Copyright 2008 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2008
To examine the level of depressive symptoms and their predictors in youth from one region of Rwanda who function as heads of household (ie, those responsible for caring for other children) and care for younger orphans.
Four adjoining districts in Gigonkoro, an impoverished rural province in southwestern Rwanda.
Trained interviewers met with the eldest member of each household (n = 539) in which a youth 24 years old or younger was caring for 1 child or more.
Serving as a youth head of household.
Main Outcome Measures
Rates and severity of depressive symptoms using the Center for Epidemiologic Studies Depression scale; measures of grief, adult support, social marginalization, and sociodemographic factors using scales developed for this study.
Of the 539 youth heads of household, 77% were subsistence farmers and only 7% had attended school for 6 years or more. Almost half (44%) reported eating only 1 meal a day in the last week, and 80% rated their health as fair or poor. The mean score on the Center for Epidemiologic Studies Depression scale was 24.4, exceeding the most conservative published cutoff score for adolescents. Multivariate analysis revealed that reports of depressive symptoms that exceeded the clinical cutoff were associated with having 3 basic household assets or fewer, such as a mattress and a spare set of clothes (odds ratio [OR], 1.69; 95% confidence interval [CI], 1.06-2.70), eating less than 1 meal per day (OR, 1.68; 95% CI, 1.09-2.60), reporting fair health (OR, 1.32; 95% CI, 0.76-2.29) or poor health (OR, 2.33; 95% CI, 1.17-4.64), endorsing high levels of grief (OR, 2.67; 95% CI, 1.73-4.13), having at least 1 parent die in the genocide as opposed to all other causes of parental death (OR, 1.83; 95% CI, 1.10-3.04), and not having a close friend (OR, 1.91; 95% CI, 1.17-3.12). There was an interaction between marginalization from the community and alcohol use; youth who were highly marginalized and did not drink alcohol were more than 3 times more likely to report symptoms of depression (OR, 3.07; 95% CI, 1.73-5.42). When models were constructed by grouping theoretically related variables into blocks and controlling for other blocks, the emotional status block of variables (grief and marginalization) accounted for the most variance in depressive symptoms.
Orphaned youth who head households in rural Rwanda face many challenges and report high rates of depressive symptoms. Interventions designed to go beyond improving food security and increasing household assets may be needed to reduce social isolation of youth heads of household. The effect of head-of-household depressive symptoms on other children living in youth-headed households is unknown.
In 2005, it was estimated that 290 000 children younger than 18 years in Rwanda were orphans, one of the highest numbers worldwide of children who have lost both parents.1 The combined effects of the 1994 genocide, in which approximately 10% of the population was killed, and the human immunodeficiency virus (HIV) epidemic make Rwanda unique.
Most African orphans have been absorbed into informal fostering systems.2,3 Such systems, however, are increasingly overwhelmed, and many orphans either head households or live on the streets.4- 7 The demographic makeup of such households and the psychological status of orphans caring for other children are not well characterized.8 Furthermore, the number of youth-headed households (defined for this study as households headed by unmarried individuals younger than 24 years) will increase with time in areas affected by HIV/AIDS.9- 11
Depressive symptoms in heads of household are of particular concern because of the association between poor developmental outcomes for children and caregiver depression.12 Across cultures, depression affects interactive patterns between the caregiver and the child, disturbing the development of children's emotion regulation.13- 15 Defining and measuring depression across cultures, however, requires consideration of how culture and language might affect symptom expression.16
Studies suggest that valid and reliable surveys for depression can be constructed using local language, though an extensive process of piloting, adapting, and translating western surveys for use in nonwestern settings, is necessary.17- 19 A primary objective of the present study was to examine the level of depressive symptoms and their predictors in youth from one region of Rwanda who function as heads of household and care for younger orphans.
In 2004, a survey was conducted in the former Gigonkoro province, a rural province in southwestern Rwanda, selected because it is particularly impoverished. Before the survey, institutional review board approval from the health sciences panel at Tulane University and from a national review panel in Rwanda was received. Youth-headed households are common in southwestern Rwanda, and in 2001, an international nongovernmental organization had begun to target the more than 700 households in this region that community leaders identified as being headed by youth; the survey was designed to guide program expansion. To be included in this study, youth heads of household had to have little or no contact with their parents (ie, be functioning as orphans).
During the course of the survey field work, 41 of the homes originally identified were found not classifiable as youth-headed households (ie, parent had returned, youth had been adopted, or other adult caregiver was present) and, thus, were ineligible; 115 of the households could not be located; most youth heads of household were reported to have moved for work, marriage, or boarding school. Only 2 eligible respondents (<1%) refused to participate. The final sample (N = 539) included those aged 24 years or younger who were caring for 1 or more other household members.
All youth heads of household were interviewed in their homes for approximately 1 hour. A small bag of household items was provided as a token of appreciation to interviewees. Thirty-eight youth younger than 25 years who reported attempting suicide or always thinking about suicide in the 2 months before the survey were immediately referred to local mental health authorities for further evaluation; these youth were included as interview participants.
The survey was designed to assist with measuring intervention effect on youth psychosocial functioning. Scales measuring symptoms of grief, depression, adult support, and marginalization from the community in addition to questions about demographic factors, general health and functioning, HIV knowledge, vulnerability/protection, sexual behaviors, and risk behaviors were included.
The development of the survey instrument was informed and refined by a multistep process detailed elsewhere.8 Focus groups and free listing with orphaned youth helped select appropriate instruments.20 Translation and backtranslation preceded review of survey items by a technical committee of youth and professionals who attended to the context, cultural, and linguistic relevance of the questions. The survey was then pretested at a local street-children's center and then piloted with 79 youth heads of households living adjacent to the survey area.
The Center for Epidemiologic Studies Depression (CES-D)21 scale was incorporated into the survey with no substantive changes. Local terms equivalent to the CES-D 4-point scale were developed. The CES-D measures affective, somatic, and cognitive components of depression and has been widely used as a screening tool, including in adolescents and high-risk groups.22,23 The Cronbach α for the CES-D in this study was 0.86.
Recent findings suggest that the distribution of CES-D scores is different for older adults than for adolescents and young adults.24 For this analysis, we used the scale as continuous and also used a cutoff score of 24 to identify those youth with moderate to severe symptoms.25 The cutoff score of 24 improves case detection in this age group,26 although, to our knowledge, there are no published studies of youth from Rwanda.
We developed several scales using our focus group and piloting to inform each scale. Principal factor analysis ensured that each scale was unidimensional. Respondents indicated their level of agreement from “strongly agree” to “strongly disagree,” with “don't know/uncertain” scored in the middle. We dichotomized each scale at the median to facilitate the estimation of odds ratios.
An adult support scale was formed from 4 items (Cronbach α, .85). Items included whether youth had an adult in their life whom they could trust to give them advice and guidance or an adult who comforts them.
A marginalization scale was formed to assess the degree to which youth felt isolated from the surrounding community (Cronbach α, .77).27 Respondents indicated their level of agreement with 6 items including whether they thought that no one cared about them or that people in the community would rather hurt than help them.
Focus group data confirmed that ongoing grief about parental loss was important. A 7-item grief scale (Cronbach α, .66) was created for this study, including questions such as whether youth often thought about their lost loved one(s) and whether the loss of their loved one(s) had led to residual anger, loss of faith in God, or loss of confidence in others.
We created 2 binary item scales. The first was an assets scale that indicates whether the head of household owns a blanket and/or a spare set of clothes or shoes and whether the household has a latrine, a light source (oil lamp or better), or a mattress. The second was an exposure to maltreatment scale, developed from our focus group data, that included having been beaten, not being paid for work, and experiencing damage to property or attempts to take land or possessions. We used item analysis to determine that the scales were unidimensional and dichotomized these scales using a median split.
Other variables empirically or theoretically linked to depressive symptoms were included in the analysis, including health status, cause of parental death, consumption of alcohol, experience of forced sex, sexual risk-taking behaviors, and sociodemographic data. Health status consisted of self-reported health (excellent, good, fair, or poor). The parental death variable was coded yes if either parent was reported to have died in the genocide. Sociodemographic variables included age, sex, assets scale, number of meals per day, area of residence, and educational achievement level.
Analyses were conducted using commercially available software (STATA version 8; StataCorp LP, College Station, Texas). The depressive symptoms score from the CES-D scale was the dependent variable. We decided a priori to group potential predictors of depressive symptoms into the following blocks: health status, social support, cause of parental death, emotional status, exposure to maltreatment, lifestyle behaviors, and sociodemographic data. Social support included the adult support scale plus a question about contact with relatives and a question about whether the head of household believes that he or she has a close friend. Emotional status included the grief and marginalization scales. Exposure to maltreatment included experience of forced sex added to positive responses on the maltreatment scale. Lifestyle behaviors included ever having had sex and drinking alcohol.
After computing frequency distributions, we explored the bivariate association of each predictor with depression using χ2 tests. Multivariate logistic regression was conducted to estimate odds ratios, 95% confidence intervals, and P values.
To determine the amount of variance in depression that blocks of variables explain and to develop a multivariate model, we calculated pseudo-R2 using the MacKelvey and Zavoina measure.28 We determined the amount of variance in depression that each block of variables explained alone and after controlling for health status, cause of parental death, and sociodemographic data.
We hypothesized that the relationships between the predictors and depression might differ by sex and initially ran the regression models separately by sex. Sex differences were, however, minimal, and we subsequently combined the sample. Nevertheless, we tested for significant interaction between each predictor and sex in the bivariate models and in the final model. There was only 1 significant interaction involving sex: sex interacted with contact with relatives in predicting depressive symptoms. We included this interaction in all of the models that include the social support block of variables. We also included an interaction between marginalization and use of alcohol. Psuedo-R2 estimates are based on the full main effects models only.
Table 1 gives the characteristics of the 539 households and youth heads of households living with 1 or more other household members. There were about equal numbers of males and females heading the households; their mean age was 20 years. Seventy-seven percent were subsistence farmers; 43% had attended primary school for 3 years or less, and only 7% had completed more than 6 years of primary school. Most respondents reported having a blanket, a latrine, and a spare set of clothes; only 10.8% owned a mattress. Forty-four percent reported eating only 1 meal per day, and almost 80% rated their health as fair or poor.
Youth in 71.4% of households reported that both parents were dead, 26.2% reported that 1 parent was dead, and 2.4% reported that both parents were alive. In cases where 1 or both parents were alive, the parents were in jail or living elsewhere or their whereabouts or status were unknown.
Almost one-fourth of the participants who reported that a parent had died indicated that 1 or both of them were killed in the genocide. Causes of death other than genocide included poisoning, and AIDS and other illnesses. It was impossible to verify youth reports of parental cause of death, and reports of poisoning represent a more socially accepted way of reporting death resulting from AIDS.
Grief clearly had a role in how youth viewed the world around them; more than half had high levels of grief (Table 2). For example, 64% of participants stated that they had lost confidence in people, and more than 40% endorsed feeling that life was meaningless or that they had lost faith in God since the death(s) of their parent(s).
The youth in this sample reported feeling stigmatized and socially isolated. For example, 76% of participants agreed or strongly agreed with the statement that the community rejects orphans. Only 26% strongly agreed that they had at least 1 friend. Nevertheless, the youth were not completely without support; 66% named 3 sources of adult support, and most of the respondents (82%) reported having contact with relatives.
Most youth (70.5%) reported some form of maltreatment such as being beaten; experiencing attempts to have land or possessions taken, or property damage; and not being paid for work. Risk behavior was common, with 32% reporting ever having had sex and 49% reporting alcohol use. Fourteen percent of the sample (n = 63 females and 13 males) reported experiencing forced sex.
Depressive symptoms were common (Table 2). The mean depression score was 24.4, which is above the most conservative published CES-D cutoff score of 24 for adolescents.24 Put another way, 53% of the sample screened positive for depression, though the range of CES-D scores (0-55) was considerable.
At univariate analysis, we found several variables to be statistically associated with depressive symptoms (Table 3). Multivariate analysis was used to further explore these findings, including possible interactions.
For the regression models, we grouped predictors as follows: health status, social support, demographic data, cause of parental death, emotional status (grief and marginalization from the community), experience of maltreatment, and lifestyle (ever had sex or used alcohol). The groups were entered in the model one at a time. Table 3 also gives the adjusted odds ratios from the final regression model.
Most of the variables significantly associated with depressive symptoms at univariate analysis remained significant in the regression model. Youth with 3 household assets or fewer were more likely to score above the cutoff for depression, as were those reporting eating fewer than 1 meal per day. Youth with fair or poor health were more likely to be depressed, though the relationship between health status and depressive symptoms was somewhat diminished in the final model.
High levels of grief were significantly associated with depressive symptoms above the screening cutoff, and those who reported that 1 or both parents had died in the genocide (as compared with all other causes of parental death) were almost twice as likely to score above the screening cutoff as well. Those who said they did not have a close friend were more likely to meet the criteria for depression.
Two interaction terms were included in the model: seeing relatives by sex and marginalization by alcohol use. Though the interaction of sex and depressive symptoms did not reach statistical significance, male respondents who did not have contact with their relatives were more likely to report high levels of depressive symptoms, whereas for female respondents, not having contact with relatives was protective. Among youth reporting high levels of marginalization from the community, those who did not drink alcohol were more than 3 times more likely to report symptoms of depression, which suggests that drinking is protective for marginalized youth. In contrast, there was no significant relationship between depressive symptoms and alcohol use for those reporting low marginalization. In sum, having few household assets and little food, reporting fair or poor health, having high levels of grief and marginalization, having lost a parent during the genocide, and having no close friend predicted depressive symptoms in the regression model.
Rwandan youth aged 12 to 24 years who care for younger children have severe challenges. Most dropped out of school in the primary (pre–high school) years, most experience food insecurity as subsistence farmers, and many report poor health, endorse feeling stigmatized, and are socially isolated. More than 70% report instances of maltreatment by others in the community, and 1 in 6 female respondents report forced sex. Grief caused by parental loss further complicates the risk profile for most respondents.
More than half of the 539 youth we interviewed were above the most conservative age-specific screening cutoff on the CES-D, a score in other cultures consistent with moderate to severe symptoms.25 Our data are consonant with our expectations about youth who head households in postgenocide Rwanda. A 1999 study conducted in Rwanda revealed that 15% of a representative group of postgenocide adults met strict criteria for depression using a locally developed and tested survey,18 though more representative surveys of adolescents and young adults are unavailable.
Orphan status in Rwanda, as in many parts of Africa, is a marker for risk conditions associated with depression. In Uganda, for example, orphan status was the only significant predictor of depressive symptoms in youth aged 11 to 14 years, with orphans more than 6 times as likely to endorse high levels of depressive symptoms than nonorphaned children.29 Orphan status also independently predicted internalizing symptoms in Tanzanian children.30Orphan status remained associated with depression after controlling for other factors in a large study comparing orphans with other vulnerable children in Zimbabwe.31 In sum, compared with other disadvantaged children in Africa, orphans are particularly likely to report symptoms of anxiety and depression.32
The predictors of depressive symptoms in the orphaned youth we interviewed were complex. We expected health status to be associated with depressive symptoms; in the final model, those with fair or poor health reported higher levels of depressive symptoms. Our finding linking depressive symptoms and having only 1 meal per day or having few household assets is less well established. Hunger has, however, been linked to depression and anxiety in children in less impoverished settings,33 and going to bed hungry was associated with depressive symptoms in orphaned children in Tanzania.30
In our sample, grief and social marginalization accounted for the largest percentage of the variance in depressive symptoms both when sociodemographic data were controlled for and when they were not. Grief is confounded with orphan status. Our data suggest that many orphans experience ongoing grief symptoms years after the loss of loved ones. In the Rwandan context, unresolved grief may be linked to exposure to trauma, which interferes with the process of grieving. For Rwandan orphans, traumatic grief, marked by an admixture of grief and posttraumatic symptoms,34 is likely prevalent. Symptoms such as intrusive re-experiencing and hyperarousal make it difficult for the individual with traumatic grief to traverse the typical grieving process.35 One clue that the interplay between trauma and grief was important in this sample is that youth who reported parental death during the genocide were almost twice as likely to report more depressive symptoms, even when accounting for other factors. Children may have particular problems making sense of violent parental death.36,37 Because our survey was conducted just before the 10th anniversary of the genocide, for which many communities in Rwanda had large-scale memorial services, it is possible that the timing of the survey may have elicited anniversary reactions and heightened grief, especially in youth who lost a parent during the genocide. There is evidence that distressed individuals flood local clinics during April when yearly public memorials occur (J.N.; personal observation; February 7, 2007).
In addition to grief, not having a close friend was strongly associated with depressive symptoms in this study. The link between social support and depression in collectivist societies is important,38 and orphaned adolescents and young adults are particularly affected by social isolation.39 The few comparative studies in Africa suggest that orphan status is a predictor of perceptions of marginalization; orphans tend to report either fewer or more negative interactions with both peers40 and adults.41,42 Given the cross-sectional nature of our data, however, it is not possible to determine whether not having a close friend precipitated depressive symptoms or resulted from them.
When we looked beyond close peer relationships, we found that perceptions of marginalization from the wider community interacted with alcohol use. Those youth who reported high levels of marginalization and no alcohol use also reported high levels of depressive symptoms. Much of the available alcohol in rural Rwanda is locally-produced and consumed in village celebrations and ceremonies; orphans with few resources are unlikely to be able to afford to drink unless they are included in such events. The most marginalized youth may be those most likely to be excluded from village events, which would mean that these youth would have few opportunities to drink. Social alcohol use in Rwanda is a proxy for social inclusion. We had only 1 general question about alcohol use, however, and so captured both common social use and more uncommon problem use. These marginalized youth may also either avoid social contact (common when depressed) or be shunned and, therefore, experience depressive symptoms. Furthermore, depressed youth may drink to self-medicate, and more data on the relationship between alcohol use and depressive symptoms in similar contexts are necessary.
Neither sex nor maltreatment experiences were significantly associated with reporting high levels of depressive symptoms. It may be that males, who are culturally less likely to have to care for other children, experience such responsibility as highly stressful. As for maltreatment, our survey included questions ranging from more common harassment (eg, threats to steal property and not being paid for work) to being beaten to uncommon experiences such as sexual assault. Relative to grief, poor health, lack of food, and not having a close friend, however, it seems that these kinds of experiences do not significantly contribute to depressive symptoms in the harsh context in which these youth heads of household live.
Our survey captured a significant amount of the variance in depressive symptoms among youth heads of household; other factors, however, seem to have a role, and both more qualitative and survey research is necessary. We focused on youth who cared for younger children (primarily siblings and offspring) because of data suggesting that caregiver depressive symptoms have important negative developmental implications for those children.12 Our data suggest that younger children living in homes headed by orphaned youth are at grave risk. Potential genetic effects (most of the heads of household are first-degree relatives) may compound the effect of the interactive effects of being cared for by a person with depressive symptoms.14 Given the psychosocial context, it seems unlikely that depressive symptoms would remit (at least not without recurring) in most of those already affected. Furthermore, the kind of social marginalization the heads of household endorsed might also directly affect younger members of the family.
There are several important limitations to this study. Our sample cannot be considered representative of orphaned youth in other areas or countries. Compounding economic and social risks combined with high levels of exposure to trauma make it difficult to generalize from data gathered in postgenocide Rwanda. Furthermore, though youth-headed households are an increasingly common phenomenon as the HIV/AIDS epidemic results in more children being orphaned, both broad cultural factors and unique local factors influence how orphans are or are not absorbed into communities.8,9
The cross-sectional nature of our data is another important limitation. Depressive symptoms are important, but whether, for example, they result from marginalization or make it more likely for youth to perceive that they are marginalized cannot be determined. Our reliance on self-report further complicates matters. Recall bias may have increased the likelihood that respondents endorsed grief, poor health, depressive symptoms, or marginalization. On the other hand, it has been argued that the cultural proscription on open displays of emotion in Rwanda might actually limit endorsement of symptoms of psychological distress among interviewees.43
Another major limitation of this study is the lack of a comparison group. Without a comparison group, we cannot be sure whether nonorphaned youth living in this social context are any less distressed. A challenge for research with orphans is to identify a group that is comparable. There is mounting evidence to suggest that orphans are uniquely at risk for adverse outcomes.
Several of the scales we used in our survey (those capturing grief, degree of adult support, and perceived marginalization) were developed and piloted for this study and, thus, need additional validation in other populations. Some of the scales had a Cronbach α reliability of less than 0.80, and refinement of these scales in other settings is indicated.
Interviews of a large sample of impoverished and orphaned youth younger than 24 years who care for dependents in rural Rwanda revealed high levels of depressive symptoms. Hunger, grief, few assets, poor health status, and indices of social marginalization were associated with more depressive symptoms in this sample. Ten years after the Rwandan genocide and in the midst of the HIV/AIDS epidemic, the effects of poverty and social disruption on the most vulnerable youth in Rwanda are evident. The effect of caregiver depression on younger children living in youth-headed households is not yet known. Further study of orphans and vulnerable children in countries such as Rwanda, in particular, studies that inform large-scale interventions, are necessary if the next generation of youth is to thrive.
Correspondence: Neil W. Boris, MD, Department of Psychiatry and Neurology, Tulane University School of Medicine, 1440 Canal St, Box TB-52, New Orleans, LA 70112 (firstname.lastname@example.org).
Accepted for Publication: February 1, 2008.
Author Contributions: Dr Brown had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Boris, Brown, Thurman, Snider, and Ntaganira. Acquisition of data: Boris, Brown, Thurman, Snider, Ntaganira, and Nyirazinyoye. Analysis and interpretation of data: Boris, Brown, Thurman, Rice, Snider, Ntaganira, and Nyirazinyoye. Drafting of the manuscript: Boris, Brown, and Snider. Critical revision of the manuscript for important intellectual content: Boris, Brown, Thurman, Rice, Snider, Ntaganira, and Nyirazinyoye. Statistical analysis: Rice. Obtained funding: Boris, Brown, and Snider. Administrative, technical, and material support: Boris, Brown, Thurman, Rice, Snider, Ntaganira, and Nyirazinyoye. Study supervision: Brown, Snider, Ntaganira, and Nyirazinyoye.
Financial Disclosure: None reported.
Funding/Support: This study was supported by the Population Council Horizons Program, which was funded by the US Agency for International Development.
Role of the Sponsor: The funder had a role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, and approval of the manuscript.
Additional Contributions: Our partners at World Vision International, especially Eleazar Mugirira, were especially helpful in data gathering. We thank the youth who graciously participated in the initial focus groups that informed our survey and in the survey itself.
Boris NW, Brown LA, Thurman TR, Rice JC, Snider LM, Ntaganira J, Nyirazinyoye LN. Depressive Symptoms in Youth Heads of Household in RwandaCorrelates and Implications for Intervention. Arch Pediatr Adolesc Med. 2008;162(9):836-843. doi:10.1001/archpedi.162.9.836