Impact statement
Refugee children globally experience a range of risk factors for poor mental health and psychosocial well-being. Aspects of their relationships with caregivers, including caregiver warmth, parenting self-efficacy and positive relationships between caregivers and children, may help to reduce refugee children’s mental health challenges. We conducted a survey of caregivers and children in the Imvepi refugee settlement in Uganda, exploring whether aspects of the caregiver-child relationship were associated with lower levels of child depression. When including economic stress in the statistical models, parental warmth and child-caregiver interaction are not associated with lower child depression. This suggests that economic factors, such as poverty and livelihood opportunities are highly influential on children’s well-being and may block the important impact of positive parenting in situations of chronic stress.
Introduction
The current scale of displacement of children globally is unprecedented. The United Nations High Commission for Refugees (UNHCR) reports that, of the 108.4 million forcibly displaced persons in 2023, 43.3 million were children (UNHCR, 2023). Between 2018 and 2022, an average of 385,000 children were born as refugees annually (UNHCR, 2023). More than three-quarters of refugees globally are hosted in low and middle-income country (LMIC) settings, often in refugee camps. Refugee camps generate a range of potential threats to refugee children’s physical and mental well-being (Vossoughi et al., Reference Vossoughi, Jackson, Gusler and Stone2016). Children in refugee camps have often experienced or witnessed traumatic events during flight, and subsequent changes in family structure, social support and educational opportunities in refugee camps in LMICs can also produce considerable stress and poor mental health. A systematic review of risk and protective factors amongst refugee children in LMICs found that exposure to premigration violence, female sex (for internalizing and emotional problems), male sex (for externalizing and behavioral problems), settlement in a refugee camp, and internal displacement were identified as risk factors for poor child mental health (Reed et al., Reference Reed, Fazel, Jones, Panter-Brick and Stein2012). A more recent systematic review indicates that evidence of risk factors for poor mental health in LMICs remains sparse, but that exposure to war-related trauma, parental mental health problems and poor parenting are risk factors for the poor mental health of refugee children (Scharpf et al., Reference Scharpf, Kaltenbach, Nickerson and Hecker2021). Empirical evidence indicates that the mental health of refugee children is inextricably connected to family functioning and well-being, which is directly impacted by displacement-related stressors in specific realms of family life (Arenliu et al., Reference Arenliu, Bertelsen, Saad, Abdulaziz and Weine2020), including changes in family roles and obligations and changes in family relationships with other family members (Weine et al., Reference Weine, Muzurovic, Kulauzovic, Besic, Lezic, Mujagic, Muzurovic, Spahovic, Feetham, Ware, Knafl and Pavkovic2004).
Research has identified several protective factors that may be protective against children’s poor mental health, even in contexts of ongoing stressors (Betancourt and Khan, Reference Betancourt and Khan2008). Family-level dynamics, including strong caregiver-child interactions and attachment relationships, and warm and supportive parenting practices, have been found to be associated with improved mental health among refugee children affected by conflict and violence (Catani et al., Reference Catani, Schauer, Elbert, Missmahl, Bette and Neuner2009; Fayyad et al., Reference Fayyad, Cordahi-Tabet, Yeretzian, Salamoun, Najm and Karam2017). Parenting practices have been identified as playing “a crucial role in children’s psychological well-being in a war context” (Catani, Reference Catani2018). A systematic review of resilience and mental health of children living in LMIC settings affected by armed conflict identified several studies that indicated family-level predictors, such as parental monitoring and support, as protective against poor mental health outcomes or promotive of positive psychosocial wellbeing (Tol et al., Reference Tol, Song and Jordans2013). The role of “engaged, responsive and stable parenting” in influencing children’s mental health in contexts characterized by extreme violence, instability and deprivation is increasingly the focus of research and intervention in humanitarian settings (El-Khani et al., Reference El-Khani, Haar, Stojanovic and Maalouf2021).
Some evidence in humanitarian contexts in LMICs on how caregiver mental health influences parenting practices exists, while evidence concerning linkages between parenting practices and children’s mental health outcomes is even more sparse. Evidence focusing on caregiver mental health and parenting practices indicates that poor mental health amongst caregivers is associated with significantly poorer parenting practices. For example, Meinhart and colleagues’ study in Uganda found that lower caregiver distress was significantly associated with lower use of parenting behaviors characterized by undifferentiated rejection (Meinhart et al., Reference Meinhart, Mangen, Hermosilla, Cohen, Agaba, Kajungu, Knox, Obalim and Stark2023). The majority of existing evidence on associations between parenting practices and child mental health in refugee contexts derives from intervention studies. A series of interventions to improve refugee caregiver mental health – including, for example, the Caregiver Support Intervention – emphasizes the role of “compromised parenting” in causing psychosocial distress and poor mental health amongst children in refugee contexts (Miller et al., Reference Miller, Chen, Koppenol-Gonzalez, Bakolis, Arnous, Tossyeh, El Hassan, Saleh, Saade, Nahas, Abboud, Jawad and Jordans2023; Miller et al., Reference Miller, Koppenol-Gonzalez, Arnous, Tossyeh, Chen, Nahas and Jordans2020). Miller and colleagues indicate that caregivers’ behaviors and practices can protect against the deleterious impacts of violence that can characterize children’s experiences in refugee contexts, “by providing a supportive and structured home environment to allay the fear, sadness, and uncertainty engendered by violence and displacement” (Miller et al., Reference Miller, Koppenol-Gonzalez, Arnous, Tossyeh, Chen, Nahas and Jordans2020). Recent evaluations of Strong Families, a parenting skills intervention implemented in several humanitarian contexts globally, indicate that an intervention focused on improving caregiver self-regulation, relationship with and responses to children, and parenting self-efficacy, can improve children’s psychosocial well-being (El-Khani et al., Reference El-Khani, Haar, Stojanovic and Maalouf2021; Haar et al., Reference Haar, El-Khani, Molgaard, Maalouf, Stanikzai, Nader, Molgaard, Calam, Allen, Coombes, Gerra, Campello, Montero Salas and Nazir2020; Haar et al., Reference Haar, El-Khani, Mostashari, Hafezi, Malek and Maalouf2021). Parenting practices have also been identified as a moderator of associations between the experience of traumatic events and children’s mental health, with some studies indicating that positive parenting practices can ameliorate the impact of traumatic events on children’s well-being (Slone and Shoshani, Reference Slone and Shoshani2017). In addition, evidence on associations between caregiver mental health and child mental health in humanitarian contexts indicates a potential pathway between poor caregiver mental health, compromised parenting practices, and adverse child mental health outcomes (Betancourt et al., Reference Betancourt, Abdi, Ito, Lilienthal, Agalab and Ellis2015). Overall, there is limited evidence from humanitarian contexts in LMICs, that links parenting practices and behaviors to children’s mental health outcomes.
Broader literature concerning the relationship between caregiver behaviors, relationships and beliefs – referred to in this study as caregiver characteristics and behaviors – has provided definitions of key concepts, as well as evidence connecting these factors to children’s mental health outcomes. Rohner and Khaleque, who developed the widely utilized Parental Acceptance and Rejection Questionnaire, propose that parental warmth can be defined as the "quality of the affectional bond between parents and their children…and the physical, verbal and symbolic behaviors parents use to express these feelings” (Rohner and Khaleque, Reference Rohner and Khaleque2012). In a meta-analysis of children’s perceptions of parental warmth and affection, and associations with their psychological adjustment, Khaleque found significant associations between parental warmth and children’s psychological outcomes (Khaleque, Reference Khaleque2013). Yap and Jorm’s systematic review of parental factors associated with child depression, anxiety and internalizing behaviors found that lower levels of parental warmth were associated with higher internalizing outcomes (Yap and Jorm, Reference Yap and Jorm2015), and evidence also indicates that experience of parental warmth and affection in childhood predicts coping and well-being in adulthood (Moran et al., Reference Moran, Turiano and Gentzler2018). Parental self-efficacy (PSE) describes “a parent’s belief in their ability to perform the parenting role successfully,” (Wittkowski et al., Reference Wittkowski, Garrett, Calam and Weisberg2017) and has been consistently found to be associated with improved parental mental health, higher quality child-caregiver relationships, and positive mental health amongst children (Albanese et al., Reference Albanese, Russo and Geller2019). Jones and Prinz identify a range of potential pathways between PSE and child mental health, including that PSE impacts parental competence, and that parents with low PSE may find effective parenting more difficult in challenging contexts (Jones and Prinz, Reference Jones and Prinz2005). Some evidence has indicated that socio-economic factors, including income and household size, are significantly associated with PSE; structural factors may reduce caregivers’ capacity for effective parenting and, subsequently, reduce PSE (Fang et al., Reference Fang, Boelens, Windhorst, Raat and van Grieken2021). Child-caregiver interactions form the basis of responsive caregiving, providing children with a strong attachment to primary caregivers, and supporting healthy social, physical and emotional development (World Health Organization, 2004).
Uganda provides a unique and important context in which to investigate associations between caregiver characteristics and behaviors and child mental health. Uganda is the largest host country of refugees in Africa, with the majority of the 1.5 million refugees residing in Uganda originating from South Sudan and the Democratic Republic of Congo [DRC] and the vast majority residing in rural refugee settlements, primarily in North-West Uganda (UNHCR, 2023). Evidence from refugee settlements in Uganda indicates that refugee youth face a confluence of factors producing significant risks to mental health and psychosocial wellbeing, including sexual and gender-based violence, poverty, early and forced marriage, and lack of access to quality health services and educational opportunities (Logie et al., Reference Logie, Okumu, Latif, Musoke, Odong Lukone, Mwima and Kyambadde2021; Logie et al., Reference Logie, Okumu, Loutet, Berry, Taing, Lukone, Kisubi, Sokolovic and Kyambadde2023; Loutet et al., Reference Loutet, Logie, Okumu, Berry, Lukone, Kisubi, McAlpine, Mwima and Kyambadde2022). A limited number of studies have explored risk and protective factors for adverse mental health outcomes amongst child refugees in Uganda, including anxiety and depression. A study conducted amongst female adolescent refugees in urban settlements in Uganda showed that having children during adolescence was a risk factor for depressive symptoms (Malama et al., Reference Malama, Logie, Okumu, Hakiza, Mwima and Kyambadde2023). Evidence from refugee settlements in Uganda indicates that adolescents who reported higher levels of violence experience was associated with elevated depression and anxiety (S. R. Meyer et al., Reference Meyer, Steinhaus, Bangirana, Onyango-Mangen and Stark2017a). In another analysis, a one-unit increase in caregiver depression, assessed using the Hopkins Symptoms Checklist, was associated with triple the odds of high levels of adolescent symptoms of depression (Sarah R Meyer et al., Reference Meyer, Steinhaus, Bangirana, Onyango-Mangen and Stark2017a).
To expand the evidence base concerning caregiver characteristics and behaviors in refugee contexts, and specifically amongst refugees from the DRC and South Sudan in Uganda, we conducted a survey to address the research question: are caregiver-level factors (parental warmth and affection, parental self-efficacy and child-caregiver interaction) protective for symptoms of child depression? In addition, drawing on self-report measures of child depression from children and caregiver-reported child depression, we assessed whether these factors are consistently associated with child depression across both child and caregiver-report of child depression.
Methods
PlayMatters is a multi-year initiative implemented by a consortium, led by the International Rescue Committee and including WarChild. The overall objectives of PlayMatters are to assess the impact of learning through play approaches, implemented within educational settings, on children’s learning and well-being, in refugee contexts in Uganda, Tanzania and Ethiopia. The present study includes quantitative data collected as part of a formative study amongst refugees from South Sudan and the DRC in the Imvepi refugee settlement, West Nile, Uganda. The formative research focused on one aspect of the larger PlayMatters project, namely, child-caregiver relationships, to identify how household- and caregiver factors influence child well-being in this context. This survey was implemented following a qualitative phase of research, which included in-depth interviews, focus group discussions, household observation and a Photovoice activity. The qualitative research phase aimed to yield insight into child and caregiver perceptions of play, child-caregiver interactions and barriers to playful interactions at home. The qualitative research informed instrument development, including the selection and development of scales specifically for this survey.
Sample and sampling
A total of 501 child-caregiver dyads were selected to participate in the survey. Inclusion criteria for a household were: a child aged between 10–14 years old residing in the household and that household members are refugees from either the DRC or South Sudan (or born in Uganda, but originating from the DRC or South Sudan). The identification of eligible households for the study followed a systematic approach. Using a random skipping pattern, each Research Assistant moved in either direction of the village, and every sixth household was screened for eligibility and inclusion. Each Research Assistant was assigned a demarcated block area to ensure comprehensive coverage, guided by community leaders who delineated block boundaries. As per the household recruitment SOP, households with a child aged 10–14 years old were identified. In households with more than one eligible child, the child was randomly selected by drawing numbers out of a hat. The child then identified the adult in the household whom they considered their primary caregiver. Participation in the study implied verbal consent from the head of the household to allow interaction with household members during the screening process by the research team, assent from the index child, and informed consent from the identified primary caregiver. A total of 501 households were included in the study: 401 households for the South Sudanese and 100 for the Congolese community, with 1002 interviews conducted; one primary caregiver and one child per household. In each household, if more than one eligible child aged 10–14 resided, one child was randomly selected.
Data collection
Data collectors participated in a five-day intensive training focused on equipping them with knowledge and skills relevant to implementing this study. The training agenda covered different sessions, which research ethics, interview skills, and principles for survey data collection and techniques; and then role-play to build an understanding of the survey instruments and practice essential Commcare skills for survey administration.
Before pilot data collection, cognitive testing of the study tools was conducted through ten child interviews and nine caregiver interviews The survey was translated following a multi-step translation process that consisted of (1) translation; (2) blind back-translation into the relevant languages: Alur, Bari, Juba Arabic and Congolese Swahili; (3) verification of the translations by language experts coming from the communities where the study was being implemented. Before pilot data collection, cognitive testing of the study tools was conducted following a three-day training for four enumerators, each assigned to a specific study language. Cognitive interviews were conducted in 10 households, and data were analyzed to refine and improve the caregiver and child survey instruments.
Pilot data collection was done in selected blocks, and a total of 100 households were recruited 100 caregivers and 100 children were interviewed. The survey was translated following a multi-step translation process that consisted of (1) translation; (2) blind back-translation into the relevant languages: Alur, Bari, Juba Arabic and Congolese Swahili and (3) verification of the translations by language experts coming from the communities where the study was being implemented.
Survey instrument
The selected measurement tool to assess the outcome, child depression, was assessed using both child and caregiver-report of the Moods and Feelings Questionnaire – Short Form (MFQ-SF). The MFQ-SF consists of 13 items assessing depressive symptoms, asking the child or caregiver respondent the extent to which certain statements were true in the past two weeks, with response options of “Not true,” “Sometimes true,” and “True,” scored as 0, 1 and 2 respectively. Possible scores range from 0 to 26 (Cronbach’s alpha of child report from this dataset = 0.83; Cronbach’s alpha to caregiver report from this dataset = 0.85). Child depressive symptoms were defined as a binary variable: no depression if the MFQ-SF score was between 0 and 11 and the presence of depressive symptoms if the score was above or equal to 12. The same cut-off points were used for the child’s and caregiver’s reports. This cut-off was selected based on existing literature and recommendations (Duke Psychiatry and Behavioral Sciences, 2024), however, the instrument has not been validated in Uganda or with this population. Conditions required to be able to do linear regression were not met, logistic regression analysis was the best fit for the analysis, with the caveat that a full validity study prior to the selection of a cut-off would strengthen the analysis and interpretation.
Exposure variables were parental self-efficacy (reported only by caregivers), parental warmth and acceptance and child-caregiver interactions both reported by children and caregivers. Parental warmth and acceptance were assessed via a sub-scale of the Parental Acceptance and Rejection Questionnaire [PARQ], which has been validated in many non-Western contexts globally (Khaleque and Rohner, Reference Khaleque and Rohner2002) and has been utilized in a similar context amongst caregivers in a refugee settlement in Western Uganda (Meinhart et al., Reference Meinhart, Mangen, Hermosilla, Cohen, Agaba, Kajungu, Knox, Obalim and Stark2023). The PARQ assesses children’s perceptions of parental warmth and acceptance using eight items such as “My caregiver says nice things about me” or “My caregiver is really interested in what I do,” with response categories of “Almost always true,” “Sometimes true,” “Rarely true” and “Almost never true.” Caregiver-report of their own parental warmth and affection towards the selected child included the same items worded to reflect the caregivers’ behaviors, such as “I say nice things about my child” and “I am really interested in what my child does,” with the same response categories (child-report Cronbach’s alpha from this dataset = 0.90; caregiver-report Cronbach’s alpha from this dataset = 0.74). Parenting self-efficacy was assessed via self-reporting from caregivers, using the efficacy sub-scale of the Parenting Sense of Competence scale, which has been validated in Uganda amongst HIV-positive mothers (Augustinavicius et al., Reference Augustinavicius, Murray, Familiar-Lopez, Boivin, Mutebe, Arima and Bass2020). The efficacy sub-scale consists of seven items, including statements such as “I feel I am able to take care of my child in the way that I feel is a good way to raise him/her” and “You honestly believe you have all the skills necessary to be a good caregiver to your child,” with six possible response options, “Strongly disagree,” “Disagree,” “Somewhat disagree,” “Somewhat agree,” “Agree” and “Strongly agree” (Cronbach’s alpha from this dataset = 0.81). The child-caregiver interactions scale was developed for this study, based on qualitative data collected in an earlier phase of the formative study. The scale included items assessing the frequency of child-caregiver interactions, reported by both children and caregivers and including items assessing the frequency of activities conducted in the past week, such as “My child and I work together on preparing food for our household/family” and “I join my child in games and playful activities,” with possible response options of “Very rarely (less than once a week),” “Rarely (a couple of times a week),” “Occasionally (at least four times a week),” “Frequently (at least once a day),” and “Very frequently (several times a day)” (child-report Cronbach’s alpha from this dataset = 0.78; caregiver-report Cronbach’s alpha from this dataset = 0.81). This instrument was developed for the purposes of this study, however, validation of the instrument beyond the calculation of Cronbach’s alpha was not possible. All items from the PARQ warmth and acceptance sub-scale, the parenting self-efficacy measure and the child-caregiver interactions scale are listed in Supplementary File 1.
Household and individual-level covariates were assessed. Caregivers were asked to report on three household-level measures of financial stress and well-being. Financial stress was assessed using four items – being worried about their general financial situation, having to borrow money for food or rent, not seeing a doctor due to not having enough money, and children missing school because of lack of money for school supplies or uniforms – and asking about frequency of occurrence in the past month. Hunger was assessed using the Household Hunger Scale, which is widely used in LMICs, including humanitarian contexts, consisting of three items and three additional questions about frequency (Deitchler et al., Reference Deitchler, Ballard, Swindale and Coates2011). Broader humanitarian needs were assessed using 12 selected items from the Humanitarian Emergency Settings Perceived Needs (HESPER) Scale, designed to evaluate the physical, social and psychological needs of adults affected by humanitarian crises (Semrau et al., Reference Semrau, van Ommeren, Blagescu, Griekspoor, Howard, Jordans, Lempp, Marini, Pedersen, Pilotte, Slade and Thornicroft2012). The measure includes items such as “Do you have a serious problem because you do not have enough, or good enough, clothes, shoes, bedding or blankets?” and “Do you have a serious problem because you do not have enough water that is safe for drinking or cooking?” and asks respondents to indicate if the specific problem is a serious problem or not a serious problem for them. Caregivers also reported individual-level socio-demographic variables – age, education level, number of years living in the refugee settlement, number of people and children in the household, and relationship with the child respondent within the child-caregiver dyad. Children reported individual-level socio-demographic variables – age, level of education, and physical limitations or difficulties with participation, which was assessed with a single binary question asking, “Do you have any physical limitations or difficulties that limit participation in your family or community?”
Data analysis
Data was transferred from Excel to STATA version 17 for analysis. Descriptive statistics were performed comparing depressive symptoms versus all explanatory variables. Frequency and proportions were calculated for categorical variables and assessed using the chi-square test. Binary logistic regression models were fitted to determine if the defined caregiver-level factors (parental warmth and affection, PSE and child-caregiver interaction) protect child depression symptoms. Independent models were fitted for child and caregiver reports of child depression. Due to the collinearity of the caregiver factors, five adjusted logistic models were fitted: three for child report on child depression and two for child depression-caregiver report AND adjusted for other potential confounders that showed significant association in the univariate models. A p-value of <0.05 was used to assess the strength of the association. The goodness of fit for all the models was assessed using the Hosmer-Lemshow test and exploration of ROC curves.
Ethics
Informed consent and assent procedures for caregivers and their children were followed at the moment of recruitment. Participation was voluntary and participants could withdraw from the study at any time. All results are reported anonymously and are treated with confidentiality. Processes for adverse events were part of the study implementation and participants were informed about available complaints mechanisms.
The study received Institutional Review Board (IRB) approval from the Makerere University School of Health Sciences Research and Ethics Committee (MAKSHS-REC), under reference number MAKSHSREC 2023-620. Additionally, approval was granted by the Uganda National Council for Science and Technology (UNCST) under reference number SS1272ES.
Results
Table 1 displays the socio-demographic characteristics of the 501 caregiver-child dyads included in the study. The vast majority of caregivers were female (80.8%) and aged 25–34 years old (32.7%) or 35–44 years old (30.1%). The most commonly reported level of education was any primary education (61.9%), followed by no formal education (19.0%). Half of the caregivers, 259 (51.7%) reported having lived in the settlement for more than five years. A little more than half the sample reported serious financial difficulties (53.1%), and almost three-quarters of the study population (74.1%) were categorized as facing moderate hunger, while 14.4% were categorized as reporting no hunger and 11.6% as reporting severe hunger. Over half the dyads reported that they were a biological parent of the child 261 (52.1%). There was an equal distribution of sex among children enrolled in the study, 50.3% female, with around a quarter (24.2%) aged 14. An even distribution of education level was observed for the children; the lowest was in P1 (17.8%), with their highest level of education 30.3% of children reported their highest level of education as P4 and above.
Table 1. Child and caregiver socio-demographics
Table 2 displays bivariate associations between caregiver characteristics, socio-demographics and child-reported depression. We observed a significant association between child-reported depression with child-report of the PARQ, parental report of child-caregiver interactions, parenting self-efficacy, years lived in the settlement, number of people in the household, number of children in the household, child-reported limitations and child-caregiver relationship. Table 3 displays bivariate associations between caregiver characteristics, socio-demographics, and caregiver-reported child depression, showing significant associations with child and caregiver-report of the PARQ, financial difficulties and HESPER.
Table 2. Bivariate associations between caregiver factors (PARQ, parenting self-efficacy and child-caregiver interactions), caregiver socio-demographic factors and child-reported child depression
Table 3. Bivariate associations between caregiver factors (PARQ, parenting self-efficacy and child-caregiver interactions), caregiver socio-demographics and caregiver-reported child depression
Table 4 displays multivariate associations between caregiver behaviors and characteristics significant at the bivariate level and child-reported depression, controlling for the socio-demographic variables significantly associated with the outcome at the bivariate level. Model 1 assesses the multivariate association between parental warmth and affection and child-reported depression, indicating no significant association between parental warmth and affection and child-reported depression. Several socio-demographic and economic variables remain significant in the multivariate model. Some of these variables were protective against depression – for example, children in households who had been in the settlement more than 5 years were significantly less likely to report depression (aOR: 0.38, 95%CI: .17, .85), whereas other variables were significantly associated with elevated symptoms of depression, for example, children in households with high HESPER scores were almost four times as likely to report higher depression symptoms (aOR: 3.79, 95% CI: 2.17, 6.61) compared to the ones with lower HESPER score.
Table 4. Multi-variate model of child-reported depression (controlling for variables significant at bi-variate level)
Model 2 assesses the multivariate association between caregiver-report of frequency of child-caregiver interactions and child-reported depression; frequency of interactions was no longer significantly associated with child-reported depression. Several covariates were protective against higher depression symptoms (years lived in settlement, number of children in household, education level), while other variables were significantly associated with increased odds of child-reported depression. For example, a higher HESPER score increases the odds of child-reported depression by four times (aOR: 4.12, 95% CI: 2.36, 7.14) compared to a lower reported HESPER score. In Model 3, PSE and child-reported depression, results indicate that higher PSE is protective against child-reported depression (aOR: 0.51, 95% CI: .29, .90). In this model, other variables that are significantly associated with child-reported depression are the number of children in a household, financial difficulties, HESPER score, child-caregiver relationship, education level and child-reported limitations status. Across the three models assessing the child-reported depression outcome, HESPER, child-caregiver relationship and child-reported limitations were all consistently significantly associated with elevated odds of child-reported depression.
Table 5 displays caregiver behaviors and characteristics significantly associated at the bivariate level with caregiver-reported child depression. Model 4 indicates that child-reported caregiver warmth and affection are not significantly associated with caregiver-reported child depression when controlling for serious financial difficulties and HESPER score. Model 5 assesses caregiver-reported parental warmth and affection and caregiver-reported child depression, displaying significant associations between higher levels of warmth and affection and reduced odds of child depression, for example, high warmth and affection compared to low warmth and affection was associated with 80% reduced odds of caregiver-reported child depression (aOR: .20, 95% CI: .08, .49). No significant association were observed between caregiver-reported child depression and child-reported parental warmth and affection and caregiver-reported child depression. Both financial difficulties and HESPER score were also significantly associated with elevated odds of caregiver-reported child depression in both Models 4 and 5.
Table 5. Multi-variate models for caregiver-reported child depression (controlling for variables significant at bi-variate level)
Discussion
Our study explored the association between caregiver characteristics and behaviors and child depression among refugees living in a settlement in Uganda. Existing evidence examining how or if parenting practices influence child mental health outcomes in humanitarian settings is sparse, yet understanding these associations can inform the development and implementation of additional research to understand pathways, as well as programming and policy to support caregiver parenting and behaviors to reduce adverse child mental health outcomes. Our study is unique in exploring a number of caregiver characteristics and behaviors – warmth and affection, caregiver-child interactions and parenting self-efficacy, as well as the inclusion of caregiver-child dyads. In addition, we assessed both child and caregiver-reported depression. Our main findings are firstly, both parental warmth and affection and child-caregiver interaction were no longer significantly associated with child-reported depression when socio-demographic and economic variables were included. Secondly, multivariate analyses indicated differences in associations between child and caregiver-reported variables, indicating the utility of including both child and caregiver-reported exposures and outcomes. This also suggests the need for more comprehensive measurement in future research, given existing evidence primarily includes reports from either child or caregiver and not both. Thirdly, economic factors were prominent in significant associations with child depression, for reporting higher levels of serious problems according to the HESPER nearly quadrupled the odds of child depression. Finally, parenting self-efficacy was found to be significantly associated with lower child-reported depression.
Multivariate model results indicate differential associations between caregiver and child-reported parental warmth and affection; parental warmth and affection are significantly associated with reduced child depression when parental warmth and affection is reported by the caregiver, but not by the child. Concordance between child and caregiver reporting on these exposures and on depression is not currently explored in the literature on parenting and child mental health in LMICs. Research on concordance of child and parent reports on mental health outcomes indicates that high levels of concordance may be driven by the low prevalence of mental health symptoms in population-based samples in high-income contexts (Gray et al., Reference Gray, Scott, Lawrence and Thomas2021); discordance between child and caregiver reporting on both exposures and outcomes may be higher in the context of this study, and therefore may influence results and interpretations. Our findings show that caregiver, but not child-reported parental warmth and affection, is protective in the case of caregiver-reported depression. This may indicate that caregiver-reported behavior is subject to self-reporting bias, with caregivers potentially perceiving their behaviors in a more positive light than their children experience them. Psychological literature has addressed discordance in child and caregiver reporting of child mental health, finding that such discordance can be due to contextual factors, in that some child behaviors occur when parents are not present, or informant characteristics, in that female caregivers may report more accurately than male caregivers about their children’s mental health (Korelitz and Garber, Reference Korelitz and Garber2016). Discordance in reporting on parenting behaviors indicates that this is due to parental tendency to provide socially desirable responses (Fung and Lau, Reference Fung and Lau2010). In addition, in our study, across both child and caregiver-reported depression outcomes, child-reported parental warmth and affection is not a protective factor as was hypothesized and may be expected from the existing literature. This unexpected finding may shed light on validity issues in the existing evidence base, suggesting that caregiver self-reports of their own warmth and affection may be less reliable than children’s reports. Child-level reporting bias may also be present. Research on methodological approaches to accurate measurement of children’s mental health recommends inclusion of child as well as close adult-reporting, i.e., caregiver or teacher (Rescorla et al., Reference Rescorla, Ginzburg, Achenbach, Ivanova, Almqvist, Begovac and Verhulst2012). This study indicates that a multi-informant approach to identifying mental health problems and associated factors in a refugee population in Uganda is feasible and effective, providing important insights into how results may be impacted by who is reporting. Further analyses of this dataset can explore whether or how child-caregiver relationship quality influences concordance or discordance of reporting of child depression between caregivers and children (Maurizi et al., Reference Maurizi, Gershoff and Aber2012).
The results of multivariate models also indicate a strong and consistent association between economic factors – higher reporting of perceptions of financial difficulties and reporting of more severe problems as assessed by the HESPER – and child depression, both child and caregiver reported. While examining the association between economic factors and child depression was not one of the objectives of the study, the strength of these specific findings indicates that while parenting factors may be important targets for intervention and support, the structural economic environment strongly impacts – likely both directly and indirectly – children’s mental health. These findings confirm an extensive evidence base identifying pathways between economic stress, including food insecurity, lack of livelihood opportunities and financial strain, and poor mental health (Lund et al., Reference Lund, Brooke-Sumner, Baingana, Baron, Breuer, Chandra, Haushofer, Herrman, Jordans, Kieling, Medina-Mora, Morgan, Omigbodun, Tol, Patel and Saxena2018). The global refugee response is grossly underfunded, affecting the provision of basic health and social services, food rations and adequate education in refugee camps globally. In our study, more than half of respondents reported having serious financial difficulties and 85% reported either moderate or severe hunger. While many humanitarian agencies support and implement programs to promote refugee self-reliance, particularly in situations of prolonged displacement (Slaughter, Reference Slaughter2020), many refugees located in LMIC settings experience severe economic hardships, which affects their own and children’s mental health. Interventions to improve child mental health in refugee settings must consider the economic context in which caregivers are parenting, which can constrain and shape parenting practices, as well as address vulnerabilities and structural factors to effectively target the primary predictors of mental health of child refugees.
Parenting self-efficacy was found to be significantly associated with reduced symptoms of child-reported depression. Parenting self-efficacy and its associations with child well-being outcomes are widely researched in high-income contexts, and a limited number of studies have examined parenting self-efficacy amongst refugees resettled in high-income countries (Boruszak-Kiziukiewicz and Kmita, Reference Boruszak-Kiziukiewicz and Kmita2020; Eltanamly et al., Reference Eltanamly, Leijten, van Roekel, Mouton, Pluess and Overbeek2023a; Eltanamly et al., Reference Eltanamly, Leijten, van Roekel and Overbeek2023b). One study examined parenting-related measures in the context of Rohingya refugees residing in Malaysia, finding that caregiver emotional distress was associated with lower parenting self-efficacy (Shaw et al., Reference Shaw, Pillai, Yang and Saasa2021a). A randomized controlled trial, conducted in the same context, found that a parenting intervention effectively strengthened parenting self-efficacy, while measures of positive discipline and parenting were not influenced by the intervention (Shaw et al., Reference Shaw, Ward, Pillai, Ali and Karim2021b). While findings regarding parenting self-efficacy in refugee populations in LMIC contexts are sparse, our results indicate that interventions and approaches to understand and improve parenting self-efficacy may be effective in reducing poor child mental health. However, a systematic review of factors influencing parenting self-efficacy globally indicated that socio-cultural and contextual variables, including socio-economic status and family dysfunction, are predictors of parenting self-efficacy (Glatz et al., Reference Glatz, Lippold, Chung and Jensen2024) and another systematic review found that household income, perceived social support and parenting stress, amongst other factors, were associated with parenting self-efficacy (Fang et al., Reference Fang, Boelens, Windhorst, Raat and van Grieken2021). The findings from this study indicate that parenting self-efficacy may be an important target to improve caregiver and child mental health in refugee settings, however, structural influences that enable or hinder parenting self-efficacy must also be understood and accounted for.
These results should be interpreted in light of some limitations. The study employed a cross-sectional study design, and therefore child depression being the causal factor driving caregiver characteristics and behaviors cannot be ruled out. However, the recall periods of the different instruments – with a two-week recall period for symptoms of depression, compared to a general recall period for the caregiver characteristics and behaviors – may reduce the likelihood of reverse causality. The child-caregiver interaction measure was developed specifically for this study, and while it was pilot-tested, it did not undergo a full-validity study. However, the measure did exhibit adequate Cronbach’s alphas for both child and caregiver reports. The PARQ, parenting self-efficacy measure and MFQ have been tested and validated in Uganda in similar contexts and exhibited medium to high Cronbach’s alphas, indicating good internal consistency of the included items. A significant limitation is an interpretation of the results in light of the use of a cut-off that was not validated in this population and context. The lack of resources for a full validity study to determine the appropriate MFQ outcome cut-off in this context, entailed using a cut-off recommended for similar populations and contexts.
Conclusion
Parenting practices and behaviors have been identified as an important target for intervention to improve child well-being and mental health in humanitarian contexts. In this study conducted in a refugee setting in Uganda, analyses indicated that many of the protective factors hypothesized to be associated with child mental health were not significant when economic factors were taken into account. Significant associations were parenting self-efficacy and child-reported child depression and caregiver-reported parental warmth and affection, which was significantly associated with reduced caregiver-reported child depression. By assessing exposure and outcome measures through both caregiver and child reports, variations in associations were identified, highlighting the importance of taking multiple measures and further exploring why these differential associations exist.
Open peer review
To view the open peer review materials for this article, please visit http://doi.org/10.1017/gmh.2025.2.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/gmh.2025.2.
Data availability statement
Data are available on request from the authors. Data that support the findings are available upon reasonable request from the corresponding author.
Acknowledgments
We extend our sincere thanks to Professor Mark Jordans for his support throughout the study’s design, implementation and analysis phases. We are grateful to Rachel Nabirinde for the overall management of the study. We acknowledge the hard work of all data collectors involved in the study and are grateful to all caregivers and children who participated.
Author contribution
SRM, ES, RK and NU led the conceptualization of the study and methodology and supervised study implementation and data quality. NM led data curation and formal analysis. SN supported project administration and supervision. SRM and ES were involved in manuscript writing and development. NM, RK, SN and NU were involved in the revision of the manuscript. All authors read and approved the final manuscript.
Financial support
This study was funded under the PlayMatters project. The views represented within are not necessarily the views of the PlayMatters project, the International Rescue Committee, War Child Holland, Innovations for Poverty Action, the Behavioural Insights Team, or Plan International. This study was funded by LEGO Foundation in pursuit of its commitment to learning.
Competing interest statement
All authors declare that they have no conflicts of interest.
Ethics statement
The study received Institutional Review Board (IRB) approval from the Makerere University School of Health Sciences Research and Ethics Committee (MAKSHS-REC), under reference number MAKSHSREC 2023-620. Additionally, approval was granted by the Uganda National Council for Science and Technology (UNCST) under reference number SS1272ES.
Open access
Comments
Dear editor
Re: Submission to Global Mental Health, “Caregiver characteristics and behaviors: quantitative associations with child depression amongst refugees in Uganda.”
Please find attached our paper entitled, “Caregiver characteristics and behaviors: quantitative associations with child depression amongst refugees in Uganda” for consideration in Global Mental Health.
Refugee children globally experience extreme and on-going stressors, especially in low and middle-income settings. Caregiver relationships and behaviors, including caregiver warmth, parenting self-efficacy and positive relationships between caregivers and children, can act as protective factors against poor mental health outcomes for children and adolescents. Identifying which factors are protective, in which contexts and for which populations is important to inform policy and programmatic response to child mental health in refugee settings. In this quantitative study, based on a cross-sectional study amongst refugees in a refugee settlement in Uganda, we explore if caregiver-level factors (parental warmth and affection, positive child-caregiver interaction and parenting self-efficacy) are protective for symptoms of child depression.
In multivariate models, which included measures of economic stress, parental warmth and affection and child-caregiver interaction – which were significant at the bivariate level – were no longer significantly associated with child-reported depression. Parenting self-efficacy was found to be significantly associated with lower child-reported depression. These findings indicate the need to examine and explore how or if caregiver-level factors, such as warmth and parenting self-efficacy, operate in the context of chronic stress. We discuss the implications in terms of policy and programming for refugee children in this context.
This is the first manuscript is distinct from any other publications from the lead or other co-authors, and is not under consideration with any other journals.
Please do not hesitate to be in touch with any queries
Sarah Meyer