 = once, 2 = twice, 4 = 3 to 5 times, 8 = 6 to 10 times, 15 = 11 to 20 times, 25 = more than 20 times). For the analyses, the items assessing a parent’s use of emotional and physical violence against the child were summed up. The internal consistency for the sum score was α = .68 for mothers and α = .69 for fathers. Two sum scores were computed based on the children’s report. Internal consistency was good for the mother (α = .80) and the father score (α = .82).
Parental childhood experiences of violence.
The Short Form of the CTSPC was also used to assess parents’ own adverse experiences of family violence during their childhood, similar to previous use of this instrument in an East-African context (Saile et al., 2014). The items were coded on a six-point Likert scale ranging from 0 = never to 5 = most of the time. Internal consistency of the two sum scores of mothers’ (α = .78) and fathers’ (α = .76) childhood experiences of violence was acceptable.
Children’s psychopathology.
The child-report version of the Strengths and Difficulties Questionnaire (SDQ) was used to screen for children’s externalizing problems. The SDQ is widely used and has been applied in Sub-Saharan Africa (Goodman et al., 2003; Hoosen et al., 2018). Children rate the extent to which the given statements applied to them within the past six months on a three-point Likert scale (not true, somewhat true, certainly true). The hyperactivity (example items: “I finish the work I’m doing. My attention is good”; “I am restless, I cannot stay still for long”) and conduct problems subscales (example items: “I get very angry and often lose my temper”; “I take things that are not mine from home, school or elsewhere”) with five items each can be combined to a score of externalizing problems (range 0-20). The low Cronbach’s alpha of this score (α = .54) can be explained by the heterogeneity of the score consisting of items assessing intrapersonal (hyperactivity) and interpersonal (conduct problems) problems. It is also in line with the average coefficients found in most studies using the self-report version of the SDQ with non-English-speaking populations (Stolk et al., 2017).
Children’s PTSD symptoms were assessed using the University of California at Los Angeles Child/Adolescent PTSD Reaction Index for DSM-5 (UCLA-RI-5; Pynoos & Steinberg, 2015) that comes with good psychometric properties and has been validated also in non-Western countries (Doric et al., 2019; Kaplow et al., 2019). The index consists of 31 items covering the DSM-5 PTSD symptoms. Children reported the amount of time they had experienced a given reaction during the past month on a five-point Likert scale ranging from 0 = none to 4 = most of the time. The sum score of all items had excellent internal consistency (α = .91).
Caregivers’ psychopathology.
The Posttraumatic Stress Disorder Checklist for DSM-5 (Weathers et al., 2013) was chosen to assess parental PTSD symptom severity due to its widespread use and good psychometric properties (Bovin et al., 2016). The PCL-5 has further been utilized in various cultural contexts (e.g., Ibrahim et al., 2018). The items were coded on a five-point Likert scale ranging from 0 = not at all to 4 = extremely, indicating how much the person was bothered by the respective problem in the past month. Scores were summed up over all 20 items and showed an excellent reliability index (α = .94 for mothers and α = .91 for fathers).
With the Brief Symptom Inventory-18 (Derogatis, 2000) we assessed symptoms of depression. The items are rated on a five-point Likert scale ranging from 0 = not at all to 4 = most of the time with regard to how much the respondent has been distressed or bothered by a symptom during the last seven days before the interview. The BSI-18 comes along with good psychometric properties and has been used internationally (Asner-Self et al., 2006). Internal consistency for the six-item subscale of depression was α =.78 for mothers and α = .72 for fathers.
The Alcohol, Smoking and Substance Involvement Screening Test (ASSIST) developed by the World Health Organization was used to assess paternal alcohol abuse as it is highly adaptable to various cultural contexts and records frequency of consumption as well as associated health, social, legal, and financial problems (Humeniuk et al., 2010). The six items referring to lifetime alcohol use and within the past three months were summed. Internal consistency of this sum score was α = .82, including subjects who scored zero on all domains (n = 87).
Six items from the Luo Functioning Scale (Ertl et al., 2014) originally developed and validated in a sample of former Ugandan child soldiers were used to assess caregivers’ impairments in different areas of daily functioning (household chores, working, social relationships, social and private leisure activities, community development activities). Respondents indicated on a five-point Likert scale ranging from 0 = none to 4 = often can`t do tasks the degree of difficulty they experienced in completing the tasks and activities. Difficulties were only considered if they were caused by physical, mental, or emotional problems and not by external factors, such as the lack of financial means (Ertl et al., 2014). Internal consistency of the sum score was good for mothers (α = .87) and fathers (α = .80).
Statistical Analyses
We developed two path models, one for the child-report and one for the parent-report of violence by mothers and fathers (Figures 1 and 2). Each model contained child characteristics (age, gender, externalizing behaviors, and PTSD symptoms), both maternal and paternal characteristics (educational level, biological parenthood, childhood experiences of violence, depressive symptoms, PTSD symptoms, and psychosocial impairment) and the average household income per month as predictors. Paternal alcohol abuse was included as an additional paternal characteristic in both models. Parental educational level and biological parenthood were only considered in the prediction of the respective parent`s use of violence. The path models were analyzed using the package lavaan (Rosseel, 2012) in the statistical environment R (R Core Team, 2019). We used the following criteria to evaluate goodness of model fit: a nonsignificant χ2-value, a comparative fit index (CFI) ≥ 0.95, a nonsignificant root mean square error of approximation (RMSEA) ≤ 0.06, and a standardized root mean squared residual (SRMR) ≤ 0.08 (Hu & Bentler, 1999). Preliminary analyses using IBM SPSS Statistics 25 had indicated that father- and child-reported paternal violence deviated from normality according to guidelines by West et al. (1995). Therefore, maximum likelihood estimation with robust standard errors and Satorra-Bentler scaled test statistic were used to estimate the models. All other statistical assumptions (linearity, collinearity, reliability, and missing value analysis) for using path analysis were met. There were 10 univariate outliers in the predictor variables deviating more than three standard deviations from the mean. Using a robust variant of the Malahanobis distance (Leys et al., 2018), we also identified 24 multivariate outliers in the child-report model and 20 outliers in the parent-report model. After calculating the models without these univariate and multivariate outliers, results remained stable and we retained them in the analyses. We used R to indicate the amount of variance in child- and parent-reported violence by mothers and fathers that could be explained by the predictor variables in the models. Regarding individual associations, our metric for a small effect size was β ≥ .10, for a medium effect β ≥ .30, and for a large effect β ≥ .50 (Shrout & Bolger, 2002). We used a two-tailed α = .05 to evaluate significance of associations in the path analysis, α = .10 was interpreted as a trend.
Path model of child-reported violence by mothers (R2 = 0.188) and fathers (R2 = 0.186), χ2[6] = 7.752, p = .257, CFI = 0.99, RMSEA = 0.036 (p = .562), SRMR = 0.03. Only associations that are significant and/or of at least a small effect (≥ β = .10) are shown. Straight lines indicate positive and dashed lines negative associations. ****p < 0.001, ***p < 0.01, **p < 0.05, *p < 0.10.
Path model of parent-reported violence by mothers (R2 = 0.170) and fathers (R2 = 0.184), χ2[6] = 7.789, p = .254, CFI = 0.97, RMSEA = 0.036 (p = .535), SRMR = 0.09. Only associations that are significant and/or of at least a small effect (≥ β = .10) are shown. Straight lines indicate positive and dashed lines negative associations. ****p < 0.001, ***p < 0.01, **p < 0.05, *p < 0.10.
Results
Descriptive statistics of the study variables are displayed in Table 2 and bivariate correlations between all model variables in Appendices A and B.
Means, Standard Deviations, and Scale Ranges of Sum Scores of Reported Maternal and Paternal Violence and Respective Risk Factors.
Note. CTSPC = Parent-Child Conflict-Tactics Scale; UCLA-RI = University of California at Los Angeles Child/Adolescent PTSD Reaction Index for DSM-5; SDQ = Strengths and Difficulties Questionnaire; PCL-5 = Post-traumatic Stress Disorder Checklist for DSM-5; BSI-18 = Brief Symptom Inventory-18; LFS = Luo Functioning Scale; ASSIST = Alcohol, Smoking and Substance Involvement Screening Test.
Child-Report of Maternal and Paternal Violence
The model predicting child-reported violence by mothers and fathers in the past year showed a good fit to the data, χ2 [6] = 7.752, p = .257, CFI = 0.99, RMSEA = 0.036 (p = .562), SRMR = 0.013. Higher levels of children’s self-reported externalizing problems and PTSD symptoms, biological motherhood and higher levels of fathers’ PTSD symptoms were significantly associated with higher levels of child-reported physical and emotional violence by mothers, while higher levels of fathers’ depressive symptoms were significantly associated with lower levels of maternal violence (Figure 1). All predictors accounted for 19% of the variance in maternal violence.
Higher levels of children’s self-reported externalizing problems and PTSD symptoms, paternal PTSD symptoms, maternal psychosocial impairment, and biological fatherhood were significantly related to higher levels of child-reported physical and emotional violence by fathers. Being a girl, a higher paternal educational level, and higher levels of paternal depressive symptoms and psychosocial impairment were significantly associated with lower levels of paternal violence. All predictors explained about 19% of the variance in paternal violence. Associations within the model that were significant, equivalent to at least a small effect, or both are graphically displayed in Figure 1. Parameter estimates and significance levels for all associations are shown in Appendix C.
Parent-Report of Maternal and Paternal Violence
The model predicting parent-reported violence by mothers and fathers in the past year showed a good fit to the data, χ2 [6] = 7.789, p = .254, CFI = 0.97, RMSEA = 0.036 (p = .535), SRMR = 0.009. Higher levels of children`s self-reported externalizing problems, maternal childhood experiences of violence, and biological motherhood were significantly associated with higher levels of mother-reported physical and emotional violence in the past year. Children’s older age, a higher monthly household income, and higher levels of paternal psychosocial impairment were significantly related to lower levels of maternal violence (Figure 2). All predictors accounted for 17% of the variance in mother-reported violence.
Higher levels of paternal childhood experiences of violence and alcohol abuse as well as biological fatherhood were significantly associated with higher levels of father-reported physical and emotional violence in the past year. The association between higher maternal psychosocial impairment and increased paternal violence failed to reach significance, but equaled a small effect. Being a girl and higher levels of children’s PTSD symptoms were significantly related to lower levels of paternal violence. All predictors explained about 18% of the variance in father-reported violence. Associations within the model that were significant, equivalent to at least a small effect, or both are graphically displayed in Figure 2. Parameter estimates and significance levels for all associations are shown in Appendix D.
Discussion
The aim of the present study was to identify factors associated with child- and parent-reported violence by mothers and fathers in refugee families living in refugee camps using an ecological systems framework. We received partial support for our first hypothesis that characteristics of children and the respective parent would be associated with parental violence: Higher levels of children’s externalizing problems were associated with higher levels of maternal violence as reported by both children and mothers. This is in line with previous findings that have demonstrated bidirectional associations between children’s externalizing problems and parents’ use of physical violence (MacKenzie et al., 2014; Sriskandarajah et al., 2015; Wang & Liu, 2018). The validity of this finding is strengthened by our multi-informant approach. Furthermore, higher levels of children’s PTSD symptoms were associated with higher levels of maternal and paternal violence, but only in the child-report. Though children’s PTSD symptoms have not been investigated as potential risk factors, our findings are not surprising considering that specific PTSD symptoms, such as irritability, anger and concentration problems, may elicit violent parenting behavior (Catani, 2018; Timshel et al., 2017).
Maternal and paternal childhood experiences of violence were only associated with their respective self-reported use of violence. This partially supports the “cycle of violence” hypothesis (Widom, 1989; Widom et al., 2008). However, the fact that these associations were not found in the child-report is concordant with Thornberry et al.’s (2012) conclusion that the support of the “cycle of violence” hypothesis is mainly based on single-reporter studies. While Saile et al. (2014) found that maternal childhood victimization was linked to children’s self-reported experiences of violence in postconflict Uganda, their outcome measure did not specify the perpetrating parent and assessed family violence in general, not only parent-to-child violence.
Further, we did not observe associations between maternal psychopathology and their use of violence as reported by themselves and their children, which is consistent with previous studies (Berlin et al., 2011; Saile et al., 2014). Although paternal psychopathology appeared to be more relevant in explaining their use of violence compared to maternal psychopathology, different types of psychopathology were associated with child- (PTSD symptoms) and father-reported violence (alcohol use), respectively. The former finding adds to the growing body of evidence documenting an increased tendency towards violent parenting among war-exposed fathers with PTSD (Montgomery et al., 2019; Timshel et al., 2017). Establishing this association for child-reported violence may be considered as more informative compared to fathers’ self-report, which may be biased by PTSD symptoms related to negative self-evaluations (Banyard et al., 2003).The latter finding, in turn, is in line with studies from postconflict settings showing consistent links between fathers’ alcohol abuse and their self-reported use of violence against children (Catani et al., 2008b; Saile et al., 2014; Sriskandarajah et al., 2015).
In the child report, higher levels of paternal psychosocial impairment and depressive symptoms were related to lower levels of violence, counter to our hypothesis. Though this is surprising at first glance and not concordant with previous findings (Wilson & Durbin, 2010), one potential explanation is that highly depressed fathers were too affected by their symptoms to engage in violent discipline. Our finding is also in line with a previous study showing that parental depression was associated with less perpetration (Pears & Capaldi, 2001). Moreover, the finding that both psychosocial impairment and depression follow the same pattern in their association with violence is consistent with the prominence of functional impairment in depression (Hammer-Helmich et al., 2018).
Our findings partially supported our second hypothesis that one parent`s distress undermines their capacity of functioning as a coparent: A consistent finding across child and parent reports was that higher maternal psychosocial impairment was related to higher levels of paternal violence. This may be because fathers experience increased stress and frustration when the mother struggles to fulfill her tasks and activities in daily life, forcing the father to adopt the mother’s traditional role as the primary agent in children’s socialization and upbringing (Song et al., 2014). Regarding maternal violence, higher levels of paternal PTSD symptoms were related to higher levels of violence, but only in the child-report. From a family systems perspective, the distress of one member will affect all other members and how they interact with each other (Riggs & Riggs, 2011). Living with a husband suffering from PTSD may especially challenge mothers’ own capacities to provide supportive and nonviolent parenting (Blow et al., 2013). Surprisingly, higher levels of paternal psychosocial impairment and depressive symptoms were associated with lower levels of mother-reported violence. While this pattern was also observed for child-reported paternal violence, it contrasts the finding obtained for maternal psychosocial impairment. A possible explanation is that when fathers are impaired in their role as household head and breadwinner, the mother may be forced to overtake this responsibility and thus may simply lack time and energy to discipline their children.
Families’ lower monthly household income was associated with a higher self-reported use of violence by mothers. Although associations between household income and child-reported violence by both parents pointed into a similar direction, these were only significant in bivariate analyses. Notwithstanding, these findings suggest that the economic hardship of camp life, which is often exacerbated by host governments’ restrictive policies regarding work opportunities for refugees, may increase children’s risk of experiencing parental violence. Low SES has been established as a strong risk factor for child maltreatment in high-income countries (Stith et al., 2009; van IJzendoorn et al., 2020), but our finding indicates that this may also be the case in low-resource settings with widespread poverty (Bermudez et al., 2018; Sim et al., 2018).
Implications for Future Research
The findings of the present study have important implications for future studies: first, it is crucial to take into account both children’s and parents’ perspectives when assessing violence. On the one hand, this can strengthen the validity of certain risk factors, such as children’s externalizing problems and maternal psychosocial impairment. On the other hand, it can help to carefully examine the role of factors that have been particularly emphasized in previous research, such as parental childhood experiences. Future studies should therefore strive to include child-report and observational data indicating potential for child abuse, which may help to overcome biases associated with distressed parents’ self-reports. Second, it is of utmost importance to consider characteristics of all family members, particularly of the coparent, when trying to understand the conditions that give rise to family violence against children. However, more studies in future are needed to provide a clearer picture of how the psychopathology of one parent may be related to the other parent’s perpetration of violence against children (Feinberg, 2003). At the same time, the results seem to suggest a more nuanced picture and that it may not be so much the coparent’s psychopathology per se, but the resulting psychosocial impairment that affects the whole family unit. Consequently, future studies should aim to elucidate the potentially mediating role of shifting family roles and processes. Third, future research using prospective and longitudinal study designs is needed to investigate underlying mechanisms and directionality of risk factors on the family and community level (Catani, 2018).
Clinical Implications
Our results indicate that characteristics of both parents (fathers’ PTSD symptoms and alcohol abuse, mothers’ psychosocial impairment, mothers’ and fathers’ childhood victimization) and children (externalizing problems and PTSD symptoms) may increase the risk of parent-to-child violence in refugee camps. This suggests that interventions aiming to reduce children’s exposure to parental violence should target the whole family unit. Although engaging fathers may pose a challenge in this specific cultural context (Doyle et al., 2014), their PTSD symptoms and drinking behavior are prime targets for interventions. Moreover, children’s traumatization and externalizing problems need to be addressed. Offering psychoeducation to parents about their children’s and their own psychopathology as well as teaching them nonviolent parenting strategies as alternatives to what they experienced during their own childhood may also be important elements of prevention and intervention (Jordans et al., 2013). In addition, psychosocial support for mothers in areas of daily functioning, such as childcare, creating livelihoods, and social aspects, may be promising ways to ease the burden on families. For instance, the Problem Management Plus Intervention (PM+; Dawson et al., 2015) poses an appropriate and evidence-based psychological intervention that can easily be implemented in settings with scarce resources, such as refugee camps. PM+ was designed specifically for people facing adversity in their day-to-day life and has the aim of reducing both psychological and interpersonal problems, such as family conflicts.
Diversity and Limitations
Apart from the general criterion that the families that participated in our study had to consist of a female and a male caregiver and a child in primary school age (between 7 and 15 years), we did not apply any inclusion or exclusion criteria in terms of caregivers’ age, participants’ ethnicity, religion, socioeconomic status, or (dis)ability. As such, we consider our sample as highly diverse, comprising different ethnic (e.g., Hutu, Tutsi, Mtwa), religious (e.g., Muslims, Christians, traditional religions), and age groups (i.e., age range for caregivers between 19 and 80 years). Factors such as language and nationality were predetermined by the fact that two of the three refugee camps in which we conducted the study are inhabited exclusively by refugees from Burundi whose only native language is Kirundi. Yet, we did not enquire about factors such as sexual orientation and gender identity.
Some limitations should be considered when interpreting the results of the current study. Due to the cross-sectional study design, we could not establish causality of the factors. The systematic sampling approach ensured the representativity of our sample of Burundian refugee families living in refugee camps. However, further generalizability of our findings is limited. In addition, self-report biases, such as social desirability or recall bias, cannot be completely ruled out. As we focused on risk factors within the proximal family microsystem and strived for parsimonious models, we did not consider other potentially relevant exosystem and macrosystem factors, such as community social support and general attitudes towards violence, respectively. However, the moderate amount of variance in parental violence explained by our models suggests that such factors may also be relevant. Future studies should aim to assess contextual factors in an objective way to avoid subjective bias and investigate how the impact of these factors on the occurrence of violence against children is mediated by microsystem variables. Moreover, we did not investigate mechanisms potentially underlying the observed associations.
Conclusion
Besides facing many adversities of camp life, children growing up in refugee camps are also at risk of experiencing physical and emotional violence by their parents, which may further jeopardize their well-being and development. In line with the ecological integration model (Belsky, 1980), our findings point to the importance of ontogenic and microsystem factors in explaining parental violence against children in families living in refugee camps. In particular, the findings suggest that paternal and children’s psychopathology, maternal psychosocial difficulties and parental childhood experiences of violence may be important risk factors for children’s victimization in these settings. In order to adequately assess children’s risk and plan prevention and intervention approaches, multiple perspectives and the whole family unit should be considered.
Appendices
Bivariate Correlations Between all Predictor and Dependent Variables Assessed With Pearson's Correlation Coefficient (r) and Spearman's Rho (ρ).
Note. *p < .05. **p < .01. Bivariate correlations were assessed with Pearson’s correlation coefficient. aAssessed with Spearman's Rho.
Bivariate Correlations Between all Sociodemographic Variables and Dependent Variables Assessed With Pearson's Correlation Coefficient (r) and Spearman’s Rho (ρ).
Note. *p < .05. **p < .01. Bivariate correlations were assessed with Pearson’s correlation coefficient. aAssessed with Spearman’s Rho.
Unstandardized and Standardized Parameter Estimates and Significance Levels for all Associations in the Child-Report Model of Maternal and Paternal Violence.
Unstandardized and Standardized Parameter Estimates and Significance Levels for all Associations in the Parent-Report Model of Maternal and Paternal Violence.
Footnotes
Acknowledgments
Special thanks go to Plan International Tanzania and International Rescue Committee Tanzania for providing space and resources for data collection. We are extremely grateful to our highly motivated assistants, to Getrude Mkinga and Faustine Bwire Masath, Bielefeld University, to Mabula Nkuba and Maregesi Machumu, Dar es Salaam University College of Education, and to Andreas Maercker and Markus Landolt, Department of Psychology, University of Zurich, for their continuous support.
Data Availability Statement
The study protocol, all assessment materials, and the datasets are available from the corresponding author on request.
Declaration of Conflicting Interests
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by the North-South Cooperation at University of Zurich [F-63212-13-01].
